Why did ChatGPT's citations to AP, TIME and TechCrunch drop so fast?
ChatGPT citations to AP, TIME and TechCrunch fell up to 90% in a week after a model update, while Google rankings for those sites stayed flat.
Starting the day OpenAI rolled out the GPT-5.6 family in ChatGPT, citations to AP, TIME, TechCrunch, the LA Times and The Atlantic fell by 50% to 90% within a week, according to Search Engine Land's reporting. Google's own rankings for these same publishers stayed largely flat over the same period. The mismatch happened because ChatGPT's citation engine and Google's ranking algorithm are separate systems that don't move together.
What actually happened to ChatGPT's publisher citations
The drop was sudden and specific. Several major publishers fell to near zero in ChatGPT citations, even though most still allowed OpenAI's search crawler and other AI providers kept citing them, as reported by the Manila Times' coverage of Muck Rack's Generative Pulse report.
The timing lines up with a model change rather than anything the publishers did. OpenAI pays AP, TIME and The Atlantic for their journalism, and on July 9 ChatGPT stopped returning the favour in its citations, per Search Engine Land. Some recovery has followed: AP has clawed back about half its old level since August 13, and AP and Axios are back on ChatGPT's top-cited list, the same source notes.
How the drop was discovered and measured
The figures came from Muck Rack's Generative Pulse team, presented by VP of data and intelligence Matt Dzugan at the Generative Pulse Summit in New York City on October 28, drawing on more than 40 million AI responses for its "What Is AI Reading?" report, according to Search Engine Land. That scale matters. This wasn't a handful of test prompts; it was a broad sample tracked over months, which is why the pattern held up as a trend rather than noise.
Why this isn't a one-off
This publisher collapse sits inside a much longer run of instability. Journalism's share of ChatGPT's citations slid from roughly a third in January to about a fifth by September, dropping in two clean steps in April and July, per Search Engine Land. Across all AI platforms tracked by Generative Pulse, journalism's citation share fell from about 27% in May to 22% in October, the same source reports.
Separate research from Ahrefs found instability is a general phenomenon across tools: at least half of citations disappeared by the next day, around 60% on Google AI Mode and ChatGPT, 50% on AI Overviews, and 80% on Gemini and Copilot, based on Ahrefs' study of citation stability. Reddit went through a near-identical sudden drop tied to a ChatGPT model change in August, with its share of citations in one continuously monitored dataset falling from roughly 4.67% to 0.38% between the two measurement windows, according to Elmo's analysis of ChatGPT citation data. We cover that specific episode in Reddit's ChatGPT citations dropping 86% in four days. Earlier in the year, ChatGPT's overall citation volume fell 86% to 94% across five tracked markets between February and April, driven by an ad rollout and a default-model change, according to Lyra's research on the 2026 citation drop, a trend also covered in why ChatGPT's zero-source answers jumped 8x in late 2026.
Why Google rankings stayed stable while ChatGPT citations collapsed
Google hit publishers with a core update in May, and ChatGPT's citations to them shrugged it off. ChatGPT then hit them in July, and Google's rankings shrugged that off too, per Search Engine Land. Several affected publishers didn't see comparable declines in Google's top-three keyword rankings during the same period, which means a website can hold its Google search visibility while losing citations in ChatGPT entirely, as iTechManthra's writeup of the study points out. A drop in one system tells you nothing reliable about the other. We've written before about why topping Google means nothing to ChatGPT right now, and this episode is a clean real-world illustration of that gap.
What's replacing journalism in ChatGPT's citation mix
As journalism's share fell, other source types grew. Something shifted in how ChatGPT selects sources around August 2026: publisher content roughly halved as a share of its citations while brand-operated domains rose twelvefold, according to Azoma's tracking of ChatGPT citation changes. The same tracker found that other major platforms collapsed too over the summer. YouTube's citation share fell by around 97%, Wikipedia's by around 96%, and Reddit's by around 91% between July and September 2026, per Azoma.
