If you've used ChatGPT to look something up this week — a celebrity, a news event, a sports moment — you might have noticed actual photographs appearing in the results. Not AI-generated illustrations. Real photographs with credits. That's new. It happened because of a deal signed last Sunday.

On June 21, 2026, Getty Images announced a multi-year display partnership with OpenAI. Getty's licensed photo library — over 400 million images covering sports, entertainment, editorial news, and stock photography — is now surfaced inside ChatGPT's search and discovery features. Getty's stock, which had been trading at 61 cents before the announcement, closed June 22 at $1.15, a 90% gain in a single session.

What's actually happening

Here's an object lesson, because "display partnership" sounds like a legal term and basically is.

Think about how Google Image Search works. You type something, images show up, they link back to their original source. That's display. Google isn't buying the photos — it's pointing you toward them. What Getty and OpenAI appear to have built is something similar: when ChatGPT answers a question that benefits from a photograph, it can now pull from Getty's library and show you a labeled, attributed image.

The word that matters is licensed. Before this deal, ChatGPT might have shown you an AI-generated image, or no image at all. Now it can show you a photograph that Getty owns the rights to, properly credited. Your brain reads "Getty Images" in the corner and knows where the image came from. That's different from what you were looking at six months ago.

What neither company will say is whether those images will also be used to train future OpenAI models. That's the harder question. Display means "show users this image." Training means "teach the AI using this image." These are legally distinct. The deal, per both companies' statements, covers display. Anything beyond that is undisclosed.

Getty's history here is worth knowing. For years, they were one of the loudest voices against AI companies using images without permission — they sued Stability AI in 2023 for training on Getty photos without a license. Their first AI deal didn't happen until October 2025, with Perplexity. This deal with OpenAI is the biggest reversal yet. Fifty-four cents a share, apparently, changes the calculus considerably.

Source spread

• Getty Images Newsroom — official announcement [hype] — covers the deal scope and "display" framing; stays silent on financial terms and training-data use.
• Bloomberg — market reaction [skeptic] — notes the unanswered training-data question and Getty's prior resistance to AI partnerships.
• Engadget — product analysis [builder] — focuses on how images surface in ChatGPT search and the user-facing implementation.
• AI Business — industry context [skeptic] — places the deal in the broader context of stock photo companies trying to survive the AI-generated image wave.

What's real:

• The photographs are credited. You can see where they came from. That's more transparency than AI-generated images, which don't come from anywhere in the way a photograph does.
• Getty needed this deal. Before the announcement, the company had shed roughly 55% of its stock value in 2026. A company that had been fighting AI tooth-and-nail cut a deal because the alternative was becoming irrelevant.
• "Display partnership" is a legally meaningful category. This is not the same as OpenAI training its next model on 400 million Getty photos — at least not as the deal is described.

What deserves a side-eye:

• Neither company will say whether Getty's content will train future models. That silence is the loudest thing about this announcement.
• The financial terms are entirely undisclosed. That means we don't know if Getty received a one-time fee, a per-image royalty, equity, or something else. We also don't know what OpenAI got in return — which, given the training-data silence, is worth asking.
• Getty sued Stability AI over training on unlicensed images. A display deal with OpenAI doesn't address whether OpenAI's existing models already trained on Getty images before any licensing agreement existed.

What to do about it

Three practical adjustments, none of which require any technical knowledge.

• Look for photo credits when ChatGPT shows you images. A "Getty Images" credit means the photo came from a real photographer who was paid. That's different from an AI-generated image, which doesn't come from anyone. The credit is a signal of provenance.
• For professional or commercial work, don't assume "licensed to ChatGPT" means "licensed to you." The deal covers what ChatGPT can show you. What you do with an image you see in a ChatGPT result is a separate copyright question. If you're using it commercially, you still need your own license from Getty.
• If you're a photographer or creative: the training-data question here is still open. Organizations like the American Society of Media Photographers are tracking the broader licensing landscape if you want to follow how this develops.

Further reading

• Getty Images Newsroom — official announcement — the primary company statement
• Bloomberg — Getty surges 145% after OpenAI deal — market reaction and deal context
• Engadget — ChatGPT image integration details — user-facing implementation
• AI Business — stock photography industry context — the broader competitive picture