Technical ReportPDF Available

Artificial Intelligence in Local News: A survey of US newsrooms' AI readiness

  • Associated Press
  • The Associated Press


The Associated Press presents a groundbreaking report that provides important insights into local news outlets’ understanding of artificial intelligence and their readiness to use AI to meet their journalism and business needs. The report is based on survey results from nearly 200 newsrooms across all 50 states and more than two dozen in-depth interviews with local news leaders. Print, radio, television and digital outlets are represented, as well as commercial and nonprofit operations. In addition to demonstrating a readiness to adopt AI and automation technologies, the report outlines what local news providers need to drive technological innovation.
Intelligence in
Local News
A survey of US newsrooms’ AI readiness
By Aimee Rinehart and Ernest Kung
The Associated Press
March 2022
  :      2  56
Introduction 3
Knight Foundation grant 8
Research methods 9
Jargon surrounding AI 11
Findings 13
Analysis and conclusions 36
Acknowledgments 44
Appendix 50
  :      3  56
Artificial intelligence, once a buzzword that might pop up in a
technology news story every so often, has found a home in the news
industry over the past decade. AI technologies are under the hood of
many news production and presentation processes, such as the
automation of brief text stories from data feeds and the
personalization of web and mobile news displays. AI technologies
also augment the work of journalists around the world by helping
them dig out information available on the internet or buried deep in
large collections of documents.
The Associated Press is among the news organizations that were
early adopters of AI technology. In one of our most significant
moves, we introduced the use of a technology called “natural
language generation” in 2014 to automate the production of
thousands of quarterly corporate earnings stories — straight from
financial data feeds without human intervention. AP now employs
AI technologies to address a variety of needs: to get early warnings
of breaking news events, generate short summaries from longer
narrative text, classify and apply digital metadata to news content
and transcribe audio from video in real time — among many other
use cases. Across the board, these technologies have been employed
to augment the work of our journalists and to minimize, and
sometimes eliminate, time-consuming chores that bog down the
daily news process.
We’re not alone in taking advantage of AI. The London School of
Economics and Political Science’s Media and Communications
department published a report in November 2019 laying out the
expanding scope of AI adoption by news providers around the world.
Oxford University’s Reuters Institute for the Study of Journalism
also has examined AI in its accounts in the past several years. Its
most recent report notes that AI technologies have become standard
for large, national and international publishers. AI “can no longer be
regarded as ‘next generation’ technologies but are fast becoming a
core part of a modern news operation at every level — from
newsgathering and production right through to distribution,” the
Oxford authors write.
The trend has clearly taken off, but as the Reuters Institute study
suggests, this phenomenon has so far been centered among “large”
  :      4  56
publishers. Even as we at AP have told our own story and observed
other big news organizations tell theirs over the past several years,
our suspicion has been that awareness and use of AI were not
trickling down to “smaller” news providers.
That perceived “gap” in AI knowledge and adoption prompted AP
to pursue a Local News AI initiative that has been funded by the
John S. and James L. Knight Foundation to understand the true
state of the art at the local level and to advance smart, ethical uses
of AI technology that can help local news providers augment their
journalism, achieve efficiencies and sustain their businesses.
As a first step in what will be a two-year project, we developed a
survey — which we called an “AI readiness scorecard” — to solicit
interest in our project and assess the current state of AI awareness
among local news providers. Leaders of 192 local newsrooms
completed our scorecard because they see AI as potentially helpful
to their journalists and the communities they serve. Twenty-five
newsrooms also sat for in-depth interviews recorded on Zoom.
Scorecards came from all 50 states, D.C., Puerto Rico and Guam;
from print, radio, TV and digital-only formats; and commercial as
well as nonprofit operations. The scorecard and interview questions
covered newsgathering, production, distribution and the business
side of news operations.
In general, we confirmed our suspicion about the AI gap between
large and small organizations. In the cohort we surveyed, AI
technologies were not in wide use. But the reasons largely came
down to a lack of any cushion required to experiment.
Experimentation with AI and automation technologies requires the
capacity of staff, a strong foundation with current technology, time
and money. Many newsrooms spoke of staff turnover, frequently
losing the one person who had been the driver of innovation. Others
spoke of being unable to spare one of a handful of reporters to take a
month to learn how a speculative technology might enhance, and
not distract, from their other duties.
What’s more, current technology in local newsrooms is patchy
and often does not sync. Adding still another layer to an already
cumbersome technology stack can be out of the question for many
newsrooms. Innovation requires customizing off-the-shelf products,
building new products and/or partnering with vendors or
universities. In all instances, someone needs to understand how
to maintain these products. Someone needs to take the lead. And,
then, some experiments ultimately fail, and that failure can sink a
local newsroom.
  :      5  56
Despite these odds, we encountered inspiring stories among
newsrooms of various staff sizes and technical capabilities. Some
newsrooms have built workarounds to technology gaps. In others,
tech-minded journalists have created basic automations to help their
newsroom contend with an onslaught of information. In the most
promising cases, some local newsrooms have devoted special teams
to technology innovation.
In this report, we share the full scope of our scorecard results and
interview highlights. Included are snapshots of local news leaders
and their unique circumstances. These profiles also serve as a
reminder that humans drive innovation, not robots. Some news
leaders expressed concern about journalists being put out to pasture
for “robot journalism.” They worried about being left behind by yet
another mysterious journalism trend that feels out of reach for their
own news operations. They asked tough questions about the ethics
of using AI to augment journalism.
With this report, we aim to level-set the AI playing field at the local
level with some basic information about the current state of the art
and what these survey respondents say they need to put AI to use in
their operations. We will move on in the next phase of this Knight-
funded project to offer a free online curriculum based on these
findings. Following that, we will undertake a limited number of
hands-on engagements with local outlets interested in developing AI
applications of their own.
At this stage, it’s all about getting a deeper appreciation of the needs
of local news outlets and a deeper understanding of how AI-based
solutions might be able to help. With AI technology emerging as the
next major step in the computerization of the newsroom, we aim to
make it accessible to all.
  :      6  56
“The acceleration of information
and the multitude of software,
platforms and tools to measure
it and apply it has become
staggering. There are few people
in any operation with the ability
to maximize all of it and few
tools that harness the power
of disparate channels to reduce,
instead of increase, information
overload or automate routine
  :      7  56
“My biggest fear is: Can we move
fast enough to keep up with new
technologies? I know things are
moving very quickly in AI, and I
ask myself organizationally, are
we going to be nimble enough to
be able to take advantage of
opportunities or to identify
needs and then be able to see
how a technology like that could
help meet those needs in a
timely fashion?”
  :      8  56
Knight Foundation grant
The John S. and James L. Knight Foundation granted $3 million
over two years in May 2021 to four organizations: The Associated
Press, Brown Institute at Columbia, NYC Media Lab and Partnership
on AI. The grant’s goal is to expand local U.S. news organizations’
adoption of AI tools and automation technologies in ways that
support their businesses. The Knight initiative is based on years-
long analysis and observations by Knight on how AI impacts
journalism, working with John Keefe, Jeremy Merrill and Youyou
AP’s work has three phases:
1. Develop an industrywide benchmark for AI readiness across
editorial and business lines, with a scorecard and interviews with
newsroom leaders.
2. Create a free, online training based on learning gaps discovered
from the scorecard and raise awareness of AI and its potential
impact across news operations.
3. Consult with newsrooms to identify opportunities for
automation and AI, implement and assess those initiatives.
  :      9  56
Research methods
Who was recruited
To establish a benchmark for AI readiness, we sought information
from local newsroom leaders about their operations. We define
“local” broadly by considering audiences served, ownership, location
and audience size. Note: “local newsroom” and “small newsroom”
are not synonymous in this report. Some local newsrooms have large
staffs, while some newsrooms consider themselves regional and have
relatively small teams.
How we recruited
We emailed our scorecard pitch to executive editors, managing
editors, news directors, publishers and station managers in AP’s
cooperative membership. We also elicited help from various
journalism industry partners to spread the word to newsroom
leaders beyond the AP membership; a list of those partners appears
in the acknowledgments. We followed up the emails with phone and
video calls to explain the initiative.
The scorecard
The findings of LSE’s November 2019 report, “New powers, new
responsibilities. A global survey of journalism and articial
intelligence,” have informed our work. We extrapolated from LSE’s
report the questions they asked of newsrooms to build AP’s
scorecard questions. AP’s Local News AI initiative is meant to build
upon the work started at LSE, explore similar topics at the level of
U.S. local newsrooms and inform other works and research for
international markets.
The scorecard is separated into five categories: general AI
knowledge, newsgathering, production, distribution and business-
side concerns. The 32-question scorecard includes statements
seeking responses on a scale of 1 to 5, where 1 represents “strongly
disagree,” 3 “neutral” and 5 “strongly agree.” Statements include:
“Our newsroom regularly uses AI in newsgathering,” and “We’re
interested in AI to simplify production operations.” There is also a
blank field in every section for automation wish lists. The scorecards
were collected from September through December 2021.
