.png){kind=small}
Reuters and AP News are the most consistently cited news sources in ChatGPT, according to Peec AI's July 2026 analysis of 1,108 prompts and 8,883 chats in the U.S. Across nine news categories, Reuters appeared in the top five cited domains every time and ranked first in seven categories. The analysis also shows that ChatGPT's preferred sources vary significantly by topic, with specialist publishers and primary institutions becoming more prominent in areas such as technology, entertainment, and health.
When someone asks ChatGPT what happened in the news, the answer is no longer based only on the model's stored knowledge.
ChatGPT can search the web, expand the original question into additional searches, retrieve multiple sources, and then decide which information to use in its answer. OpenAI describes ChatGPT Search as a system designed to connect users with current information from the web while providing links and citations to the underlying sources.
That creates a new layer of competition for publishers.
A news organization is not only competing to appear in search results. It is potentially competing to become one of the sources an AI system selects and cites when constructing an answer.
The results reveal a strong role for established news organizations, but they also show that there is no single source hierarchy across every topic.
Key takeaways
- Reuters and AP News have the strongest overall presence in the analysis, appearing in the top five cited sources across all nine categories.
- Reuters ranked first in 7 of the 9 categories.
- The strongest Reuters result was International Politics, at 38% of citations.
- AP News ranked first in Crime & Safety, at 27%.
- Specialist sources become more important in narrower categories. The Verge appears in Technology, ESPN in Sports, Variety and People in Entertainment, and the CDC and FDA in Health.
- Reuters and AP News together accounted for 58% of citations in International Politics, but only 14% in Entertainment.
- ChatGPT searched the web in 100% of the tracked news cases in the analysis.
- 94% of fan-out queries included "2026", even though only 1% of the original prompts explicitly mentioned the year.
- The broader AI search ecosystem shows that being retrieved and being cited are separate outcomes. Peec AI's analysis of more than 1 million citations found substantial concentration among highly cited URLs.
- AI is becoming a more established way of accessing news: 10% of people across 48 markets reported using AI chatbots for news weekly in 2026, up from 7% in 2025.
The most cited sources by news category
The category breakdown below comes from a month-long analysis of 1,108 news prompts and 8,883 ChatGPT chats, tracking trending stories across nine news categories over repeated six-hour intervals.
The clearest way to understand the research is to compare the categories directly.
Source: Malte Landwehr, Linkedin
The percentages represent each domain's share of citations within its category. The table shows the five highest-cited domains displayed in the analysis, rather than all cited domains.
Reuters and AP News show the strongest cross-category consistency
The most obvious pattern is the breadth of Reuters and AP News.
Reuters ranks first in Business & Economy, International Politics, Sports, U.S. Politics, Technology, Health, and Climate & Weather.
AP News takes first place in Crime & Safety and remains in the top five in every other category.
Looking across the nine categories, their combined citation share is particularly high in politically focused news:
This variation is important.
It shows that Reuters and AP News do not simply dominate every subject to the same degree. Their strongest concentration appears in categories where broad, current, factual reporting is especially important.
By contrast, entertainment and technology have much more room for specialist sources.
Specialist sources become more visible in niche categories
The category comparison reveals a second pattern: topic expertise changes the source mix.
Technology is a good example.
Reuters leads the category at 12%, but The Verge follows at 8%, while Space and NASA each account for 4%.
Entertainment is even more specialized. People leads with 15%, followed by EW at 10% and Variety at 6%.
This suggests that broad publisher authority does not eliminate the importance of topical relevance.
A source can have enormous overall visibility but still face strong competition from a specialist publication when the query is tightly focused on a particular subject.
For GEO, this has an important implication: building authority across an entire domain is useful, but earning recognition in a specific topic area can matter just as much for the queries that actually concern that topic.
Health shows the importance of primary sources
Health has a notably different structure.
Reuters leads with 15%, followed by AP News at 10%, but the CDC and FDA account for 7% and 5% respectively.
That means institutional sources are competing directly with major publishers for citation share.
This is understandable in topics where answers depend on medical recommendations, regulatory decisions, public-health information, or official guidance.
The broader lesson extends well beyond healthcare.
AI systems do not always need another article summarizing a fact. Sometimes the strongest supporting source is the organization that created, published, regulated, measured, or documented the information in the first place.
