There is no public, current ranking of news domains in AI search and nothing else. News brands show up inside mixed lists, usually behind a platform or an encyclopedia. Forbes is the name that recurs near the top of ChatGPT's. The 12% figure people quote is not news's share of citations. It is how often ChatGPT cites a URL after retrieving it from a news channel: 12.01%.
- Forbes is the recurring news brand on ChatGPT: 1.1% of all citations in Profound's window through June 2025 (rank 3), and 3.1% of the US top-source pool in September 2026 (rank 4).
- On AI Overviews, a news domain is a rounding error of the head. Business Insider and the New York Times are at 0.5% of the top-source pool. CNBC is 0.3%.
- 12.01% is a conversion rate. Of 3.94 million URLs retrieved through ChatGPT's news channel, about one in eight was cited. Search-channel URLs were cited 88.46% of the time.
- Germany's top 1,000 ChatGPT pages are not a news list either. Editorial outlets were 12.0% of citations inside that head. FAZ was the first general news domain, at rank 10.
- Promptwatch's August 2026 "News Article" slice is 5.6% of a classified mix that itself sums to about 56% of citations. It is not a complete partition, and it is ChatGPT only.
The leaderboard that is not there
A page called "most-cited news domains" should be a table of Reuters, the AP, the Times, the Post, the BBC, and their shares. That table, on one prompt panel, across engines, was not found. Building it by picking the news-looking rows out of five incompatible studies would be a new study, and this site has not run it. Broken chain for a pooled news top ten. Partial for the rows below, which are sightings.
What follows is every news-looking row that survived a pass through the engine lists, with the denominator still attached. If a name is missing, the honest reading is "not in the published head," not "never cited."
Where news brands do rank
| Domain | Where it ranked | Share | Denominator |
|---|---|---|---|
| Forbes | ChatGPT, Profound, through Jun 2025 | 1.1% | All ChatGPT citations |
| Business Insider | ChatGPT, Profound, through Jun 2025 | 0.8% | All ChatGPT citations |
| NYPost | ChatGPT, Profound, through Jun 2025 | 0.7% | All ChatGPT citations |
| Reuters | ChatGPT, Profound, through Jun 2025 | 0.6% | All ChatGPT citations |
| Forbes | ChatGPT, Ahrefs US, 2 Sep 2026 | 3.1% | Top-source pool |
| New York Times | Perplexity, Ahrefs US, 2 Sep 2026 | 1.6% | Top-50 pool, rank 10 |
| Forbes | Copilot, Ahrefs US, 2 Sep 2026 | 1.4% | Top-50 pool, rank 19 |
| New York Times | AI Overviews, Ahrefs US, 2 Sep 2026 | 0.5% | Top-source pool, rank 26 |
| FAZ | ChatGPT, Ahrefs Germany, 21 Sep 2026 | 1.94% | Inside the 1,000 most-cited pages |
Profound's January 2026 category table is the class version, and it is not a domain list. Media was 25% of ChatGPT citations, 33% of Copilot, 24% of Claude, and 19% of Perplexity. "Media" includes more than newsrooms. The sample sentence on that page does not match every engine in the table. Use the direction: media is a quarter of some engines and a fifth of others, while any one news domain is around 1%.
On a German page-level cut the same pattern holds one country over. Ahrefs, 21 September 2026, classified the 1,000 most-cited ChatGPT pages out of 3,698,476 cited pages that day. Editorial outlets were 132 pages and 12.0% of citations inside that thousand. The top of the domain table was ADAC, the consumer centres, statute law, and Stiftung Warentest. FAZ was tenth.
The 12% figure, read correctly
Ahrefs, Louise Linehan with Xibeijia Guan, published 15 April 2026, on 1.4 million ChatGPT prompts the article dates "February 2025" and also calls ChatGPT 5.2. That pairing was not reconciled. Keep the date conflict visible. Partial
ChatGPT labels retrieved URLs with an internal type. Citation rate by channel: search 88.46% (25,563,589 rows), news 12.01% (3,940,537), Reddit 1.93%, YouTube 0.51%, academia 0.40%. Overall, half of retrieved URLs were cited. The news row answers a useful question: given that a story was retrieved as news, how often did it become a footnote. It does not answer "what share of footnotes are news sites." Cited news pages were younger than discarded ones, a median of about 200 days against about 300, and title-relevance scores of the two groups were nearly identical.
Promptwatch's August 2026 content-type series, classified citations only, shows News Article at 5.6% beside Product Page at 28.7% and Listicle at 10.1%. The displayed slices sum to about 56%. News gained a point or more during the month. Millions of citations are claimed. The classified-coverage rate is not on the page. A 5.6% of a partial mix and a 12% conversion rate can sit in the same month and describe different events.
What this does not prove
It does not say AI search has stopped citing news. It says no one has published the news-only league table, and the mixed tables put individual newsrooms well behind Reddit, YouTube, Wikipedia, or a retailer, depending on the engine. HasData's ten-query AI Mode probe on general news is a list of who gets cited when the question is the news. That probe is on the crawler-block page, because its useful finding is about robots.txt, and the sample is ten queries.
Pew's "5% of AI summaries linked to a news website," BuzzStream's 14% news share, and a Qvery 0.11% were not re-read at the original pages. They are not in the table. The Tow Center test remains the best opened evidence that identifying the right article is a failure mode, and it is seventeen months old as of this page.
How to use what exists
Pitch the story to the engine that actually cites newsrooms at a visible rate, and do not expect the homepage of the Times to outrank YouTube on an Overview. ChatGPT is the list where Forbes, Insider, the Post, and Reuters have printed shares. Copilot's category row has the highest media share and a domain list led by retailers, which is a reminder that "media" and "the New York Times" are not synonyms.
If the goal is the citation rather than the brand name, the page-type evidence says a buying guide on a news domain is what gets the links, not the news story. That cut is on most-cited pages.
The 12% news figure is a conversion rate. About one in eight retrieved news URLs was cited. It is not news's share of ChatGPT.
Share on XVerification status
Rows are taken from the engine lists and from Ahrefs' channel table. No news top ten was assembled by averaging them. Figures whose original pages were not opened are named as omitted. This site has not classified a citation corpus into news and non-news.
Namdev, R. (2026). Most-cited news domains in AI search (v1). Retrieved from https://ritiknamdev.com/blog/most-cited-news-domains-ai-search Published under CC BY 4.0 — reuse freely with attribution.
The mixed lists that news brands actually appear in start at ChatGPT.