A leading newspaper says most of its articles are AI-written

Will The New York Times, The Wall Street Journal, the Financial Times, The Guardian, Le Monde, or the Asahi Shimbun state that a majority of the articles it publishes are written by AI systems?

Chance by Q4 2031

3%

Cumulative probability

0%25%50%75%100%
  • Latest forecast
Chance by the end of each quarter
YearQ1Q2Q3Q4
2026<1%
2027<1%<1%<1%<1%
2028<1%<1%<1%<1%
20291%1%1%2%
20302%2%2%2%
20313%3%3%3%

Forecast history

Forecast made on 12 Oct 2026

Latest of 4 forecasts · Methodology v2.1

Model reasoning

All six newspapers have publicly committed to human-written journalism. Independent audits find AI wrote under 1% of news articles at the American titles checked, and saying most articles are AI-written would hurt the trust these papers sell. Even the most AI-enthusiastic of them, Asahi, quickly insisted that AI "only assists" when criticised. The development that would most change this forecast is the Wall Street Journal or Asahi moving to having AI write most articles and openly describing it that way.

Where things stand (12 Oct 2026)

None of the six papers has said anything close to "most of our articles are written by AI," and none is close to it in practice. All six have publicly committed to human-written journalism:

  • New York Times. Its AI principles say AI work "should always be managed by and accountable to journalists" (nytco.com, May 2024). In 2026 it tightened the rules: freelance work "must be the product of human creativity," and AI is banned from guest essays (Semafor, 26 Aug 2026; Futurism).
  • Financial Times. Its 2023 editor's letter says FT journalism "will continue to be reported and written by humans." Its code bans AI in writing; it put a note on a guest column in August 2026 after AI was used to shorten it.
  • Le Monde. Its 2024 AI charter says generative AI "ne peut en aucun cas remplacer les équipes éditoriales" ("can in no case replace the editorial teams") (ActuIA). Its main AI uses are translation and audio, and AI translation of human-written articles does not count here.
  • The Guardian. It revised its guidance in March 2026 to allow only limited uses; AI should "support our journalists' expertise, never replace it" (Nieman Lab, 29 Jul 2026).
  • Asahi Shimbun. It is the most aggressive of the six. Its president calls the strategy "AI all-in," has pictured a newsroom of 100–200 "super journalists," and has said ordinary reporters will disappear within 10 years (Oct 2025 talk). Today it publishes only AI-drafted notices in its Tokyo edition (StoryHub case study). When the "AI all-in" interview drew public criticism in April 2026, Asahi quickly issued a statement that "AI is only an aid to humans" (ITmedia, 9 Apr 2026; Asahi's AI principles).
  • Wall Street Journal. Its editor-in-chief, Emma Tucker, praised Fortune's AI-written stories (Semafor Media, Apr 2026). Its opinion editor lets outside contributors use AI, but staff must still write their own pieces (The Atlantic, Aug 2026).

On actual practice, an audit of the Times, Post and Journal found only 0.7% of news articles flagged as partly or fully AI-written, against 9% across U.S. newspapers as a whole (Russell et al., ACL 2026). AI-drafted stories are spreading fastest at local chains (McClatchy, Advance Local's "Express Desk") and at Fortune, not at these six titles. New York's FAIR News Act, which would require labels on AI-generated news, is awaiting the governor's signature. If signed, it would make heavy AI drafting more visible and so more costly for the Times and the Journal.

Reference class and base rate

I found no case of any mainstream newspaper officially saying that most of its articles are AI-written, let alone a prestige title, so the base rate is close to zero. The nearest past examples are disclosures of automation within one category: AP's automated earnings stories, and Bloomberg's estimate that about a third of its content used automation. Neither was a newspaper-wide majority, and neither was AI-written prose in the sense used here.

Paths to YES and NO

Resolving YES needs two unlikely things together:

  1. At least one title moves to AI being the main writer of most of its articles.
  2. It then says so explicitly, in an official statement that calls the articles AI-written or AI-generated rather than "AI-assisted."

The most plausible candidates are Asahi (shrinking staff, a pro-AI president) and the Journal (a pro-AI editor). The scenario most likely to produce this is very rapid AI progress in 2028–2031, which industry leaders were openly warning about in September 2026, making AI writing normal and removing the stigma.