  :      10  56
The interviews
A subset of the newsrooms responding to the scorecard were invited
to participate in recorded Zoom interviews. Those interviews were
selected to balance format, geography, staff size and business type.
AP membership was not a requirement. The interviews were
conducted from October through December 2021.
Five students at Northwestern University’s Knight Lab conducted
the interviews with a digital interview guide. The question set was
developed as a collaboration between AP and the university. All
interview participants were asked the same basic set of questions,
and the student interviewers also asked follow-up questions.
  :      11  56
Jargon surrounding AI
AI is short for artificial intelligence, and it generally refers to a range
of technologies that are human-designed to automate, accelerate or
extend the human work required for specific tasks. Many
applications can be used for journalism that include AI. Some
applications involve simple automation of time-consuming tasks.
Others are more advanced, helping journalists with tasks that they
couldn’t reasonably accomplish with human effort alone.
In general:
AI helps to process data.
AI does not have a mind of its own.
AI depends on data that humans feed it.
AI produces results from human-fed information.
There are multiple subcategories within AI. In general, four types of
AI technologies stand out in journalism: machine learning, natural
language generation, natural language processing and computer
vision. We lean here on AP’s previously published report, “A Guide
for Newsrooms in the Age of Smart Machines,” for definitions.
  (ML) enables an application to adjust without
being told what to do once humans feed it a lot of data. ML takes a
complex idea and breaks it into a series of smaller, more
approachable tasks that lead to a designated endpoint. But for
machines to learn, they need to be taught by humans, again and
again to perfect the output. This is sometimes referred to as having
a “human in the loop.”
   (NLG) turns structured data into a
digestible written narrative. Structured data means that the data is
organized in a predictable, formatted way to render similar results
every time. AP had success in 2014 in automating business earnings
reports and later extended that to certain kinds of sports reports.
  :      12  56
   (NLP) can help journalists sift
through and draw insights from large collections of data or
documents, such as FOIA requests. NLP also includes
summarization technology, which AP and its technology partner
Agolo have used to transform long-form text stories into short
broadcast scripts.
  can help extract insights from pictures and video.
Image recognition was used to help AP journalists identify sea
vessels as part of the 2016 Pulitzer Prize-winning series on abuses in
the seafood industry. Further, AP uses computer vision to help tag
photos with descriptive information as they are processed, which
helps with searching for them later.
Programs that include AI are already common in many local
newsrooms. Examples include NewsWhip and Google Analytics for
understanding audience consumption, DocumentCloud and Google
Pinpoint for analyzing large collections of documents, Otter and
Trint for automated transcription. (A list of technologies in use by
local newsrooms that participated in this research is in the
What most newsrooms and their audiences need is the automation
of basic information, such as social media content and high school
sports scores. They need help managing information overload. While
automation does not always involve AI, it is often the first step
toward the adoption of more sophisticated AI applications.
The findings of this research cover this full gamut of technology
needs and have the potential to benefit newsrooms of any size.
  :      13  56
Let’s dive into the results of our scorecard with a look at overall AI
understanding and readiness by local U.S. newsrooms. We will then
cover four segments of the newsroom: newsgathering, distribution,
production and the business side. Each part describes what’s
happening in newsrooms and then highlights what could be helpful
to automate.
AP received 192 scorecards. While the scorecard remains open, the
data that informs this report is for scorecards submitted by Dec. 20,
2021. Results are split into print, radio, television and digital to
account for format differences.
Of the 192 scorecards, 187 were from individual local news outlets;
the others were group responses. The individual responses were
used for the quantitative results, where news managers responded to
statements on a scale of 1 to 5. We use median values to analyze
these numerical data and eliminate wide swings. The full results
with averages are available in the Appendix.
A total of 135 scorecards were used for qualitative results, including
the automation wish list where managers could write into a text box.
Fifty-two newsrooms did not specify anything on their wish lists.
We also measured how much AI is being used today across the
industry at the local level. On a scale of 0 to 6, where 0 represents no
AI usage, to 6 representing widespread usage, the median was 0 or 1
for all formats, indicating that little AI is being used in U.S. local
Our interviews were conducted with a broad range of newsrooms,
from format to geographic region.
  :      14  56
AI readiness and understanding
U.S. local newsroom managers we surveyed are condent that AI
can take on repetitive tasks to free up time and are concerned about
falling behind on AI. However, they are less condent in their
understanding of AI in journalism.
Scorecard ndings
We created a composite score to measure AI readiness,
understanding and usage. The scores were calculated on a scale of 24
to 120. Digital newsrooms had the highest composite scores with a
median of 82. The medians for the other three formats were lower,
with television newsrooms at 75, print 74 and radio 73.
Newsrooms rated the following statements on a scale of 1 to 5.
The bold scores on the left reect median values:
We have a good understanding of what AI is and how it relates to journalism.
3 News managers across all formats are neutral on this broad statement.
Were condent that AI can take on repetitive tasks to free up resources for more
substantive work.
4 Radio, TV and digital news leaders feel that AI can help increase eciency by
handing o specic tasks to automation.
3 However, print organizations have a neutral score.
Our organization has a solid strategy for AI that crosses all departments.
1, 2 With low scores across all formats, managers expressed concern that news
organizations are unprepared for AI adoption.
We feel ready for AI technologies in our operations.
2 Print, radio and TV managers expressed pessimism.
3 Digital managers gave higher scores but were neutral overall.
We are concerned about falling behind in AI.
4 Radio, TV and digital managers expressed concern.
3 Print managers were neutral.
We have the nancial resources to invest in AI.
3 In an industry with well-documented nancial challenges, survey participants
across all formats were neutral about having nancial resources to invest in
  :      15  56
We have people with AI skills in our organization.
3 News leaders in all formats were neutral about having people with AI skills in
their organizations.
We can allocate time to work on AI projects.
3 Print and radio leaders were neutral. Some managers said they weren’t sure
how much time they could spare for working on AI.
4 Overall, TV and digital leaders felt they could nd the time.
Interview top ndings
     . People need to have the final say
about what will be published. News leaders said they don’t want to
be in a position where they let AI post reports without humans
knowing what was posted. AI is best suited for data sifting, but
interpreting that data is a human task.
   . Some news leaders question if AI will
consider privacy and understand not to publish a person’s name in
certain situations. How AI handles gathering and storing sensitive
records, such as personal contact data, is vital to consider and needs
to be transparent across the news operation and to the public.
   . Many news operations said they
struggle with finding the most impactful problem to automate. They
also listed three requirements for adopting new technology: low
cost, low learning curve and low maintenance.
  . Many managers said automation would
free people to do more important things for the brand. They want to
ensure that automations make news operations more efficient
without introducing complications. However, we also heard
concerns from some newsrooms that ushering in AI has some
staffers worried about their jobs becoming obsolete.
 . Newsrooms we surveyed often rely on a single tech-
minded journalist. When that person moves on, it can leave a hole in
the newsroom’s ability to continue that same work.
 . News leaders spoke about the importance of
delivering quality news, and equally important, about the
communities they serve. Some feel they know their audience well
with two-way digital communications, primarily through the
website, app or social media comments.
  :      16  56
“We envision AI rounding out
coverage from rural areas within
our DMA, which is diverse,
and travel is not always possible
— either from a time standpoint
or logistically. AI could help
us service news deserts that
are emerging.”
  :      17  56
“We cover a lot of important
news that makes national
and international headlines,
so we want to be prepared
for when news hits our area.
AI could potentially help make
some of our newsgathering
and production easier, allowing
us more time to focus on
our content.”
  :      18  56
APs AI readiness scorecard reveals U.S. local newsrooms sur veyed
do not regularly use AI in newsgathering. Few have tried tools that
use AI. However, most indicated that theyd be willing to use
automation and AI if it helped to reduce workloads.
Some newsrooms said journalists use saved Google searches to get
alerts to beat-specic information and use third-party tools to get
alerts when a website has been updated. Other newsrooms employed
services like Dataminr, CrowdTangle and Chartbeat for social
listening. Most newsrooms interviewed said their reporting teams
manually check social media pages and groups, sift through city hall
meeting agendas, press releases, court records, etc., to find helpful
What’s happening now
Scorecard ndings
Newsrooms rated the following statements on a scale of 1 to 5.
The bold scores on the left reect median values:
Our newsroom regularly uses AI in newsgathering.
2 Print, radio and digital news managers reported little use of AI in
3 TV managers gave higher scores but were neutral overall.
We have a few people in the newsroom who have tried AI technologies for
2 Print, radio and digital leaders report having few people who have tried AI
for newsgathering.