Depending on the subject, that could be:
- a government agency
- a regulator
- a scientific institution
- an official company source
- a standards organization
- a research institution
- a database or original dataset
For GEO, original information can therefore be more valuable than simply producing another secondary article.
Citation concentration is not evenly distributed
The news analysis fits into a broader pattern seen across AI search.
Peec AI's analysis of more than 1 million AI citations found that on ChatGPT, 31% of URLs in its dataset had a citation rate of 2.0 or higher, and these URLs accounted for 59% of all citations.
In practical terms, a relatively small portion of pages can account for a disproportionately large share of AI citations.
That helps explain why publishing large volumes of content does not automatically translate into AI visibility.
A site might have thousands of indexed pages while only a much smaller group repeatedly gets used as evidence.
The strategic question therefore shifts from:
How much content do we publish?
to:
Which content becomes useful enough to cite repeatedly?
Being found by ChatGPT is not the same as being cited
This distinction is critical when measuring AI visibility.
A source can be discovered during search without appearing in the final answer.
Peec AI describes a retrieval as the point at which a webpage enters the candidate pool, while a citation occurs when the page is actually used and referenced in the generated answer.
That creates several different levels of visibility:
These stages should not be combined into one metric.
For example, a page could be retrieved frequently but have a lower citation rate. Another might receive relatively few retrievals but perform extremely well for a narrow set of highly relevant prompts.
This is why GEO reporting increasingly needs to distinguish retrieval visibility from citation visibility.
ChatGPT searches beyond the words users type
News queries also show how much the AI's own research process matters.
In the July 2026 study, web search occurred in 100% of the tracked cases.
The analysis also found that 94% of fan-out queries contained "2026", despite the original year appearing in only 1% of prompts.
This illustrates a fundamental difference between conventional keyword research and AI search.
A user might ask:
"What happened with the latest trade announcement?"
The AI system still has to work out what "latest" means.
It may need to establish:
- which announcement the user is referring to
- when it happened
- what changed
- which organizations are involved
- what the latest reporting says
- whether official information exists
- whether newer developments have superseded earlier reports
The user's prompt is therefore only the starting point.
The AI is effectively constructing its own research path before producing the answer.
Why current context matters so much for news
The heavy use of date terms in fan-out queries is particularly relevant to publishers.
News has a short shelf life compared with evergreen content.
A page about an event from six months ago may still be useful as historical context, but it may be inappropriate as the primary source for a question about what is happening today.
This makes several types of date information useful:
- Publication date tells readers when the article first appeared.
- Update date indicates when the page was changed.
- Event date tells the reader when the underlying development occurred.
- Current status explains whether the situation described on the page is still true.
These should not be treated as interchangeable.
Simply changing a publication date without materially updating the content does not make information current.
For news publishers, genuine freshness means keeping the factual state of the story accurate and understandable.
AI is becoming part of the news discovery journey
The citation data matters because people are increasingly using AI systems as part of their news consumption.
The Reuters Institute's Digital News Report 2026 found that 10% of respondents across 48 markets used AI chatbots for news during the previous week, compared with 7% in 2025. Among people under 35, the figure was 16%.
However, AI has not replaced traditional news sources.
Only 1% said AI was their main source of news, suggesting that chatbots currently function more as an additional layer in the news ecosystem than as a complete replacement for publishers.
People are using AI for functions that are especially suited to conversational search, including asking follow-up questions and getting explanations.
The Reuters Institute found that 42% of AI-news users cited the ability to ask follow-up questions as a reason for using chatbots for news.
This creates a different relationship between a publisher and its audience.
A person may encounter a news story through an AI-generated summary, ask for more context, compare competing accounts, and then decide whether to open the original source.
The publisher is no longer necessarily the first destination.
It can instead become the evidence layer behind the conversation.
What makes a source useful in AI-generated news answers?
The available research does not establish a single formula for earning a ChatGPT citation.
The category data suggests a combination of factors is likely involved, including relevance, authority, freshness, source type, and the ability of a page to directly support the claim being answered.
For publishers and brands, several principles stand out.
Original reporting
Information that comes directly from the source of an event or development can provide stronger evidence than a page simply repeating other coverage.
