Journalism has 63 AI job titles on file. The biggest desk is the BBC's five

Key findings

1. 63 of the 127,602 editorial rows in our published base carry an AI word in the job title, 0.049% of the base, and they sit at 48 of the 8,800 outlets with at least one row.
2. The BBC has the most, 5 of its 3,935 editorial rows, followed by Bloomberg with 3 of 1,785; nine outlets have 2 each and 37 have exactly 1.
3. Four of the BBC's five AI titles name a job running AI inside the newsroom and one names a job reporting on it.
4. Under our keyword rule, 27 of the 63 titles are about deploying AI in the newsroom, 21 are about covering AI as a subject and 15 are unclear.
5. Reading the LinkedIn headline as well as the title raises the count to 325 of 127,602 rows, 0.255% of the base, across 225 outlets.
6. The vendor stores "A.I." as "a i", which our lexicon misses. A variant regex for that spelling matches 65 titles, 60 of them among the 63, and reading the 5 new ones by hand takes the title-only count to 65 rows at 49 outlets.
7. The New York Times has 0 AI job titles among its 2,168 editorial rows under the strict lexicon and 2 under the variant spelling.
8. 34 of the 50 largest publishable outlets on file have no AI job title, and 16 of the 50 have no AI word even in a headline.
9. Sports job titles outnumber AI job titles 4,877 to 63 in the same 127,602 rows, and AI job titles outnumber crypto job titles 63 to 42.
10. Of the 43 countries with 200 or more rows in the base, 30 have no AI job title at all, and the highest rate is Finland's 2 titles on 408 rows.
11. 14 of the 63 AI titles carry head, director, lead, chief or senior, 22%, against 12.4% of all 127,602 rows.
12. The Reuters Institute reports that Newsquest created more than 30 AI-assisted reporters. We hold 42 editorial rows across 6 of the 74 Newsquest domains we could name, and 0 of the 42 carries an AI word in the title.

63 of 127,602 editorial rows carry an AI word in the job title

We ran an AI lexicon over the job title field of every row in our published base. The lexicon is \b(ai|artificial intelligence|machine learning|generative ai|llm|chatgpt|openai)\b, matched against the lowercased title with & replaced by "and". 63 rows match, 0.049% of the 127,602. Those 63 sit at 48 outlets, out of the 8,800 outlets that have at least one editorial row in the base.

Source: JournalistLabs, September 2026.

We expected a small number. We did not expect it to be small enough to print every row on one page, which is what we have done below.

Newsquest has more than 30 AI-assisted reporters and 0 AI job titles on file

The first objection to a title count is that newsrooms run AI teams without anybody changing their title, so we checked it against an outside figure before writing anything else. The Reuters Institute's Journalism, Media and Technology Trends and Predictions 2026, fetched on 2026-09-16 from https://reutersinstitute.politics.ox.ac.uk/journalism-media-and-technology-trends-and-predictions-2026, says: "Newsquest, a leading regional UK publisher, has created more than 30 'AI-assisted reporters' across its titles, staff who are equipped with an AI tool called News Creator to draft stories - up to 30 every day - which are then checked and supplemented."

Newsquest's mastheads carry their own domains, so a group-domain count would miss most of them. We built the masthead list from Wikipedia's List of newspapers published by Newsquest and the official-website field on the matching Wikidata items, which gives 74 Newsquest domains. They are in newsquest_domains.csv. Six of the 74 have any editorial row in our base, 42 rows between them: thenational.scot 20, heraldscotland.com 11, newsquest.co.uk 7, edp24.co.uk 2, dorsetecho.co.uk 1 and herefordtimes.com 1. None of the 42 carries an AI word in the title. More than 30 AI-assisted reporters work at Newsquest by the Reuters Institute's count, and 0 of our 42 Newsquest rows says so in the title.

Source: JournalistLabs, September 2026.

So the 63 counts titles. At regional groups like Newsquest it misses the people doing the work under an older title. Matching the same lexicon against the title plus the LinkedIn headline gives 325 rows at 225 outlets, 0.255% of the base. The headline is free-text profile copy that carries former jobs, side projects and keyword stuffing, so we do not treat a headline match as a newsroom AI job, but it is the widest reading of the same data.