3 TV leaders again gave higher scores but were neutral as a group.
Were interested in AI to potentially help reduce the workload for our journalists.
4, 5 While AI experience levels may be low, news managers are willing to learn
about AI to reduce newsgathering workloads. Managers in all formats gave
positive scores, with digital the highest.
Our journalists support exploring AI for their work.
3 Support among journalists for using AI with their work is potentially lower,
with print managers giving a neutral score.
4 Radio, TV and digital managers say there is support for AI in newsgathering.
  :      19  56
What could be useful
Based on the scorecard and interviews, the following automations
are what U.S. local newsrooms would like to have for newsgathering:
Transcription of interviews in audio and video formats of public and government
meetings, of published audio and video stories, and with the ability to handle
multiple languages.
Delivering meaningful story recommendations to online audiences.
Flagging and gathering social media content like trends, quotes from newsmakers,
ag and gather content on government websites, e.g., COVID-19 data, court records,
law enforcement records.
For investigative journalism, revealing patterns in crime; streamlining data-based
reporting of government records, home sales, cleaning datasets; election reporting,
informing editorial decisions and source audits.
Processing large sets of public records like campaign nance records, state
legislation, civil complaints, municipal budgets; in combination with text summari-
zation to help reporters.
Translating published stories into multiple languages and processing raw data in
various languages.
Bringing in scores from local high school teams, tracking team performance over
time and competition schedules.
Applying tags — people, things — to photos as they’re ingested to search for them
quickly later.
Sending and processing permissions requests from UGC sources and editing UGC
Moderating story idea submissions and questions; verifying tips.
Streamlining fact-checking processes, alerts to mis- and disinformation.
  :      20  56
“News decision-making is still
an art and a craft, but the art
in the craft can benefit greatly
from better data and more
information, and AI could help.”
“I used to have 12 reporters,
now I have five, so everything
is out of reach from what we
used to do. By pursuing
automation, we will be able
to focus human intellect
on more complicated stories.”
  :      21  56
“Lots of important stuff happens
at board and council meetings. If
there were AI that could take the
minutes, decipher it, relay
newsworthiness, then we could
take that as a signal to send a
reporter to do more exhaustive
  :      22  56
Few U.S. local newsrooms surveyed use AI for news production. The
scorecard results reveal moderate to strong interest in doing so.
Automation in news production can include automated text writing
and social media content creation.
What’s happening now
Scorecard ndings
Newsrooms rated the following statements on a scale of 1 to 5.
The bold scores on the left reect median values:
We regularly use AI in production operations.
2 Print and radio managers say they don’t use AI in production.
3 TV and digital managers gave neutral scores.
We have a few people who have tried AI technologies for production.
2 There is little experimentation with AI for production. Print, radio and TV
leaders gave this low scores.
3 Digital leaders gave higher scores but were neutral overall.
We're interested in AI to simplify production operations.
4 Managers are interested in simplifying production operations using AI. Print,
radio and TV leaders expressed solid backing.
5 Support was strongest among digital newsrooms.
Our production managers support exploring AI for their work.
3 Support for production AI needs to be built in print organizations where
managers returned neutral scores.
4 Radio, TV and digital leaders say their production managers would support
  :      23  56
What could be useful
Based on the scorecard and interviews, the following automations
are what U.S. local newsrooms would like to have for news
Generating content (e.g., text, video, photo, audio) and scheduling optimized posts
to Twitter, Facebook, Instagram; cropping photos and videos for dierent formats.
High school sports, college sports, weather, natural events (e.g., tides, res),
restaurant report cards, police logs, elections, agriculture grain bids, business
licenses, real estate and community calendars.
Obituaries, press release briefs, event previews, etc.
Newspaper layouts (weather, scoreboards, obituaries, comics, specials) and website
Identify photos related to stories, photos related to other photos (archives), archive
Identify videos related to stories, videos related to other videos (archives), archive
management and transcoding.
Personalize newsletters and optimize newsletter delivery times.
High school and college sports (e.g., box scores, leader boards) and competition
Maps (e.g., story-related, weather, re perimeters), knowledge graphs, suggest ways
to visualize a data set and broadcast graphics (e.g., banners, tickers, full screens,
multi-lingual subtitling and captioning).
Generate summaries (e.g., briefs, from government meetings and cut downs of
broadcast packages).
Monitoring that identies potential topics to cover that have not been covered.
Warnings to editors of overuse of specic photos.
Monitoring of the diversity of sources used in news coverage.
  :      24  56
“If we can find a way to
implement AI tools, I could see
us freeing up several hours of
the various mundane tasks
included in our day-to-day,
which would allow our reporters
and editors to focus on
producing more and better
stories for our communities.”
  :      25  56
“We’re looking at ways that we
can use more templating
technology to simplify the
placement process, but also for
data-driven presentations like
sports, agate and other kinds of
fixtures of the paper that require
manual work.”
  :      26  56
The U.S. local newsrooms surveyed view AI for news distribution
similarly to newsgathering and production; that is, with moderate
interest. Distribution includes everything involved in getting
content out to consumers. Examples include publication to multiple
platforms, personalization, search engine optimization, social media
postings, comment moderation and push alerts.
What’s happening now
Scorecard ndings
Newsrooms rated the following statements on a scale of 1 to 5.
The bold scores on the left reect median values:
We regularly use AI in distribution operations.
2 Print and radio managers say they do not regularly use AI in distribution.
3 TV and digital managers gave higher but neutral scores overall.
We have a few people who have tried AI technologies for distribution.
3 Print, TV and digital managers stayed neutral on this topic.
2 Radio managers were more condent that no one had experimented with AI
in distribution.
We're interested in AI to potentially deliver more relevant content for the audience,
which includes personalization.
5 Digital leaders are intensely interested in using AI for distribution.
4 Print, radio and TV leaders also expressed solid interest.
Our distribution managers support exploring AI for their work.
4 Further, there was consensus across all formats that distribution managers
support exploring AI for their work.
  :      27  56
What could be useful
Based on the scorecard and interviews, the following automations
are what U.S. local newsrooms would like to have for news
Personalization of homepage and learning what a subscriber is interested in and
delivering more of that content.
Filtering by language, duplicate commenter accounts and compiling comments for
Analytics to select the best time to publish and increase posting frequency.
Full-spectrum automation of publishing from website to email, to push alert and to
social posts.
Distribution to syndicators and aggregators.
Format articles as structured data to enable reuse in dierent platforms, format
broadcast scripts for the web.
Integration with A/B headline testing, oering recommendations and integration
with archives to recommend evergreen content.
Extending story recommendation abilities to personalize mobile push alerts.
  :      28  56
“It’s really interesting to think
about how to do distribution
efficiently without big system
replacements; those are hard
for a lot of smaller newsrooms
to deal with.”
“It’s a lot of very manual
stitching of all the data together
to tell the story. You don’t really
have one piece of technology or
an analytics tool that can do it
all for us.”
  :      29  56
The U.S. local newsrooms surveyed seek ways to improve business
eciency and see AIs potential to fill a prominent gap. Examples
include chatbots for customer and donor service, adaptive paywalls
and ad design.
What’s happening now
Scorecard ndings
Newsrooms rated the following statements on a scale of 1 to 5.
The bold scores on the left reect median values:
Our organization regularly uses AI in business operations.
3 Print, TV and digital managers gave neutral scores.
2 Radio managers returned lower scores overall.
We have a few people who have tried AI technologies for business operations.
3 Managers consistently provided a neutral score across all formats.
We're interested in AI to potentially improve business eciency.
4 There is solid backing for AI to be used across the industry for business
Our business leaders support exploring AI for their work.
4 Radio, TV and digital managers provided consistent support.
3 Print managers provided lower but neutral scores as a group.
  :      30  56
What could be useful
Based on the scorecard and interviews, the following automations
are what these U.S. local newsrooms would like to employ in their
business operations:
Tracking user journeys in detail to understand retention and sales, assisting
coverage decisions, demographics, referral details, prospecting for donations,
tracking which paragraph a consumer stops reading in a story and new ways to
understand what kind of content engages audiences.
Chatbots to handle routine subscriber, member and donor service issues, self-ser-
vice tools and implementing a CRM.
Design service for newspapers and digital allowing easy ways for clients to create
and proof advertisements.
Making it easier to order ads at all levels including hyperlocal, suggesting ways for
advertisers to improve their results, steering advertisers to proper services when
they need help and reaching out to potential advertisers.
Identifying potential donors so the audience team can follow up and tools to power
community engagement.
Personalize ads, underwriting messages to prospective donors and A/B testing of
donor messaging.
Determine the best time and content to show paywall content.
  :      31  56
Publisher David Blomquist
on learning new tricks
Jersey Journal Publisher and Editor David Blomquist said that at 65,
he’s as young a person as you might find who started journalism in
the “hot-metal” days.