Several forces point to NO:

  • These papers sell trust and human expertise.
  • Unions and contracts at all six resist AI replacing journalists.
  • Several are suing AI companies (the Times against OpenAI; Asahi and Nikkei against Perplexity), and saying most of their own articles are AI-written would undercut their own lawsuits.
  • Surveys show readers distrust AI-written news.
  • Papers almost always describe AI use as "assisted" and "journalist-led," which does not count here.

My rough estimates were:

  • P(at least one title has a majority of AI-drafted articles by end-2031): about 10–15%.
  • P(it says so in qualifying terms, given that): about 15–25%.
  • Plus small tail risks, such as an AI-maximalist new owner.

Forecast construction

I used quarterly hazards rising from 0.04% (the rest of 2026) to 0.28% per quarter in 2031, reflecting improving AI and the lag before institutions change. I computed the cumulative series in Python: about 0.3% by end-2027, 1.5% by end-2029, and 3.4% by end-2031.

Strongest case against my estimate: if AI capabilities take off by around 2028, AI writing could become normal fast, and a pro-AI leadership at the Journal or Asahi could openly describe AI-written output. That could justify 5–8%. I give it partial weight; institutional lag and the requirement for an explicit statement keep me near 3.5%.

Resolution criteria

Question. By the last day of each quarter, has The New York Times, The Wall Street Journal, the Financial Times, The Guardian, Le Monde, or the Asahi Shimbun stated that a majority of the articles it publishes are written by AI systems?

Definitions.

  • Named newspaper: any one of the six named titles, including its print and digital journalism. A parent company's other publications, separate sister titles, and unrelated publications with similar names do not count. A change of ownership does not disqualify the named title.
  • Majority: strictly more than 50% by number of articles, not by word count, readership, revenue, or staff time. An explicit statement that “most” or “a majority” of its articles are AI-written qualifies without a numerical percentage. Exactly 50%, “up to 50%,” “many,” and “a substantial share” do not qualify.
  • Articles it publishes: the newspaper's overall output of distinct editorial text articles, including news, features, reviews, opinion, and externally contributed or syndicated articles. Duplicate print/digital appearances, translations, and updates to the same article are not additional articles. Advertisements, sponsored content, reader comments, market-data tables, standalone headlines, and audio or video without an accompanying article are excluded. A figure limited to one desk, section, edition, language service, newsletter, or experimental product does not establish the newspaper-wide majority.
  • Written by AI systems: AI generates all or most of the article's substantive prose. Human commissioning, reporting, fact-checking, editing, or approval does not disqualify it if AI remains the principal writer of the published text. An unqualified publisher statement that articles are “AI-written” or “AI-generated” satisfies this definition. Merely using AI for research, transcription, translation of human-written articles, proofreading, headlines, summaries attached to human-written articles, or minor rewriting does not. AI may be supplied by any developer.
  • Reporting period: any identified completed period, with no minimum duration, or an explicit statement describing the newspaper's current overall output. The Observation period convention applies; a historical statistic for a period ending on or before the opening date does not count.

Resolves YES if at least one named newspaper makes a qualifying official statement that the majority threshold has actually been reached for its own overall article output. A numerical numerator and denominator establishing more than 50% also qualify when the accompanying definitions meet these criteria. An official statement of present practice qualifies even if it does not specify calendar dates.

Resolves NO if no qualifying statement has appeared by the quarter-end. Plans, targets, predictions, permission to use AI, assertions that most journalists use AI, or a majority of articles being merely AI-assisted do not qualify. Neither isolated AI-written articles nor outside audits, AI-detector scores, leaks, or allegations establish the required self-statement.

Resolving source. Official statements by the named newspaper, its publisher, or its editor-in-chief speaking institutionally, published on the newspaper's or publisher's own domain, in its print edition, or in a publisher-authored report. Eligible domains include nytimes.com and nytco.com; wsj.com and dowjones.com; ft.com and aboutus.ft.com; theguardian.com and theguardian.org; lemonde.fr; and asahi.com, plus their official successor domains. Editorial-policy pages, signed editor's letters, official newsroom announcements, and corporate reports qualify; ordinary commentary or an employee's personal statement does not. This is a narrower source rule than Public confirmation (A2): third-party reporting, including reports of executive interviews, cannot independently resolve the question. Apply the Resolution window convention and Resolution date (A3); a newly dated substantive update may qualify, but an undated page or an unchanged older statement does not.

Download all forecasts (JSON)