A display-only deal that rewrites AI image search
The Getty Images OpenAI deal is a multi‑year display agreement that lets ChatGPT Search show Getty’s licensed photographs directly inside answers, promising richer visual responses while stopping short of granting full AI training rights and leaving contributor economics undisclosed. That definition matters because it draws a bright line: this is about what users see, not what models learn. OpenAI gains a licensed image channel inside ChatGPT Search, while Getty tests whether its vast archive can earn money in an AI‑first discovery world. In plain terms, this is a shift from scraping to signing contracts. It signals that, at least for one major library, AI image licensing agreements are no longer a hypothetical future; they are how search will work. But it also exposes how fragile creator rights AI training still are when the most important terms stay behind closed doors.
Markets love the model; creators are still in the dark
Investors treated the Getty Images OpenAI deal as proof that old‑guard photo libraries can plug into the AI economy. Getty’s stock shot up over 200 percent in premarket trading after the announcement, and later traded as much as 108 percent higher at $1.26 as optimism poured in. Another report notes the shares closed at $1.35, up 123 percent from $0.61 once the deal was public. That kind of spike is not a bet on art; it is a bet on licensing. Yet the same disclosures that thrill investors leave creators guessing. Financial terms remain private, and there is no public breakdown of contributor economics or any separate payments if training rights are added later. One quotable takeaway is this: “Investors treated a display‑license announcement as a test of whether Getty can convert its library into AI‑search revenue.” Photographers are left to hope those revenues reach them.
What changes for ChatGPT users and AI image search
For ordinary users, the impact is concrete: ChatGPT Search licensed images from Getty’s archive will appear directly in responses instead of sending people off to separate stock sites. Getty argues this will “significantly improve the richness of visual responses” and make search results more useful and trustworthy. OpenAI can now display professional photos and editorial images inline when you ask about events, people, or concepts, turning the chat window into a mixed media answer box. That matters in two ways. First, discovery of licensed photography becomes invisible; users may not even realize they are looking at stock, unless credits and source links are prominent. Second, it normalizes the idea that AI chat interfaces are also storefronts for licensed media. ChatGPT becomes not only a question‑answering tool but a front door to commercial photo archives. Whether that boosts paid licensing or dilutes brand identity depends on how clearly attribution and rights information are shown.
Display rights today, training rights tomorrow?
The most interesting part of this AI image licensing agreement is what it does not cover. Getty makes clear that the deal does not allow its images to be used to train OpenAI’s image generator, DALL·E. Other reporting says model‑training rights and any use of the library to teach or tune AI systems remain undisclosed. In other words, this is explicitly a display‑only arrangement: it governs what shows up in ChatGPT answers, not how models are built. That separation sounds reassuring for creator rights AI training, but only up to a point. Commercial terms, pricing, and contributor economics are all hidden, and no one outside the contract knows whether a future add‑on could quietly grant training rights. Model‑training rights remain central to interpreting any display‑only OpenAI arrangement. Until those rights are either granted or ruled out in detail, creators cannot tell whether they are licensing visibility, feeding the model, or both.
From lawsuits to licensing: a fragile win for creator rights
This deal also marks a strategic pivot for Getty. After suing another AI developer in 2023 for allegedly scraping more than 12 million images without permission, and largely losing that legal fight, the company appears to have concluded that court battles are a weak defence for its catalogue. One source notes that “having tested the courtroom route and found it wanting, Getty appears to have concluded that licensing is the surer path to getting paid.” At the same time, Getty’s catalogue of about 609 million images has not insulated it from fears that AI generators could erode demand for traditional photo libraries; its shares had fallen by more than half before this deal. So the Getty Images OpenAI deal is both survival strategy and precedent. It shifts the conversation from infringement to negotiated access, but does not yet deliver a transparent, enforceable model for how photographers share in AI‑driven revenue. The conclusion is blunt: licensing beats scraping, but opaque deals are a thin foundation for long‑term creator trust.