“Everyone who’s around still as a legacy journalist understands that
our continued survival is rooted in accepting and leading the charge
for new technology.”
He said that if AI is going to accomplish more than put journalists
out of jobs, it needs to increase opportunities for expanding
coverage in institutions the size of the Jersey Journal.
“I would even argue there is greater opportunity to do good in
businesses of our size than at The New York Times. Those of us this
size, with a 3,500 daily print circulation, our mission will only be
fulfilled if we continue to find more ways to produce news creatively
and imaginatively.”
His AI dream is a tool that could scan an 850-page city council
meeting document, and the AI having learned from previous news
decisions, recognize that a redevelopment plan might be a
significant move in an ongoing discussion about city growth, that
there is something significant involving the pay of city employees
and an ordinance is set to increase municipal water bills by 7%.
“I know all that’s a reach too far,” he said. “But as Robert Browning
might have said, a bot’s reach should exceed its grasp, or what’s a
heaven for?”
  :      32  56
News director Ernesto Romero
on removing repetitive tasks
KYMA-TV in Yuma is emblematic of many smaller market stations
because it broadcasts newscasts under the CBS and NBC banners,
and airs programming on multiple channels for ABC, Fox, CW and
Telemundo. Its designated market area of southwestern Arizona and
southeastern California is vast.
As a border community, Mexico figures into its coverage as well.
What KYMA-TV News Director Ernesto Romero would like most for
the stations he directs is more reporters who could go deeper into
Mexico to report on migrants earlier in their journey. He’d also like
to cover more political stories, but Yuma isn’t considered large
enough by some candidates and politicians for campaign stops.
“They only come to Yuma a handful of times, whereas they’ll go to
Phoenix and Tucson. We’re part of the state and should be allowed
that same access to them that some of the bigger markets have.”
With two different news brands, there is a lot of repetitive work
from the one staff that powers both. Among the things that must be
changed for each broadcast are graphics and signoffs.
“It would be nice for us to be able to have a system where a reporter
would input something once and then it could automatically go
where it needs to go. It’s kind of my whole thing with having such a
small team is being able to have someone do something once and
then it puts itself on social media, online, or on air.”
  :      33  56
Owner Erin McIntyre on performing
the weekly miracle
Erin McIntyre has run Ouray County Plaindealer with her husband,
Mike Wiggins, in Ouray, Colorado, for almost three years. The
weekly paper is the only source of news for the entire county, so the
staff of three spends a lot of time at public meetings.
“We are the paper of record in Ouray County since 1877. We’re
known for being a trusted, reliable source of news that you can’t get
anywhere else. There are so many small, family-owned publications
like ours. We’re probably the youngest newspaper owners in
Colorado now, but there’s a lot of people who are looking to retire.
Whoever takes over a small newspaper is going to need a lot of help.”
McIntyre said tools that include automations could help new owners
like herself, who do everything to ensure the paper goes to press;
she referred to herself as “co-publisher, co-owner, co-editor, co-
janitor.” Ouray County Plaindealer has an audience that is seasonal
and continues to rely on its printed publication, and the printer is 40
minutes north of the office.
“They are the only press in the region anymore,” she said. “If we
didn’t print there, I would have to go five hours east or five hours
One of her biggest tech needs is a way for subscribers to be able to
change their own addresses and to purchase past issues. For her to
consider using a new tool or technology, it has to be mission critical.
“I live from week-to-week putting out this thing called a paper, which
is like, frankly, a miracle every damn week. If I’m going to
implement a tool, it better work because I can’t be babysitting it.”
  :      34  56
Publisher DeAnna Tisdale Johnson
on crossing the digital divide
DeAnna Tisdale Johnson became the publisher of the Jackson
Advocate, a weekly newspaper that serves the Black community in
Mississippi, in March 2020. Her parents had owned the paper since
1978. She wants to make the Jackson Advocate a multimedia
company and would like to have more reporters to cover farther
reaches of the state.
“Our motto is ‘the voice of black Mississippians,’ so those are places
where we want to expand in 2022,” she said. Another goal is to do
more reporting that is solutions driven and investigative, like the
collaborative project the Advocate did with the Mississippi Free
Press on how Black women have been affected by the pandemic.
“My parents took the ‘advocate’ part of our name very seriously and I
do, too,” she said. “My dad would go to court with people, and he
would pay someone’s rent. We aren’t just a newspaper. We’ve always
been doing the solutions part.”
The paper has long-time subscribers who are aging, and Tisdale
Johnson sees an opportunity to expand the digital footprint to reach
a younger audience but is uncertain how connectivity issues might
impact progress.
“Letting people know about what they need to be aware of for
COVID-19, it was difficult for us to reach some people digitally
because of lack of broadband,” she said. “We can have tools that use
AI, but if people aren’t attuned to digital mediums, it doesn’t
necessarily help us.”
History is a critical part of the publication, as it’s the oldest Black
publication in the state founded in 1938. Tisdale Johnson places
archiving as something she’d like to automate, so that all articles,
photos and multimedia files are easy to find.
  :      35  56
Digital specialist Brad Gowland
on identifying the ‘bird calls’
Michigan Radio’s Digital Specialist Brad Gowland developed
“Minutes,” based on an idea from reporter Dustin Dw yer. Minutes
is an application that downloads transcripts and uses machine
learning to transcribe audio to text.
The app monitors sites where local governments post videos
or audio of meetings, then transcribes that audio into searchable
transcripts that reporters can use to research stories.
They would love to see a tool that could scan a four-to-eight-minute
audio clip and determine the best parts to edit. To illustrate the
point, Gowland said that as a graduate student he had written code
that used machine learning to sort bird calls from field recordings.
The concept was to help ecologists take surveys to estimate
population sizes by how often different types of bird calls appeared
in the recordings.
“Most of what you have to do there is cut out all of the silence and all
the things that aren’t birds,” he said. “So, it’s kind of the same task …
how much of this audio is trash and how much of it do I want?”
  :      36  56
Analysis and conclusions
In this final section of the report, we focus on the needs that were
surfaced in the scorecard and selected interviews and draw some
conclusions about where AI solutions might be targeted at the local
level to help alleviate common pain points, streamline workflows
and enhance opportunities for monetization.
News managers surveyed are looking for ways to make it easier for
journalists to gather news more eciently. The scorecards and
interviews revealed strong support for handing off newsgathering
work to automation and AI. Editor Kyle Ocker of the Ottumwa
Courier in Iowa made this observation in the scorecard: “I see great
potential to replace functions that used to be designated to ‘news
clerks’ in the days of larger staffs but are now placed upon reporters
and editors.” The top items on managers’ wish lists were
transcription, along with content discovery of government
documents and social media monitoring.
Transcription, viewed by some news managers as grunt work, was the
most-requested automation wish in our scorecard. In technical speak,
transcription tools use NLP to understand what is being said in a
recording and who is saying it, then transcribes the results. With
multiple commercial tools available, some newsrooms have already
adopted or tested AI-based transcription technology. At KSAT-TV in
Texas, News Director Bernice Kearney called the transcription tool
Trint “a game-changer,” especially during the pandemic when most
civic meetings have been recorded. Oregon Public Broadcasting also
uses transcription, and Chief Content Officer Morgan Holm
describes transcription as something that makes newsgathering
more efficient, allowing the newsroom to put more resources “into
the human part of newsgathering.”
Additionally, some newsrooms want to take transcription further by
automating alerts of potential newsworthy topics discovered in
Multiple newsrooms in scorecards and interviews asked for help with
gleaning insights from government meetings and documents. Ouray
County Plaindealer in Colorado already applies AI transcription to
  :      37  56
recordings of government meetings. Co-founder Erin McIntyre
wants to go a step further and get alerts from those transcripts.
Technically, this type of content discovery can use both ML and
NLP. The San Francisco Chronicle also wants to get to the point
where AI delivers alerts on newsworthiness based on a transcript of
a city council, school board or board of supervisors meeting. With
hundreds of government bodies in its coverage area, the Chronicle
wants automation to take over initial coverage of some
governmental meetings. PACER, the website used to search court
documents, evoked frustration from several reporters, who spend a
lot of time and money pulling records. Kate Hessling, editor of the
Michigan-based Midland Daily News, said she would have liked to
see automated alerts for when records, dockets and rulings are
Some news managers asked for help with social media monitoring.
While commercial tools exist, some would like to go further.
WTAE-TV News Director Jim Parsons said that sometimes events
happen in the communities the Pennsylvania station serves, and the
reporting team misses it on social media. “We don’t have a great
system in place, other than keeping our eye on TweetDeck 24/7,” he
said. WTAE-TV relies on manual scouring of local websites and
social media that AI may help solve. “It would help to have some
‘robotic eyeballs’ trying to find these events.” Social listening tools
combine both ML and NLP.
