OpenAI Licensing Deals Buy a 48% ChatGPT Citation Premium. Nobody Else's Do.

Publishers with an OpenAI licensing deal earn 48% more citations per page on ChatGPT. Google's and Perplexity's licensees get nothing on their own platforms. A joint study by Press Ranger and OtterlyAI, announced via GlobeNewswire on 20 August 2026, matched 129.3 million June 2026 AI citations against every confirmed licensing agreement between AI companies and news publishers. Publishers signed with OpenAI exclusively earned 112% more. The asymmetry, not the premium, is the finding — and the whole thing is a vendor press release with no methodology paper behind it, so read the direction and hold the decimals loosely.

Key facts

Key takeaways

The dataset is citation-side rather than survey-side: 129.3 million citations collected in June 2026 across ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Microsoft Copilot, Gemini, and Claude, matched against the public record of licensing agreements between AI companies and news publishers.

The normalization matters. The study compares citations per cited page, not total citations — which strips out the most obvious objection, that licensed publishers are simply bigger and publish more. A page-level premium says that when a licensed publisher's page and an unlicensed publisher's page are both in the citable pool, ChatGPT returns to the licensed one more often.

Publishers signed with OpenAI and nobody else did better still, at +112% on ChatGPT against unlicensed publishers — the study reports the premium for that cohort without the underlying per-page rates.

The exclusivity result is the sharpest edge in the data: more than double the per-page citation rate of unlicensed publishers on ChatGPT, well above what the same publishers averaged elsewhere. Whatever mechanism connects a deal to citation behavior, it concentrates where the deal is.

If licensing deals simply made content more available — cleaner feeds, better crawl access — you'd expect every licensor's platform to favor its own licensees. That is not what the data shows.

Google's result is consistent with everything else known about AI Overviews: citation there tracks organic search systems, and rank and AI citation already come apart in measurable ways — a licensing agreement evidently doesn't reroute that pipeline. Perplexity's parity result is consistent with its revenue-sharing model, which pays publishers when cited rather than privileging them in retrieval.

ChatGPT is the outlier, and the study stops short of claiming intent. A deal typically bundles crawl permissions, structured content feeds, and sometimes prominence features, so the premium could be plumbing rather than favoritism. But the practical consequence is identical either way: the engine with the largest share of AI search activity has a measurable, deal-correlated skew in whose pages it cites.

The distribution effect compounds it. OpenAI-licensed publishers now take 57.9% of all their AI citations from ChatGPT alone, while unlicensed publishers keep a more balanced mix across ChatGPT and Perplexity. Thomas Peham, CEO of OtterlyAI, put it plainly:

Almost nobody reading this can sign a content deal with OpenAI. As of August 2026, the roughly 40 partnerships OpenAI has struck skew heavily toward large English-language premium news organizations, per Press Gazette's analysis of Tollbit's State of the Bots data. The same analysis found licensed publishers enjoy roughly seven times the clickthrough rate from ChatGPT — the premium extends past citation into traffic.

But the premium attaches to pages, and your brand can appear on those pages. Third-party sources already dominate where AI citations come from; this study adds a weighting to that map. A feature in an OpenAI-licensed outlet now sits on real estate ChatGPT structurally favors, while the same story in an unlicensed outlet of equal authority does not. For anyone allocating a PR budget with answer engine optimization in mind, that is a concrete, per-engine input into target selection — and a reason the licensed-outlet list is worth keeping next to your media list.

The equally useful half of the finding is negative: on AI Overviews and Perplexity, the licensed press has no edge. The study's own conclusion is that a deal "is not the only way to win" and that the smarter move for most organizations is earning coverage in the vertical publishers engines already trust. Trade press, category-specific outlets, and community sources remain the accessible tier — the sourcing mix differs sharply per engine, and none of the non-ChatGPT engines pays a deal premium at all.

The deals are announced publicly, but no maintained, category-filtered list exists — and a list of roughly forty national news organizations is not the same as a media plan for a B2B software category. Build your own, and keep it next to your existing media list:

The last row is the reason to keep the list rather than build it once. The useful half of this study's finding is negative: on AI Overviews and Perplexity there is no premium at all, so an outlet list that ignores which engine it is serving is already leaving value on the table.

The discounts here are larger than usual, and worth stating in order.

This is a press release from two companies with products to sell — Press Ranger distributes press releases and Otterly sells AI search monitoring — and "earned media drives AI citations" is a conclusion both benefit from. That doesn't make the numbers wrong, but no methodology paper accompanies them, nothing is peer-reviewed, and the matching of "comparable unlicensed publishers" is not documented.

The design is correlational with a built-in selection effect: OpenAI didn't license random publishers, it licensed prominent ones. Per-page normalization controls for volume, not for the underlying citability that made those publishers worth licensing in the first place. Some fraction of the 48% was there before any contract was signed.

The window is one month. Our own measurement of citation volatility found the cited-domain set for a prompt turning over 60–70% day to day, so a June snapshot is a photograph of a fast-moving subject.

And a citation is not an endorsement of accuracy. The Tow Center study covered by TechCrunch found ChatGPT misattributing and inventing citations even for publishers with licensing deals. More citations per page says nothing about whether the answer described the page — or your brand — correctly.

The clean version of the story — pay OpenAI, get cited — is not what the data shows. The messy version is more useful: one engine exhibits a large, deal-correlated citation skew; two engines exhibit none; and the deals themselves are reshaping where publishers' AI visibility concentrates. For brands, the actionable layer is that earned media now has engine-specific exchange rates, and the outlets that clear best on ChatGPT are, for the first time, a published list.

Elmo is an open-source, self-hosted AI visibility platform that runs your prompt sets across ChatGPT, Claude, Gemini, Perplexity, and Google's AI surfaces, recording every citation with its full URL, every mention, and the complete answer text. Because the data is yours to query, checking which engines your citations concentrate on — and whether licensed outlets are carrying your name into ChatGPT answers — is a query over your own results rather than a vendor report.

For the fundamentals, start with what AI citations are and where AI citations come from. For the practice, see the AEO guide and how to track your brand in AI search. For the vocabulary, the AI search glossary.