Clear factual statements
Pages that make important information explicit and easy to identify can be easier to use as supporting evidence.
Topical expertise
The category data shows that specialist sources can become highly visible when their expertise closely matches the subject.
Current information
For rapidly changing topics, current reporting and accurate update practices are essential.
Strong source reputation
The repeated appearance of established publishers across categories suggests that source reputation matters within news search, although the data does not establish exactly how much weight any individual factor receives.
Accessible content
A page cannot contribute to AI search if the relevant information cannot be accessed. OpenAI states that websites can choose to appear in ChatGPT Search and provides guidance for publishers on allowing its search crawler to access content.
What this means for GEO strategy
The biggest lesson is that GEO should not be treated as simply "SEO for ChatGPT."
Traditional search visibility and AI citation visibility overlap, but they answer different questions.
SEO might focus on:
Can this page rank for the query?
GEO needs to additionally consider:
Can this page be retrieved for the questions AI systems generate around the topic?
And then:
Does it provide information worth citing in the final answer?
For news publishers, that means monitoring more than rankings and traffic.
A useful measurement framework can include:
This makes it possible to distinguish between being present in AI search and actually influencing the answer.
The most important insight: source selection is contextual
The biggest mistake would be to take the Reuters result and conclude that ChatGPT simply prefers one fixed group of domains.
The category-level data shows otherwise.
Business and politics are heavily concentrated around large international news providers.
Technology introduces specialist technology and science sources.
Entertainment shifts toward entertainment publishers.
Health brings government institutions into the top five.
This tells us that source selection is contextual.
AI systems need different kinds of evidence for different questions.
A publisher looking to improve AI visibility should therefore understand the source landscape of its own topic rather than copying the strategy of a completely different industry.
The most important source for one category may be a global news agency.
For another, it may be an industry publication.
For another, it may be a government agency or original research organization.
Conclusion
The latest research into ChatGPT's news citations provides a useful snapshot of how AI search is selecting information in 2026.
Reuters and AP News have an unusually broad presence, but the category-level comparison shows that their dominance is not uniform. Specialist publishers become more important in technology and entertainment, while primary institutions such as the CDC and FDA appear prominently in health.
The broader lesson is about how AI chooses evidence.
ChatGPT can search beyond the user's original wording, introduce additional context such as the current year, retrieve multiple sources, and then cite only a subset of what it finds. As AI becomes another pathway through which people discover and understand news, that selection process becomes increasingly important for publishers.
For GEO, the objective is therefore not simply to be visible.
It is to create information that is relevant enough to retrieve, authoritative enough to trust, specific enough to use, and useful enough to cite.
Omnius helps brands combine SEO and GEO to understand how AI search discovers, evaluates, and cites their content - and turn those insights into a measurable search visibility strategy.
FAQ
Does ChatGPT cite the biggest news websites for every topic?
No. The July 2026 data shows strong cross-category performance from Reuters and AP News, but specialist sources become more prominent depending on the subject. Technology, entertainment, and health all show noticeably different source mixes.
Why might two highly authoritative websites receive different citation rates?
Authority is only one potential factor. A source may be highly reputable but less directly relevant to a particular claim, while another page may provide more specific, current, or original information. AI search also works from individual queries, so citation patterns can change substantially from one prompt to another.
Can original research help a company earn AI citations?
Yes, original research can give a company information that other sources do not have. Proprietary data, surveys, studies, benchmarks, original experiments, and industry statistics can all create citable information. There is no guarantee that original research will be cited, but it gives AI systems a more distinctive reason to reference the source.
Why are specialist sites important if Reuters and AP News dominate?
Because AI search is contextual. A general news provider can cover many subjects, but a specialist publication may have deeper reporting, expertise, or original information for a particular query. The category comparison shows this clearly: The Verge appears strongly in Technology, while People, EW, and Variety are prominent in Entertainment.
How should publishers measure AI visibility?
AI visibility should be measured across several stages rather than through one metric. Retrievals show whether content is being found, citations show whether it is being used as supporting evidence, and AI referral and conversion data show whether that visibility results in measurable audience or business value. Peec AI's broader research also demonstrates that citation performance can be highly concentrated among a smaller group of URLs.
.png)
.svg)
.svg)
.svg)