Across the survey and interviews, news managers said they want to
streamline their production workows using AI and automation.
Executive Editor at the Traverse City Record-Eagle in Michigan
Nathan Payne described tasks like laying out pages that have a rigid
design, such as weather pages, as “busy work” and having “low
human demands.” In addition to automating page layouts, news
managers also said they want to automate social media content
creation, along with photo, video and audio suggestions.
Automating page layouts is one of the most-requested items on the
wish list by print newsrooms. From a technical perspective, an AI
layout tool would use ML and NLP. “There’s a big potential for the
layout of obituaries and sports agate,” said Ohio Lima News Editor
David Trinko. “Anybody who’s ever designed either knows what a
pain that can be, but they’re very standardized processes.” Some
newsrooms, however, wonder which layouts should be automated.
Web Editor Amy Libby of the Washington-state based The
Columbian sees layouts as a creative endeavor. It’s “where I go and
play and you have art and do fun things and cutouts and all that kind
of stuff,” she said.
  :      38  56
Newsroom leaders described automating social media content
creation as a signicant need, along with optimizing and scheduling
posts. Some managers spoke of wanting to automatically repurpose
existing content for social media, while others spoke of generating
original content for social media. Managing Editor Halle Stockton of
Pittsburgh’s PublicSource said that automation would “lighten the
load of social media posting while keeping our feeds fresh and
lively.” Sharing original stories on social media takes “a lot of time
and energy,” said KYMA-TV News Director Ernesto Romero of the
additional duty on his Arizona digital team. Sometimes, social media
duties are split among multiple people, leading to inconsistent
posting. News Editor Jason Ubay of Hawaii Public Radio said some
days they have one or two social posts, while on others five to 10,
and that it all depends on workloads. Executive Editor Payne in
Traverse City said he’d like to optimize posts, schedule posts to
match or maximize audience reach and tailor social media posts for
different audiences. Several commercial tools exist to optimize and
schedule posting to social media and use AI techniques including
ML and NLG.
Automated text writing was the third-most requested production
wish among those surveyed and interviewed. The technology, which
AP and its partners pioneered, is now widely available. These
automated writers use the AI technique of NLG and require a
structured data source. If data is available in a spreadsheet, the
information can be plugged into an auto writing system and produce
a similar article every time. Whereas AP automated reports on
business earnings and college sports, local newsrooms surveyed
would like to automate stories on high school sports, weather, police
logs, election results, restaurant report cards, grain bids, business
licenses, real estate and community calendars. In the scorecards,
multiple newsrooms also saw a need for automated text writing to
craft COVID-19 case updates. Data Editor Matt Kiefer of WBEZ-FM
in Illinois noted in the scorecard that “dumping data on our
audience may not help them completely understand the issue, so
automated text generation could help us bridge the divide from
structured data to readable text.”
Another area mentioned for streamlining newsroom workows was
the potential for automating suggestions for photos, video and audio
to accompany text stories. In an interview, a digital director said that
having automation suggest matching video clips for a story package
would help her to produce stories more efficiently. Further,
automation here could tie into an engagement strategy to increase
the amount of time audiences spend on digital platforms. AI-
technology powering suggestions could include computer vision, ML
and NLP.
  :      39  56
Across our research, newsroom managers often described
production automation needs that don’t require AI technology. A
typical example is automatically publishing to multiple social media
channels. Commercial tools exist for basic process automation. Kate
Hessling, the Michigan editor, said her team used IFTTT to post
items to Facebook and Twitter simultaneously. Such comments led
us to conclude that some newsrooms would see significant time-
saving benefits by implementing simple automation.
As audiences increasingly consume news through digital products,
local newsroom managers expressed a strong desire to deliver more
timely and relevant content across all channels. In surveys and
interviews, newsrooms sought automation for audience analytics,
story recommendations, website personalization and comment
moderation. WBUR-FM in Massachusetts Executive Editor for
Digital Tiffany Campbell said she was looking for anything to help
her newsroom “publish more and more efficiently without
compromising quality.”
Audience analytics turns raw digital data into usable information,
allowing news and business managers to make coverage and strategic
decisions. ML powers many of these tools. “We are especially
interested in using AI to create a more personalized and relevant
experience for diverse audience members with different interests,”
wrote KOSU-FM in Oklahoma Executive Director Rachel Hubbard in
the scorecard. Hubbard’s views represented many newsrooms, as
audience analytics ranked fourth on the automation wish list.
Analytics tools are commercially available and are used by multiple
newsrooms that took our survey. Even newsrooms that have access
to analytics wanted to learn more about their audience. Editor
Alison Gerber of the Chattanooga Times Free Press in Tennessee
described wanting to know what paragraph, or perhaps sentence,
that a reader stopped reading in an article. A station manager at a
public broadcaster said in an interview that they’d like to have an
AI-powered solution that combines potential content with market
data to make suggestions as to what content would perform better
on a specific day. Additionally, from a business intelligence
perspective, Hearst Newspapers Senior Manager of Content
Marketing Rachael Gleason would like a tool that automatically
stitches together data held in disparate locations to tell a story about
customer journeys and to make optimization recommendations.
(Gleason spoke as part of our interview with the San Francisco
  :      40  56
Several newsrooms said they’d like to personalize ads and
subscription oers. An executive at a public broadcaster said a
system for streamlining promotional advertising across different
social channels could be valuable. Web Editor Libby of The
Columbian said in the interview that the Washington state paper
would like to target subscription offers and advertising. “If you live
in Camas, we’re going to show you Camas ads. It would be nice for
the business to use, as long as we’re not trading on people’s privacy.”
Public broadcasters reported in the scorecard that more insights into
their audiences could help with donations. A California public
broadcaster said that it would like to have more personalized donor
messaging online and WBEZ-FM in Illinois reported that there
might be some opportunities around pledge drives in finding the
most effective messaging for potential members.
Story recommendations and website personalization were fth and
sixth, respectively, on the automation wish list. Both involve ML and
NLP and are examples of what can be done with audience analytics.
Survey participants viewed story recommendations in two distinct
ways. The first was from the producer side with techniques that
identify related content before posting. For example, North Carolina
WRAL-TV Digital Product Manager Jake Seaton would like
automation to recommend stories based on the historical
performance of that topic. It seeks “to elevate a story that may not
otherwise be on the producer’s radar.” Story recommendations were
also viewed by local news leaders from the audience side, where
stories would be recommended based on the reader’s viewing
history and preferences.
Many newsrooms indicated that they would like to personalize the
website homepage, so readers could get stories that they want above
the fold. “I would rather have a site that is reading the person and
using flexible content,” said Kirk Dougal, The Lima News Publisher.
“If they’re always going to local sports first, serve that upfront.”
Widening the scope of personalization, Libby of The Columbian
suggested news that is important and specific to the Pacific
Northwest would be ideal to surface to their audience. “Boeing
stories are important to us because we have a Boeing facility within
our distribution area,” she said. “Reports on Boeing could then be
put out into the world on Twitter or Facebook or wherever we would
want them to go without me having to touch every single one.”
Automating comment moderation was a top-10 wish in our survey.
The technology is an exercise in ML and NLP, with some notable
services developed in recent years. However, the scorecards and
interviews tell us more work needs to be done in this field. In Texas,
  :      41  56
KSAT-TV News Director Kearney wrote that their AI-based
comment moderation system “still requires significant human
input.” Some news organizations have removed comments on
website articles or limited who can comment. The San Francisco
Chronicle limits comment to its website subscribers. It also has
reporters circulate within the comments on high-interest articles to
answer reader questions. Michigan’s Traverse City Record-Eagle
uses Facebook for comments because people must log in and use
their name, limiting runaway comments. The Columbian’s Libby
said comments were eliminated from the Washington publication’s
website in March 2021. “I was just sick of it,” she said. “They were
bad for my mental health.” The Columbian now uses commenting
features on its Facebook page instead of hosting those conversations
on the website.
Newsroom managers were interested in supporting the business side
of the operation with enhanced revenue-generating solutions and
better decision-making data. In addition to audience analytics
described previously, local news leaders told us they’d like to have
automated customer and donor service capabilities, automation for
advertisers including ad design and tools to increase subscribers and
Automating customer service and donor service functions ranked in
the top 10 of the survey. Depending on the level of automation,
perhaps with chatbots, such systems can incorporate NLP and ML.
At a fundamental level, multiple local newsrooms wanted access to a
customer relationship management system (CRM). Newsrooms that
use a CRM wanted to do more, perhaps adding chatbots to handle
routine service inquiries. In an interview, Ouray County
Plaindealer’s McIntyre said she would like automation to help with
mailing address changes because many subscribers are seasonal
The scorecard also elicited requests to automate ad services and ad
design. The Santa Fe New Mexican wrote that they would like to
have “AI-driven design of online ads.” Having automation help
advertisers create their own ads is an opportunity multiple
newsrooms would be interested in. The technology would likely rely
on ML and NLG. Dave Fellabaum, Executive Director of Information
Systems of the Tribune-Review in Pennsylvania, pushed for making
it easier for digital advertisers to place orders.