In a cross-engine test from September 2026, ChatGPT drew around 34.6% of its citations from government and nonprofit sources, while Gemini leaned most heavily on company-owned content, according to Axios' reporting on AI citation sources. That matters for founders because it suggests the gap left by publishers isn't being filled by one winner. It's scattered across official sources, brand sites and whichever outlets a given model happens to favour that month.
Why AI citation and Google ranking are fundamentally different systems
The low overlap between the two systems isn't new to this episode. Ahrefs has previously found only 12% of AI-cited links rank in Google's top 10 for the same prompt, per Search Engine Land's summary of that research. A separate Ahrefs study found Google rankings and organic traffic show a stronger association with AI citation stability than Domain Rating or backlinks, and that around 75% of pages cited by AI Mode and ChatGPT didn't rank in Google for the same question, based on Ahrefs' citation stability research and its methodology of tracking the same prompts over time. A different cross-engine study found the overlap between AI Overview citations and organic top-10 results dropped from 76% in mid-2024 to 17% in February 2026, according to DEV Community's writeup of the BrightEdge and Ahrefs findings.
Our view: treating AI citation volume as a proxy for Google rank, or vice versa, is a mistake that will keep catching people out as long as teams run SEO and GEO off the same single dashboard. If your reporting conflates the two, you'll misread a genuine AI-visibility problem as a search ranking problem, or miss one entirely.
What founders and marketers should do about it
Stop assuming your Google position protects your presence in AI answers, and track the two separately. Because so much citation volatility traces back to model swaps and version changes rather than anything publishers or brands did, the useful response is monitoring, not panic. Our AI search readiness checklist is built for exactly this: tracking how your brand shows up across engines so a sudden drop doesn't go unnoticed for weeks.
It's also worth understanding what a "citation" even means in each engine's output, since not every link ChatGPT shows functions as a true source, a distinction unpacked in why don't most of ChatGPT's links count as citations.
One limitation worth being honest about: none of this research tracks causation back to a specific model mechanism. The studies measure before-and-after citation counts around known release dates. They don't show what inside GPT-5.6 or any other model caused the shift, so any explanation beyond "this coincided with a model change" is speculation.
Frequently asked questions
Did ChatGPT stop citing AP, TIME and The Atlantic on purpose?
The notes don't show OpenAI explaining a deliberate decision. What's documented is a timing coincidence: citations fell sharply starting the day GPT-5.6 rolled out, and OpenAI has paid AP, TIME and The Atlantic for journalism, per Search Engine Land. AP has since recovered about half its prior level.
Does losing ChatGPT citations mean a site is losing Google rankings too?
No. Several publishers hit by the July ChatGPT drop didn't see comparable declines in their Google top-three keyword rankings in the same period, according to iTechManthra's coverage. The two systems have shown low overlap generally, with Ahrefs finding only 12% of AI-cited links also rank in Google's top 10 for the same prompt.
Is this citation volatility unique to publishers?
No. Reddit, Wikipedia and YouTube all saw steep citation share declines across similar periods in 2026, according to Azoma's tracking, and Ahrefs found at least half of citations across five AI tools disappeared within a day in its own stability testing.
What's filling the gap left by falling journalism citations?
Government and nonprofit sources, brand-owned content and third-party corporate content have all grown as shares in different engines, according to Axios' cross-engine citation analysis. The mix varies notably by platform, with no single source type replacing journalism across the board.
Sources
- ChatGPT citation drops look like unannounced Google core updates
- Generative Pulse: Earned Media Still Drives More Than 84% of AI Citations
- How Stable Are AI Citations? 10 Things We Learned Tracking 37,500 Citations
- ChatGPT UI Reddit Citations Fell to API Levels
- ChatGPT citation drop 2026: what happened, and what to do
- ChatGPT Citation Drops: Why They Look Like Unannounced Google Core Updates
- Why AI visibility drops: what changed in AI Search citations, August 2026
- GEO starts with the question, not the answer
- 81% of ChatGPT-cited brands don't rank in Google's top 10. The decoupling is now data - DEV Community
Want more Sited in your Google results?
Add Sited as a preferred source on Google (opens in a new tab)

