You ask ChatGPT about your category and the same large publisher is cited near the top again. The usual advice is to write better content than they did. Sometimes that advice is aimed at the wrong target, because the publisher has a signed licensing agreement with the company behind the assistant, and no page you write changes a contract. The way to tell is to ask the same question on an assistant that signed nothing and see whether the publisher still leads. We also counted this across 31 of our own stored audits. For the finance, legal and home-services businesses we audit, large publishers showed up in a small share of answers, and they showed up less on ChatGPT and Copilot than on Perplexity and Google.
What the licensing study measured
In August 2026, Press Ranger and OtterlyAI published a study of AI citations and publisher licensing deals. They took 129.3 million citations captured in June 2026 across seven AI engines (19,812,878 cited URL records) and joined them to 91 confirmed licensing agreements covering 314 publisher domains.
On ChatGPT, licensed publishers received 48% more citations per cited page than unlicensed ones: 10.2 against 6.9. On Copilot the gap was 85%. Publishers that signed with OpenAI and nobody else received 112% more on ChatGPT. On Google AI Overviews, Google AI Mode, Gemini and Perplexity, the licensed publishers sat at or below the unlicensed ones.
The authors state that this is an association and that they have not shown the deals cause the extra citations. Large publishers with deals may simply be the ones assistants already preferred. It is one month of data. News is 7.2% of the citations they track, so most AI answers are outside its scope.
The engine split is the useful part
The gap appears on ChatGPT and Copilot, where deals were signed, and is absent on the Google products and on Perplexity. That gives a business owner a test that costs nothing. Ask your buying question on ChatGPT, then on Perplexity, then in Google's AI Mode. A publisher that leads on all three is probably there because its page ranks well and answers the question, and that is a position other sites can compete for. A publisher that leads on ChatGPT and Copilot and is missing elsewhere deserves a second look before anyone sells you a plan to beat it.
One answer proves little in either direction, since the same question returns different sources on different runs. Run the comparison several times before reading anything into it.
What we found in our own audits
The study covers news publishers. Our clients are mostly finance, legal and home-services businesses, so we checked whether the same split shows up on ordinary buying questions. On September 15, 2026 we scanned all 31 audits we have stored. 30 of them had cited-domain data. We counted the runs in which an assistant cited at least one domain from a list of 43 large news and business-media publishers, such as nytimes.com, forbes.com, wsj.com and cnbc.com.
The licensing split did not appear. Large publishers were cited more often on the two engines without the deals, which is the opposite direction from the study. The hits were concentrated in our finance audits. Several audits outside finance showed no large-publisher citation on either side.
This does not contradict the study, because it measures something else. The study asks how many citations a licensed publisher gets once it is cited. We asked whether a large publisher appears in a run's cited sources at all. Our list of 43 domains is a rough proxy we assembled, and it is a different list from the 314 licensed domains in the study. The large-publisher hits in most individual audits are in single digits. We read the result as the absence of a visible licensing effect in our client categories and nothing stronger.
Which domains do the assistants cite for your buying questions, engine by engine? The free audit records the cited sources for every run.
Get my free auditWhat to do if a big publisher is above you
Start by checking how often it happens. In our audits a large publisher was cited in 6.4% of ChatGPT and Copilot runs. The rest of those runs cited no domain on our list. The sources they did cite hold positions a business can compete for with the levers that have evidence behind them.
If a large publisher does lead your category, run the engine comparison above. When it leads everywhere, treat it like any strong competitor and look at what its page offers that yours lacks. When it leads only where a deal exists, do not budget for displacing it. We cannot promise to move a licensed publisher out of an answer, and you should be wary of anyone who does. The realistic goals are to be named in the same answer, to be cited on the engines where the publisher is absent, and in some categories to be mentioned in the publisher's own coverage.
This is a question to put to any agency quoting you AI visibility work: which of the domains above me hold licensing deals, and do they also lead on the engines that signed none? An agency that has not looked may sell you six months of content work against a contract. More questions like it are in questions to ask an AI SEO agency.
See who is cited above you, per engine
The free audit asks the assistants your buyers' questions and lists every domain they cite, split by engine, so you can see whether the names above you are the same everywhere.
Get my free auditSources
- Press Ranger with OtterlyAI, AI citations and news licensing deals (August 21, 2026; June 2026 data): otterly.ai/blog/ai-news-licensing-deals
- EZ Web, count of large-publisher citations across 31 stored AI visibility audits, run September 15, 2026. Method and limits are described on this page. The audit protocol is on the methodology page.