The other correction is a spelling one. The vendor drops the periods in "A.I." and stores the string as "a i", which our lexicon does not match. We ran \ba\.?\s?i\.?\b over the titles. It matches 65 titles in the base. 60 of those 65 are already among the 63. Five are new, and we read all five by hand:

• a i Projects Editor, The New York Times, United States
• Associate Editorial Director, a i Initiatives, The New York Times, United States
• Managing Editor a i, Hearst Magazines, Netherlands
• Picture Editor Art Director a i, Harper's Bazaar Netherlands, Netherlands
• Visual Editor a i, Vogue, Netherlands

The two New York Times headlines spell the word out as A.I. The three Dutch rows use a.i. in its Dutch sense, ad interim, and we read them out. The variant regex also misses 3 of the 63, which spell it out and carry no bare ai token:

• Artificial Intelligence Editor, Financial Times, United Kingdom
• Reporter Electric Vehicles, Artificial Intelligence, PCMAG, United States
• Artificial Intelligence & Special Projects Reporter, Politico, United States

Taken raw the variant rule gives 68 title-only rows at 52 outlets, and with the three Dutch titles read out it gives 65 rows at 49 outlets. We publish the strict 63 as the headline and the 65 beside it. The New York Times has 0 AI job titles among its 2,168 editorial rows under the strict lexicon and 2 under the variant.

We read all 63 strict matches one by one, checking for AI as a personal name, AI inside another acronym such as AIB, and the Dutch ad interim. Nothing was removed: 0 of 63. Five of the 63 are flagged borderline in the borderline column of ai_titles.csv: three Dow Jones risk-and-compliance data research editor titles, one at Barron's and two at the Wall Street Journal, the "Ai Transcript Editor" at CalMatters, and the VFX video editor at Sinclair Broadcast Group. We publish 63, and 63 minus those 5 is 58. All five sit on the deploying side, so dropping them leaves 22 deploying, 21 covering and 15 unclear of 58.

The BBC has 5 AI job titles and Bloomberg has 3

The 48 outlets with at least one AI title are ranked in ai_desk_ranking.csv. The BBC is first with 5 of its 3,935 editorial rows. Bloomberg is second with 3 of 1,785. Nine outlets have 2 each: Aftonbladet, Amagi, the Financial Times, MIT Technology Review, People Inc., The Verge, The Wall Street Journal (Dow Jones), Time and ZDNET. The remaining 37 outlets have exactly 1, and 5 plus 3 plus the nine twos plus the 37 ones is the 63.

Source: JournalistLabs, September 2026.

The BBC's five are:

• Ai Correspondent, Bbc News
• Editorial Executive, Generative Ai Programme
• Head of News Ai (Senior News Editor, Ai)
• Innovation & Ai Assistant Editor - Driving Change in Journalism & Broadcasting
• Journalist, News Ai Team

Those are the strings as the vendor stores them, title-cased, so AI reads as Ai.

Four of those five are about running AI inside the newsroom and one, the correspondent, is about reporting on it. We could not check the five against anything the BBC publishes. The BBC's AI principles page at https://www.bbc.co.uk/supplying/working-with-us/ai-principles/ gives no headcount, and the two AI pages we tried under aboutthebbc returned 404 on both the .co.uk and the .com spelling. No outlet in the 48 publishes the size of its AI team anywhere we could find, so every count in this section stands on the profile titles alone.

Small outlets in the ranking carry high rates because they hold few rows. Amagi has 2 AI titles on 8 editorial rows, The Verge 2 on 31 and ZDNET 2 on 19. We would not read anything into the per-1,000 column for those.

27 of the 63 titles are about deploying AI and 21 about covering it

We split the 63 with a keyword rule applied to the title, first match wins.

1. Deploying, the title names a newsroom function performed with AI: innovation, development, integration, operations, adoption, validation, transcript, assisted, automation, accelerator, strategy, programme, program, project, intrapreneur, ai-hub, ai hub, news ai, editorial ai, generative video, vfx, data and ai.
2. Covering, the title names a reporting job: reporter, correspondent, writer, journalist, columnist, critic, reviewer.
3. Unclear, neither list matches.