The San Francisco Chronicle said in an interview that the newsroom
and business side share the same interest to increase subscribers
  :      42  56
and engagement and to limit churn. Recently the paper has met its
subscription goals and said they’d like to focus now on retaining
subscriptions and analyzing churn. “Technology that gives insight
into the customer journey and how that can be optimized would be
valuable,” said Hearst Newspapers Senior Manager of Content
Marketing Rachael Gleason. Tampa Bay Times said it would like to
have better tools for targeting potential subscribers, while digital
publications and public broadcasters told us they would like tools to
increase membership and donations.
Several leaders spoke about the need to make the transition from
traditional output — print, radio and television — and help their
readers find them on the web and social media. Traverse City
Record-Eagle Publisher Paul Heidbreder said “developing, improving
and increasing our digital presence” was an overarching business
goal. Oregon Public Broadcasting Chief Content Officer Morgan
Holm said, “to survive in a digital world, the secret for us is creating
content primarily for digital platforms and then figuring out how we
can put it on broadcast. We have to flip that model 180 degrees and
then realize that different platforms bring different audiences.”
New technology, of course, is needed to accomplish the transition
from legacy to digital. Newsrooms in the scorecard and interviews
said that making the business case for new technology involves a
range of challenges from determining need, sifting through various
application options and then figuring out how to integrate
something new into existing infrastructure. For the Tampa Bay
Times, implementing new technologies is “a mixed bag,” said
Product Manager James Collins. “One of the primary drivers is
when a system contract is about to expire. Do we renew or replace?”
The second driver is when the software “just bites the dust.” A third
driver is when a staff member goes to a seminar and sees a pitch,
then it will come down to cost. To make the budget work, some
newsrooms partner with university talent and seek foundation
funding to address technological gaps.
Reflecting the views of many respondents, a commercial TV station
engineering director said that if there is a technology or a product
that will help in producing news better and delivering it more
efficiently to audiences, “We’ll chase it, if we believe it will make a
  :      43  56
Key Takeaways
Significant workload reductions could be seen with increased use
of transcription tools.
Alerts from the posting of new government documents could help
journalists keep watch over more agencies.
Improvements to social media monitoring could help newsrooms
stay on top of content discovery.
Tools to help generate, optimize and schedule social media posts
could be big time savers, especially for small newsrooms where
staff add social to their other duties.
Automating page layouts could reduce workloads for print
Having automated writing applications pick up briefs, such as for
high school sports and weather, could help newsrooms reallocate
journalists to focus on writing more substantive stories.
Automation that helps suggest photos, video and audio items to
match text stories could speed up the production process.
Implementing audience analytics could help newsrooms create a
baseline for personalizing news content and advertising, along
with honing pitches to potential subscribers and donors.
Content moderation systems could help newsrooms deal with a
high volume of audience feedback.
A robust CRM and chatbots would be helpful to handle routine
subscriber and donor-service issues.
Benefits are possible from helping digital advertisers place and
design their own ads with help from AI-powered services.
Many local outlets would be interested in tools to increase
subscriptions, memberships and donors.
  :      44  56
AP sta
The two leaders on AP’s Local News AI initiative, Aimee Rinehart
and Ernest Kung, were new to the organization. They quickly came
to know the teams across many departments that provided expertise
on every facet of this project.
From the Strategy team, Jim Kennedy has provided his decades of
experience as a journalist, editor and innovator to guide this project
and report. Our daily meetings have added clarity and gave room to
brainstorm, all with good humor. Gloria Sullivan has acclimated the
team to AP and ensured swift payment to project support people
and organizations.
The Partnerships team Lisa Gibbs, Jin Ding and Bryan Pollard have
made this project possible by securing funding from the Knight
Foundation and maintaining project updates. They also served as
careful final readers of this report and we thank them for that
additional work.
For the beautiful illustrations that support our work throughout this
project Peter Hamlin and Philip Holm, and for the design of this
report, Hal Hilliard and Julie Yee.
Many thanks to the Revenue team who shared the scorecard link
and programmatic details with AP members: Nicole Baugh, Ivett
Chicas, Jim Clarke, Dawn DeGuzman, Michael Fabiano, George
Garties, Paul Memoli, Sam Moore, Nicole Morales, Eva Parziale, Jim
Pollock, Dave Rizzo, Starr Talley, Sara Trohanis, David Smith, Chris
Weis, David Wilkison, Michelle Williams and Adam Yeomans.
AP Engagement team reached out to members with email alerts and
held an AP Academy session so that we could explain the initiative:
Nancy Nussbaum and Ashlee Schuppius. The Marketing team
launched the project page and help with our web presence: Brian
Barth, Kaya Bieler-Rasmussen and Antoine Vessaud. The
communications team refined and posted our project blog posts:
Lauren Easton and Patrick Maks.
The News team explained their project to localize AP content for
  :      45  56
members, Noreen Gillespie and Jake Kreinberg. Karen Mahabir
detailed AP’s fact-checking processes. Reporter Garance Burke
provided great insights based on her reporting on AI in society and
our work, and Sharon Lynch is working with the team for the online
course development. The AP Data team guided our course structure
and possible future collaborations: Justin Myers, Michelle Minkoff
and Troy Thibodeaux.
To the AP team that walked us through the features of the broadcast
system ENPS: Brian Doyle, Brian Hopman, Jason Smith and Andrew
Wormser; and Ted Anthony for providing a snapshot of AP
Local newsroom leaders
We appreciate the time local newsroom leaders took to share
what is happening in their news operation. We circled back to 25
newsrooms to conduct interviews on a recorded Zoom meeting
with some conversations lasting nearly two hours. Thank you for
your generosity of time and for relaying your newsroom experiences.
Four newsrooms asked not to be identified in the list below.
The Lamar Democrat
Daily Sitka Sentinel
Arizona Daily Star
Helena World
Bay City News
Lodi News-Sentinel
Outlook News Group
San Francisco Chronicle
The Appeal-Democrat
The Bakersfield Californian
The Sacramento Bee
Voice of San Diego
Yuba Net
CPR News
Ouray County Plaindealer
Delaware State News
Center for Public Integrity
Tampa Bay Times
USA TODAY Network / Gannett
Atlanta Journal-Constitution
Forsyth County News
The Korea Daily Atlanta
The Times
The Walton Tribune
Pacific Daily News
  :      46  56
Hawaii Public Radio
East Idaho News
Christianity Today
The Pantagraph
South Bend Tribune
Lee Enterprises
Globe Ga zette
Kossuth County Advance
Ottumwa Courier
Telegraph Herald
The Gazette
The Emporia Gazette
The Kansas City Beacon and The
Wichita Beacon
Wichita Eagle
Casey County News
Clinton County News
Crittenden Press
Daily News
The Lebanon Enterprise
The Messenger
The Advocate and The Times-Picayune
Portland Press Herald
Cumberland Times-News
Boston Herald
The Berkshire Eagle
The Sun Chronicle
Cadillac News
Great Lakes Now
Michigan Radio
Midland Daily News
The Alpena News
Traverse City Record-Eagle
Star Tribune
Jackson Advocate
Houston Herald
The Joplin Globe
Lincoln Journal Star
Omaha World-Herald
The Nevada Independent
This Is Reno
The Telegraph
The Jersey Journal
Westfield Leader
Albuquerque Journal
New Mexico News Port
Santa Fe New Mexican
The Buffalo News
The Haitian Times
Times Union
Sandhills Sentinel
The News and Observer
Ohio Center for Journalism
Richland Source
The Business Journal
The Lima News
The Plain Dealer
  :      47  56
Mvskoke Media
Herald and News
Oregon Public Broadcasting
The Register-Guard
Centre Daily Times
Gettysburg Connection
Philadelphia Inquirer
The Daily Review
The Times-Tribune
