Getty + OpenAI Bring Licensed Images to ChatGPT Search

Getty Images and OpenAI have struck a deal that puts Getty’s licensed editorial and stock photography inside ChatGPT search — and the most important word in that sentence is licensed. Announced on June 21, 2026, this is a multi-year display partnership: Getty’s content is meant to appear in ChatGPT’s answers with attribution, not be ingested to train OpenAI’s models.

The market read it as a rescue. Getty’s stock, which had fallen under a dollar earlier in 2026 and drawn an NYSE minimum-bid-price warning, surged somewhere between roughly 145% and 200% in a single session depending on the outlet and the moment of measurement. But the financial terms were never disclosed, the rollout timeline inside ChatGPT was never specified, and how much of the licensing money reaches Getty’s ~600,000 contributors remains an open question.

This guide explains what the deal actually is, why the distinction between display licensing and training licensing is the part marketers need to understand, and why it turns “visual GEO” from a talking point into a concrete distribution channel. Every number and quote below is sourced; where the reporting is uncertain, we say so rather than paper over it.

On June 21, 2026, Getty Images announced a multi-year display partnership with OpenAI that makes Getty’s licensed content libraries available inside OpenAI’s search and discovery experiences within ChatGPT. In plain terms: when ChatGPT answers a question that calls for a real-world image, the reply may include a Getty-licensed photograph to help illustrate the topic, rather than an AI-generated approximation or an unattributed scrape.

This is not Getty’s first AI display deal. In October 2025 it signed a multi-year global licensing agreement with Perplexity AI on similar terms — attributed, linked images without training use — which sent Getty shares up roughly 19% on the day. The OpenAI deal extends the same pattern to ChatGPT, a platform OpenAI has said reached on the order of 900 million weekly users by early 2026. The scale is the difference.

Most coverage explained this deal in financial and legal terms. The part that matters for marketers is a taxonomy almost nobody spelled out: there are two fundamentally different kinds of content deal in the AI era, and they change different things. A display licence means the platform shows your content — attributed and identifiable — inside its answers. A training licence means your content teaches the model, dissolving into its weights without ever appearing as itself.

The Getty–OpenAI deal is squarely the first kind. Getty’s images are meant to be shown and credited; the announcement is explicit that they are not being used to train OpenAI’s models. That precision matters because it determines what you actually see in search. Display licensing creates an attribution and discovery channel a brand can be present in. Training licensing — absent here — would mean the content shapes how the model talks, but produces no identifiable visual answer to optimize for.

Put the announced deals side by side and the display-versus-training distinction becomes a working reference rather than a definition. The table below tracks the publicly confirmed visual- and content-licensing deals relevant to AI answer surfaces as of late June 2026. We include only rows we can source; where a platform has not announced a comparable visual deal, it is left off rather than filled with a guess.

Two things jump out. First, the visual deals — both with Getty — cluster on the display side: attributed, no training, terms undisclosed. Second, OpenAI’s earlier publisher agreements (AP, Reuters, the Financial Times, the Washington Post, Axios and others) sit in a different column: text-centric, with training treatment that varies deal by deal. The Getty partnership is the most significant visual-content display arrangement OpenAI has announced at this scale — and that is exactly why marketers should treat images as a distinct GEO surface, not a footnote to text. For the mechanics of getting surfaced in AI answers in the first place, our generative engine optimization (GEO) guide is the companion read.

The price action was the headline everywhere, and it deserves a careful read. Coming into the announcement, GETY had fallen roughly 55% year-to-date, closing at about 61 cents on June 18 — low enough that the NYSE had issued a sub-$1 non-compliance warning back in March. On the news, reported single-session moves ranged from about 145% to 200% across premarket and early trading; Bloomberg framed the move as “more than 150%.” The outlets differ because they measured at different moments. By 12:44 PM ET, shares had settled to roughly +118% at $1.32 — still Getty’s best trading day since July 2022. Shutterstock, tied to Getty through a pending merger, rose more than 18% the same day.

The surge was relief: the AI threat had visibly become an AI opportunity. Analysts framed the shift in striking terms — Benchmark analyst Mark Zgutowicz described the deal as repositioning Getty as a licensed visual-content supplier to AI-native search platforms, moving part of the AI narrative from substitution risk toward monetizable distribution. Other commentary cast Getty’s transition as a move from a transactional stock-photo utility toward a high-margin intellectual-property tollbooth for generative search. The throughline: AI operators must pay for premium, authenticated visual content — a structural change from traditional stock licensing.

Here is the practical shift. Until now, generative-engine optimization has been overwhelmingly a text discipline: structure your content, earn citations, and hope ChatGPT or Perplexity quotes you. Visual presence in AI answers was mostly out of reach — the model either generated an image or pulled an unattributed one. A licensed display deal changes that. When ChatGPT chooses to illustrate an answer with a real photo, it now has a licensed, attributed source to draw from. Brands whose imagery is distributed through that library gain a visibility channel inside the answer itself.

That is what makes visual GEO concrete rather than aspirational. It sits alongside the text-citation work brands already track — the same discipline as understanding the factors that determine AI search citations and the parallel ecosystem where images appear in Google AI Overviews. Measurement extends too: the same muscle you use for brand citation tracking in AI search now has a visual dimension to monitor. Treating image distribution as a deliberate part of an agentic SEO and GEO program is the obvious next move.

There is a structural catch that almost no coverage named. Unlike text GEO — where a brand can publish a blog post on its own domain and hope to be cited — visual GEO through ChatGPT requires the imagery to live inside Getty’s licensed library. That makes Getty an intermediary gatekeeper for this channel. Brands that already distribute visual content through Getty, iStock, or the broader licensed ecosystem gain automatic eligibility; brands that self-host every image are, for now, invisible to it.

This is the genuinely new dynamic for marketers, and it cuts both ways. The upside is a fresh, defensible visibility surface for brands with strong, well-distributed visual libraries. The downside is dependence on an intermediary whose selection logic is opaque and whose own economics — debt, a pending merger, undisclosed creator splits — are in flux. The smart posture is to participate where it is cheap to do so, while keeping your owned-channel imagery strategy intact.

A clear-eyed view keeps four uncertainties on the table. None of them makes the deal less real; together they keep the hype honest and the planning sober.

• Creator compensation is genuinely unknown. Getty will receive licensing fees from OpenAI, but how much flows to the roughly 600,000 contributors whose work powers this deal has not been disclosed. Getty’s royalty structure has historically been contentious among contributors — worth watching, not assuming.
• Rollout timing was not specified. The announcement said Getty content “will appear” in ChatGPT; it did not say when. Do not tell stakeholders the feature is live today — frame it as announced and verify before building campaigns around it.
• Training rights are a future open question. Not part of this deal, but the announcement is silent on later phases. Precision protects you: “not part of the current deal,” not “permanently excluded.”
• Getty’s balance sheet still matters. A 150%-plus single-day surge does not erase roughly $2 billion in debt or settle the Shutterstock merger’s completion status. The stock story and the business story are not the same story.