Deploying is tested first because a title can carry words from both lists. "AI Validation Journalist" at Amagi checks machine output and "AI Assisted Reporter" at the Miami Herald writes with a tool, and both would land on the covering side if the order were reversed.

That gives 27 of 63 deploying, 21 of 63 covering and 15 of 63 unclear.

Source: JournalistLabs, September 2026.

Five of the deploying rows name the function plainly: Head of News AI (Senior News Editor, AI) at the BBC, Global Editor, AI Development and Integration at Reuters, Lead Editor, AI Innovation in Journalism at McClatchy Media, Director, AI Editorial Operations at People Inc., and AI Assisted Reporter at the Miami Herald.

We also ran a broader rule that counts any of editor, head, director, lead, operations, innovation, development, integration, product or strategy as deploying. It gives 35 deploying, 25 covering and 3 unclear. We did not publish that split because it reads "AI Editor", "Artificial Intelligence Editor" and "Editor, AI" as deployment. A title like that can be a desk covering the subject or a job running AI, and our rule leaves it unclear for that reason. Nine of the 15 unclear rows are editor titles whose only qualifier is the AI subject, marked yes in the title_names_a_desk column; the other 6 are not. The 20 rows where the two rules disagree can be read off ai_titles.csv by row_id, where class and class_alt sit side by side, so a reader who prefers the broader rule can switch to it row by row.

Here are all 63.

"Ai" with a lower-case i is the vendor's title-casing of AI, which people write in capitals. Country is the person's own country as the vendor records it, not the outlet's, which is why one BBC row reads United States and both Amagi rows read India. The two identical "Director, AI Editorial Operations" rows at People Inc. are two different people; the build dedupes on name, title and domain.

34 of the 50 largest publishable outlets have no AI job title

We ranked the publishable outlets by editorial rows and took the top 50. Publishable means media-typed, named, typed, with a country, and not a parent group domain. This is the ranking that feeds the /outlets/ table on this site, and the site gives its own page to the first 40 of it. 34 of the 50 have no AI job title. 16 of the 50 have no AI word anywhere, in a title or a headline: Associated Press, SABC, ABC News, DR - Danmarks Radio, CBS News, Fairfax Media, SVT, The Telegraph, USA Today, RTE, ABS-CBN News, Vogue, RTV Slovenija, ETV, PA Media and Mediacorp.

Source: JournalistLabs, September 2026.

The Associated Press has 1,057 editorial rows on file and no AI word in any of them. CBS News has 577 and none. The full 50 with both counts are in top50_newsrooms_ai.csv.

Sports job titles outnumber AI job titles 4,877 to 63

We ran seven other beat lexicons over the same title field and the same 127,602 rows, so the comparison is like for like.

Source: JournalistLabs, September 2026.

Sports job titles outnumber AI job titles 4,877 to 63. Health titles outnumber AI titles 590 to 63, technology titles 492 to 63 and climate titles 385 to 63. AI titles outnumber crypto titles 63 to 42 and cybersecurity titles 63 to 23. Every regex is in lexicon.txt. Each figure counts words in a job title, and we have no measure of what any of these people file.

30 of the 43 countries with 200 or more rows have no AI job title

Counted by the person's own country, over the 43 countries that clear the 200-row cut a country page needs on this site. The 63 AI titles sit in 13 of the 43, and one of the 63 carries no country at all.

The United States has 32 AI titles on 69,377 editorial rows, 0.46 per 1,000. The United Kingdom has 11 on 12,094, 0.91 per 1,000. India has 4 on 7,656, Sweden 3 on 1,879, Germany 2 on 1,643, Singapore 2 on 787 and Finland 2 on 408. Finland's 4.90 per 1,000 is the highest rate of the 43, and Australia's 0.24 per 1,000, on 1 title across 4,170 rows, is the lowest that is not zero.

Source: JournalistLabs, September 2026.

