Dossier mode
The Doom Loop: What AI Does to Journalism
The same investigation, restaged one beat at a time. Drive it with the arrow keys, space, or autoplay. Nothing is cut from the piece — long runs are split across frames. Read the full investigation or open the Who Controls What You Get to Know hub.
The doom loop.
A Microsoft executive's phrase for the company's own AI: a machine that answers your question by summarizing the news, so you never visit the newsroom — starving the reporting the machine was trained on. If it runs unchecked, what happens to journalism?
The decline figures and the AI click-through drop are FACT. That answer engines will break the business model funding original reporting is a forecast — well supported, not certain — graded PROBABLY TRUE. And AI did NOT start the fire: the collapse began years before ChatGPT.
Disclosure: drafted with Claude, made by Anthropic — an AI company whose products also summarize news. Full COI in the companion, An Astonishing Theft.
Journalism was already collapsing before AI — and AI didn't cause that.
Per Medill's State of Local News 2024: 3,200+ print newspapers have closed since 2005 (~two a week; 130 last year), leaving ~5,600, 80% weeklies; 206 counties have no news source and 1,561 only one, so ~55 million Americans have limited or no local news; fewer than 100,000 people work in newspaper publishing (BLS). Per Pew, daily circulation fell from ~55 million in 2005 to under 21 million by 2022. The cause was the internet's capture of advertising — classifieds to Craigslist, ads to Google and Facebook — not AI. This is the baseline AI now acts upon.
AI answer engines are draining the last lifeline: search and referral traffic.
Pew's July 2025 analysis of 68,879 real Google searches: when an AI summary appeared, users clicked a traditional result link in only 8% of visits (vs 15% without), clicked a link inside the summary just ~1% of the time, and ended their session 26% of the time (vs 16%). In the OpenAI/Microsoft case, Microsoft's own data showed 83–93% drops in click-throughs to Times and Daily News domains (51–94% for Ziff Davis) from Copilot vs Bing, and Cloudflare's CEO put OpenAI's pages-scraped-to-visitors ratio at 1,500:1 by mid-2025 (Google 18:1).
The same tools flood the market with synthetic 'pink slime.'
As real reporting gets harder to fund, fake reporting gets nearly free. Per the newspapers' brief, at OpenAI's API prices it costs ~$6,800 to generate one million 500-word news-style articles with no reporter. One operation, Prism News, ran 200 AI-generated outlets posing as local newsrooms with four employees, repackaging others' articles as 'new' stories in the originals' markets. Pink slime dilutes the market, siphons readers and ad money, and is often indistinguishable from genuine local news — hitting the communities already becoming news deserts.
The fork: sign or sue — and the money flows to the big, not the local.
Both paths are visible now. Some publishers sue (Times, Daily News, Ziff Davis, CIR, Intercept). Others sign: OpenAI has reported licensing deals with News Corp (~$250M over five years), Axel Springer, the Associated Press, The Atlantic, Vox Media, and more. The fork is FACT; the inference graded PROBABLY TRUE is that licensing entrenches inequality — national/legacy brands can extract deals or afford lawyers, while local weeklies and the 206 no-news counties can neither license nor litigate. A subsidy for the strongest newsrooms and nothing for the weakest, accelerating the deserts rather than filling them.
The doom loop: the machine eats the supply chain it needs to work.
Put together: if answer engines keep the readers, subscriptions, and ad views that pay for reporting — and the click-through data says they do — the newsrooms producing the training data shrink and the synthetic flood grows. A Microsoft memo called this a 'doom loop' that 'will hurt the performance of our models and the entire web at the same time'; an internal line said large language models are 'a product that destroys its supply chain.' Graded PROBABLY TRUE, not FACT, because it's a forecast outcomes can still change (courts, licensing, policy). But it's the industry's own description of where the arrangement leads: fluent answers about a world fewer people are left to report.
You cannot summarize reporting that no one did.
AI didn't start the fire — the internet's capture of advertising gutted newspapers for two decades before ChatGPT. What AI changes is the last step: the survivors pulled readers through search to their own pages, where a subscription or ad still earned something, and the answer-engine model removes that visit. A chatbot can only tell you what happened at the school board meeting or the corruption trial if a reporter was in the room; strip that funding and the machine keeps talking — repeating, guessing, or inventing, with no one left to check. It's the sequel to An Astonishing Theft (the same firms' own 'doom loop' and 'substitutive' admissions) and a companion to The Wrong AI Debate (attention kept on distant risks, off present harms). The transfer underway is of power over information itself — from the people who gather facts to the machines that repackage them.