El Vocero de Puerto R ico
The Providence Journal
The Post and Courier
Chattanooga Times Free Press
Knoxv ille News Sentinel
Main Street Nashville
El Paso Matters
Houston Chronicle
Texarkana Gazette
The Salt Lake Tribune
Brattleboro Reformer
Daily News-Record
The Columbian
The News Tribune
The Seattle Times
The Intelligencer
The Record Delta
Kenosha News
La Crosse Tribune
Oil City News
Outreach and AI landscape
America Amplied
American Press Institute
Big Local News
Brown Institute at Columbia
Center for Community Media
Center for Public Integrity
Center for Cooperative Media, “The Local
Hearst Connecticut Media Group
Institute for Nonprot News
Jackson Advocate
Knight Foundation
Lee Enterprises
Local Media Association and Word in
Local Media Consortium
National Federation of Community
News Media Alliance
News Product Alliance
NYC Media Lab
Partnership on AI
Poynter, “Local Edition,” AP wants to
help local newsrooms with AI and
Reynolds Journalism Institute, University
of Missouri
RTDNA, How AI can help journalists and
deliver better journalism
Solutions Journalism Network
South Bend Tribune
The National Trust for Local News
Tiny News Collective
Trusting News
  :      48  56
University professors, center
directors, and practitioners who
shared their expertise and project
Paul Cheung, The Center for Public Integrity, Chief Executive
Janet Coats, University of Florida College of Journalism and
Communications, Managing Director, Consortium on Trust in
Media and Technology
Kat Duncan, Reynolds Journalism Institute, Director of
Laura Frank, COLab, the Colorado News Collaborative, Executive
Tim Franklin, Medill, Professor and John M. Mutz Chair in Local
John Keefe, CNN, Senior data and visuals editor
Damon Kiesow, Reynolds Journalism Institute, Knight Chair in
Digital Editing and Producing
April Lindgren, Ryerson University, Professor and Velma Rogers
Research Chair
Edward C. Malthouse, Medill, Erastus Otis Haven Professor and
Research Director of Spiegel Research Center
Randy Picht, Reynolds Journalism Institute, Executive Director
Hilke Schellmann, New York University, Assistant Professor of
Patti Sontag, Ryerson University, Local News Data Project Editor
Reuben Stern, Reynolds Journalism Institute, Director, New York
  :      49  56
Ryan Thornburg, University of North Carolina, Associate
Journalism Professor
Serdar Tumgoren, Stanford University, Lorry I. Lokey Visiting
Professor in Journalism
If we missed you or your organization, we apologize for the oversight
and ask that you email to let us know. If you would like to
help AP share upcoming AI programmatic details like the free online
course and the AI project pitch information, please email
Northwestern students
Northwestern University: Knight Chair for Digital Media Strategy at
Northwestern University Medill School Jeremy Gilbert and
Executive Director at Knight Lab Joe Germuska devoted their fall
2021 class to help us shape the scorecard and interview questions,
conduct the interviews with 25 newsrooms, and provide an AI
product overview. Many thanks to students Hannah Barton, Helen
Bradshaw, Joshua Hoeflich, Grace Lee and Sammie Pyo. Some of the
students shared their insights from the interview experience:
Adopting tech
“Interviews revealed how mysterious AI seems to non-technical
people. Adoption of AI technologies could be improved by making
tools more approachable to people without a formal tech
—Joshua Hoeflich
Ethical concerns
“It’s important to start creating legal guidelines for newsrooms in
using AI, because even though few might intentionally misuse it,
less experienced newsrooms could use some best practices.”
—Hannah Barton
Smaller, the better
“We interviewed the director of a radio station of “two-and-a-half
staff and realized that it is imperative that local newsrooms like
these have access to automation tools to serve their communities.”
—Helen Bradshaw
  :      50  56
Automation wish list
We’ve organized the automation wish list items presented in the findings section
here by the number of times they were requested overall. It’s important to note that
journalists share many common needs despite differences in format. Our results
for the automation wish lists are based on 135 scorecards where people could ask
for whatever they wanted. We ranked each need by the number of times they were
mentioned by a news organization, with each newsroom getting one vote for each
item on the wish list.
transcription * 29 16 12 8 65
social media content creation * 18 14 6 10 48
auto writer (structured data) * 21 11 4 4 40
audience analytics * 16 7 8 5 36
recommendations, story * 19 7 6 4 36
website personalization * 17 6 5 4 32
CRM *, customer service functions, subscriber and donor services 14 6 2 6 28
page layouts * 22 2 2 1 27
comment moderation * 12 5 4 1 22
photo suggestions * 7 5 3 3 18
newsletters 7 2 1 5 15
data analysis 10 1 2 2 15
high school sports coverage 8 2 1 3 14
auto writer (unclear data type) 6 2 2 1 11
sports score layout/agate 7 0 2 1 10
social media scheduling 6 2 1 1 10
chatbots, subscriber service * 9 1 0 0 10
document analysis, large data sets 5 3 0 1 9
ad design * 7 0 0 2 9
ad services 6 1 1 0 8
graphics, data visualization 3 2 1 2 8
content discovery (structured data) 4 2 1 0 7
translation 2 2 1 2 7
publish to multiple platforms 3 1 3 0 7
SEO 2 1 2 2 7
auto writer (unstructured data) 2 1 2 1 6
publish from CMS to web 4 0 2 0 6
audience/community engagement 1 0 3 2 6
archives, photo 3 3 0 0 6
community calendar 4 0 0 1 5
analytics (unclear what type) 2 3 0 0 5
website curation 2 3 0 0 5
  :      51  56
CMS improvements 3 2 0 0 5
content discovery (unstructured data) 4 0 0 1 5
government meetings, coverage 2 1 1 1 5
photo ingestion with metadata 2 2 0 0 4
donor messaging personalization 0 2 1 1 4
personalization (unclear what type) 2 2 0 0 4
captioning video and audio 0 1 2 1 4
government meetings, sorting agendas 3 0 0 1 4
metadata, stories 2 2 0 0 4
photo, editing 2 0 1 1 4
chatbots, donor service 0 2 0 2 4
summarization, text 3 0 1 0 4
archives, story 2 1 1 0 4
data entry 2 1 1 0 4
subscribers, increase 3 0 0 0 3
identify manipulated or out of context media 1 1 0 1 3
metadata, unspecied 2 0 1 0 3
social media monitoring 2 0 1 0 3
police log publishing 3 0 0 0 3
video suggestions 0 2 1 0 3
AP Services 1 0 1 1 3
attract subscriptions 1 0 0 1 2
media monitoring 0 2 0 0 2
cover employee o-hours 0 2 0 0 2
push alert personalization 0 1 0 1 2
agate, no detail 1 0 0 1 2
text to speech 1 0 0 1 2
collect state wire stories 2 0 0 0 2
paywall, exible and adaptive 2 0 0 0 2
middleware 2 0 0 0 2
crowdsourcing, unspecied 1 0 0 1 2
tips, responses 0 1 1 0 2
election coverage 2 0 0 0 2
ugc (user-generated content), processing 2 0 0 0 2
teleprompter 0 0 2 0 2
content transformation and reuse 1 0 1 0 2
archives, video 1 0 1 0 2
donors, prospecting 0 0 0 2 2
analytics, business 1 0 0 1 2
push alerts 0 0 2 0 2
scheduling (unclear what it refers to) 0 0 0 1 1
scheduling broadcast times 0 1 0 0 1
advertiser communications 1 0 0 0 1
audio imaging 0 1 0 0 1
  :      52  56
video editing 0 0 1 0 1
marketing materials 1 0 0 0 1
content syndication 0 1 0 0 1
alerts for programming and radio operations 0 1 0 0 1
podcast production 0 1 0 0 1
identify overused elements 1 0 0 0 1
pages, preparation for printer 1 0 0 0 1
oer price optimization 1 0 0 0 1
identication of new human sources 0 1 0 0 1
crowdsourcing, images for web stories 0 1 0 0 1
inventory management 0 0 1 0 1
content discovery (evergreen content) 0 0 1 0 1
track content reuse by other services 0 0 0 1 1
photo galleries 1 0 0 0 1
scheduling, publication 1 0 0 0 1
tips, verication 0 0 0 1 1
agriculture coverage (grain bids) 1 0 0 0 1
identify mis/disinformation 0 0 1 0 1
metadata, video 0 0 1 0 1
graphics, broadcast production 0 0 1 0 1
recommendations, video 0 0 1 0 1
weather, unspecied 0 1 0 0 1
college sports coverage 0 1 0 0 1
minor league stories 0 1 0 0 1
grant mining 0 1 0 0 1
grant writing 0 1 0 0 1
content scheduling, dayparting 1 0 0 0 1
chatbots, news delivery 0 0 0 1 1
business license publishing 1 0 0 0 1
government coverage 1 0 0 0 1
press release processing 1 0 0 0 1
source diversity tracking 0 1 0 0 1
audio suggestions 0 1 0 0 1
fact checking 0 0 0 1 1
identify under-covered topics 0 0 0 1 1
review photos and videos 1 0 0 0 1
sponsorships, prospecting 0 0 0 1 1
website publishing 0 0 0 1 1
FOIA request management 1 0 0 0 1
real estate transactions 1 0 0 0 1
graphics, suggestions 0 0 0 1 1
* This wish list item was among a list of examples provided in the digital survey to help clarify the types of automations
we were seeking.