The 30 countries with none include France on 1,369 rows, South Africa on 1,530, Denmark on 1,125, Brazil on 1,024 and Italy on 907. The full table, with the headline-inclusive column beside the title-only one, is ai_by_country.csv.

14 of the 63 AI titles carry a rank word, 22% against 12.4% of the base

Testing the 63 titles for head, director, lead, chief or senior matches 14 of them. The per-word counts are senior 8, director 3, head 2 and lead 2, which sum to 15 over 14 distinct rows because "Head of News Ai (Senior News Editor, Ai)" at the BBC carries head and senior. The same five words over the whole base match 15,793 of the 127,602 rows. So 22% of AI titles carry a rank word against 12.4% of all editorial titles.

Source: JournalistLabs, September 2026.

Our own seniority column on the same 63 rows reads unknown 45, senior 9, executive 5, chief 2, deputy 1 and staff 1. It is assigned by the classifier rather than by the string match above, so the two need not agree.

The funnel: 173,133 rows in, 127,602 published, 63 AI-titled

The base starts at 173,133 classified editorial rows. We drop 12,540 of them because they sit on a link-in-bio, social or marketplace domain rather than a newsroom. We drop another 32,991 because the outlet is not media-typed, which removes the universities, academic publishers and government agencies that reach our outlet table as employers of journalists. The 127,602 rows left are the published base, and every page on this site counts from it. 325 of the 127,602 carry an AI word in the title or the headline, and 63 carry one in the title. The full method is at https://journalistlabs.com/method/.

Counts below five are printed as themselves here, against our usual rule, because all 63 rows are published with their outlet and country and every per-outlet count can be derived from the list.

Every one of the 63 titles is self-written, and the profiles were fetched in July and August 2026

Job titles here are profile text people wrote about themselves. A person can put AI Editor on a profile with no desk behind it. A person running an AI programme may never update the title on the profile, which is what the Newsquest check above points at.

The profiles behind the 63 rows were fetched between 2026-07-23 and 2026-08-11, median 2026-08-06. A title created after that date is not in this file, and a title that has since gone is still in it. We cannot tell from the data whether a person with an AI title still holds it.

Coverage is thin below the large outlets. 8,800 outlets have at least one editorial row in the base, and most local and regional titles have a handful or none. Group domains split newsrooms too: People Inc. carries 183 editorial rows and Hearst Newspapers 18, both company domains covering several mastheads.

Not every outlet among the 48 is a general newsroom. Biotechtv, Fashionunited and Boston Magazine contribute one row each, and all are media-typed. Every row is on this page, so a reader who does not accept some of them can drop those and recount.

No first-party page from any outlet gives the size of its AI team, so the BBC's five and Bloomberg's three are unchecked against the outlets themselves.

Cite this

JournalistLabs, "Journalism has 63 AI job titles on file. The biggest desk is the BBC's five", September 2026, https://journalistlabs.com/research/journalism-has-63-ai-job-titles-on-file-biggest-desk-bbc-five/.

Dataset

Every number on this page is in /static/data/journalism-has-63-ai-job-titles-on-file-biggest-desk-bbc-five/. The person-level base is not published, so these files are the grain a reader can check.

• ai_titles.csv, the 63 rows, one per line, with row_id, class, class_alt, title_names_a_desk, borderline, outlet_domain and fetched_at.
• ai_desk_ranking.csv, the 48 outlets with at least one AI title.
• top50_newsrooms_ai.csv, the 50 largest publishable outlets with their AI counts and their zeros.
• ai_by_country.csv, the 43 countries with 200 or more rows.
• beat_title_counts.csv, the eight beat comparisons under both rules.
• newsquest_domains.csv, the 74 Newsquest domains with the rows we hold for each.
• lexicon.txt, every regex used here.
• report_ai_desks.py, the script as it ran.
• facts_console.txt, the console output of that run, and sanity_console.txt, the outside check with its HTTP statuses.
• README.md, the columns and denominators, and METHOD.md, the base, the two match rules and the limitations.

fetched_at on the 63 rows runs from 2026-07-23 to 2026-08-11, median 2026-08-06. Data: CC BY 4.0, JournalistLabs, September 2026.