  :        
Scorecard stats
Here’s a detailed look at the scorecard data from 187 newsrooms used to inform this report.
The scorecard composite score measures AI readiness, knowledge and usage. It is calculated
on a scale of 24 to 120. We also calculated how much AI is already being used in a newsroom
on its own; that measure is on a scale of 0 to 6. The balance of the scorecard was answered on
a scale of 1 to 5, with one being strongly disagree, three neutral and five strongly agree.
(90 newsrooms)
(44 newsrooms)
(31 newsrooms)
(22 newsrooms)
Number of sta 104.41 45 85.36 50 104.35 100 23.82 11
Scorecard composite score 71.89 74 71.43 73 76.71 75 77.05 82
AI usage 1.17 1 0.93 1 1.61 1 1.09 0
1. We have a good understanding of what AI is and how it relates to
journalism. 2.89 3 2.64 3 2.71 3 3.14 3
2. Our organization has a solid strategy for AI that crosses all departments. 1.74 1 1.59 1 1.65 2 1.91 2
3. We feel ready for AI technologies in our operations. 2.40 2 2.34 2 2.55 2 2.7 7 3
4. We are concerned about falling behind in AI. 3.48 3 3.43 4 3.48 4 3.32 4
28. We have the nancial resources to invest in AI. 2.58 3 2.39 3 2.84 3 2.32 3
29. We have people with AI skills in our organization. 2.62 3 2.55 3 3.13 3 2.73 3
30. We can allocate time to work on AI projects. 3.09 3 3.20 3 3.42 4 3.55 4
31. We're condent that AI can take on repetitive tasks to free up resources
for more substantive work. 3.82 4 3.82 4 3.84 4 4.36 5
7. Our newsroom regularly uses AI in newsgathering. 2.19 2 2.27 2 2.7 7 3 2.23 2
8. We have a few people in the newsroom who have tried AI technologies
for newsgathering. 2.56 2 2.64 2 2.87 3 2.45 2
9. We're interested in AI to potentially help reduce the workload for our
journalists. 3.92 4 4.07 4 4.23 4 4.50 5
10. Our journalists support exploring AI for their work. 3.40 3 3.73 4 3.58 4 3.64 4
  :        
(90 newsrooms)
(44 newsrooms)
(31 newsrooms)
(22 newsrooms)
12. We regularly use AI in production operations. 2.16 2 1.91 2 2.61 3 2.32 3
13. We have a few people who have tried AI technologies for production. 2.49 2 2.43 2 2.48 2 2.77 3
14. We're interested in AI to simplify production operations. 3.90 4 4.09 4 3.97 4 4.36 5
15. Our production managers support exploring AI for their work. 3.31 3 3.7 7 4 3.55 4 3.91 4
17. We regularly use AI in distribution operations. 2.49 2 2.02 2 2.81 3 2.55 3
18. We have a few people who have tried AI technologies for distribution. 2.72 3 2.27 2 3.06 3 2.68 3
19. We're interested in AI to potentially deliver more relevant content for
the audience (this includes personalization). 4.10 4 4.14 4 4.29 4 4.36 5
20. Our distribution managers support exploring AI for their work. 3.69 4 3.7 7 4 4.10 4 3.77 4
23. Our organization regularly uses AI in business operations. 2.42 3 2.23 2 2.42 3 2.41 3
24. We have a few people who have tried AI technologies for business
operations. 2.54 3 2.45 3 2.61 3 2.59 3
25. We're interested in AI to potentially improve business eciency. 3.87 4 3.89 4 3.97 4 4.23 4
26. Our business leaders support exploring AI for their work. 3.51 3 3.80 4 3.77 4 4.18 4
  :      55  56
Technologies in use today
Here are some of the technologies used by newsrooms that
completed the scorecards and interviews. This is by no means a
comprehensive list of the technologies used and does not reflect the
breadth of tools commercially available. Not all the technologies
contain AI. AP does not endorse third-party products.
AP Newsroom, Dataminr, Follow That Page, Klaxon, LexisNexis, NewsWhip,
PACER, Workbench
Airtable, Slack, Teams, Trello, WhatsApp
Grammarly, Otter, Trint
Document Cloud, Google Pinpoint
AP ENPS, Arc, Canva, Chorus, Dalet, IgniteTech Olive, Inception, NewsBoss,
Storyform, StoryMate, Tecnavia, TownNews BLOX, Triton Digital, Wildmoka,
Colaboratory, GitHub, JavaScript, Python, R, SQL
DataWrapper, Flourish
Airship, Chartbeat, Coral Project, CrowdTangle, Disqus, Futuri, Google Analytics,
Hearken, Looker, Mather, OpenWeb,, Viafoura
Hootsuite, SocialFlow, Social News Desk
Listrak, Mailchimp, Sailthru, Second Street Lab
BlueConic, Inka, Salesforce
Semrush, Wunderkind
For more information, go to
... B. Diakopoulos, 2019;Thurman, Lewis & Kunert, 2021;Wilczek et al., 2022) und Branchenberichte (z. B. Goldhammer, Dieterich & Prien, 2019;Kalim, 2021;Rinehart & Kung, 2022), dass Medienorganisationen zunehmend auf Künstliche Intelligenz setzen. Doch Patentrezepte fehlen, es bleibt die Frage nach dem Wie: Wie lassen sich Wertketten in Medienorganisationen -sogenannte Nachrichtenwertkettenmithilfe von Künstlicher Intelligenz optimieren? ...
Full-text available
Angesichts der vielfältigen Herausforderungen, welche die Digitalisierung an Medienorganisationen stellt, müssen Medienmanager:innen und Journalist:innen ihre Prozesse und Erlösmodelle optimieren – und mithin ihre Nachrichtenwertketten effizienter ausrichten. Anwendungen der Künstlichen Intelligenz (KI) bieten in diesem Zusammenhang Chancen. Entwicklung und Implementierung von KI sind aber ressourcenaufwändig und berücksichtigen nicht immer die normative Tragweite, die dem Journalismus gemeinhin innewohnt. Kooperationen – etwa zwischen der Medienindustrie und der Wissenschaft – dürften daher in Zukunft weiter an Bedeutung zunehmen.
Full-text available
La inteligencia artificial ha llegado al periodismo en diferentes fases del proceso de producción de noticias, desde la identificación de tendencias informativas, al tratamiento de datos o a la producción automática de textos, entre otros. Su potencial se manifiesta, sobre todo, cuando existe una gran cantidad de datos, algo que pueden ofrecer secciones como las de deportes y economía. De las dos opciones, en este trabajo se ha elegido el periodismo deportivo y, así, buscamos conocer y entender cómo los periódicos, radios, televisiones y productos nativos digitales de Brasil y Portugal se han relacionado con la inteligencia artificial (IA). Para lograr los objetivos propuestos, enviamos una encuesta a los responsables de varios medios de comunicación de ambos países con preguntas que nos ayudaron a darnos cuenta de que los editores y responsables conocen la contribución que la IA puede ofrecer a sus redacciones. Entre otros resultados obtenidos, destacamos que el uso de IA en los medios brasileños está más presente que en Portugal, pero hay una convergencia en sus usos y en las dificultades para obtener un mayor desarrollo: la falta de recursos económicos y los bajos conocimientos sobre el potencial de la IA. Entre las razones destacables para usar dicha tecnología está el objetivo de hacer más eficiente el trabajo de los periodistas y de ahorrar tiempo en la producción. Se concluye que los decisores de medios deportivos portugueses y brasileños son conscientes del potencial de la IA, pero ahora mismo las dificultades económicas y profesionales son el principal adversario para su implementación en las redacciones.
Full-text available
Les défis des agences de presse internationales AFP, Reuters, AP et Bloomberg à l'ère des GAFA et de l'Intelligence artificielle. par Paloma Martínez Méndez (UQAM), sous la supervision de Patrick White (UQAM). Les agences de presse internationales sont des « courtiers en informations » internationaux qui recueillent, rédigent et distribuent des informations provenant d'un pays spécifique ou du monde entier à des journaux, des périodiques, des diffuseurs de radio et de télévision, des agences gouvernementales et d'autres utilisateurs. En général, une agence ne publie pas les nouvelles elle-même, mais les fournit à ses abonnés qui, en partageant les coûts, obtiennent des services qu'ils ne pourraient pas se permettre autrement. Dès la dernière moitié du XXe siècle, ces agences sont devenues omniprésentes dans le paysage médiatique, car c'est à cette époque que le public du monde a commencé à accéder à des informations nationales et internationales produites par l'une ou plusieurs d'entre elles et par l'intermédiaire des médias de masse qu'elles desservent. Dans un monde post-Covid 19, les agences de presse doivent se réinventer pour affronter les défis venant des géants du web et des avancées technologiques.
ResearchGate has not been able to resolve any references for this publication.