The Confusing Getty Images/Perplexity Deal

Last week, the photo licensing company Getty Images announced a new multi-year deal with the AI company Perplexity. 

Though AI licensing deals have become increasingly common, this deal stands out for several reasons.

First, the deal doesn’t involve AI training. Since Perplexity doesn’t train its own models, this isn’t a license to train AI generators using Getty content. For their part, Getty launched its own image-generating AI based on its library of licensed content. 

Second, the deal is heavily focused on attribution and image licensing. Perplexity will gain access to Getty’s application programming interface (API) and will be able to access both the images Getty licenses and the metadata around them.

According to Getty, Perplexity will be allowed to show Getty-licensed images in its search and discovery tools. However, those images will include both attribution and instructions for licensing them correctly.

On the surface, this seems like nothing but a good thing for Getty. The company had been working with Perplexity for over a year. It was already a part of Perplexity’s Publishers’ Program, which shares ad revenue with publishers when their content appears in search queries.

However, shareholders appear to disagree. Their stock dropped by more than 6% after the licensing deal was announced. A decline that hasn’t been helped by a recent legal defeat in their UK case against Stability AI.

But even without the recent legal defeat, the deal between Perplexity and Getty is confusing. It is both unnecessary and unhelpful. Worst of all, it’s with a company that has already established it doesn’t care about attribution.

A Strange Deal

The benefits of the deal are seemingly straightforward. Perplexity can and will showcase Getty Images’ photos in its results. For Getty, those images will be properly attributed and include licensing information.

This deal presumably applies across all of Getty’s sites, including iStock, Unsplash, and Photos.com. 

But nothing in this deal was impossible before. The Perfect 10 ruling found that using image thumbnails to aid search was already fair use. The recent Warhol decision didn’t significantly impact that ruling. 

As such, Perplexity was already free to include thumbnails of Getty’s photos in its results. As for attribution, nothing prevented Perplexity from providing it before now. Though having access to the API makes it easier, I refuse to believe an AI company couldn’t figure this out through other means.

A quick test of both Google Gemini and OpenAI’s ChatGPT showed that they could easily find images and provide attribution. They also both linked to pages where I could license the images I searched for. 

Though this deal may mean Perplexity can bake the licensing process into its AI tools, that’s a relatively small benefit. We’ve already seen how AI reduces website traffic by encouraging people to end their session. There’s no reason to believe that adding such a feature to AI results would have much of a boost.

In short, Perplexity either couldn’t or wouldn’t do this before, even though others can and have. But that’s not the biggest worry. Even among AI companies, Perplexity has long been one of the least trustworthy partners and has become famous for its shady practices.

A Dubious Partner

Over the past few months, Perplexity has earned a reputation as a company not to be trusted, repeatedly violating both online and academic norms.

In August, Cloudflare exposed Perplexity’s practice of ignoring robots.txt and using undeclared bots to index content on websites. To make matters worse, the company disguises its bot as a regular user on Safari on a macOS system when its main bot is blocked. 

Though the company provides an official IP range for its bots, these “stealth” bots use a different address range to access content that has been blocked. 

In short, even if you block all bots from accessing your site and then deliberately block Perplexity’s IP range, the company will still index your content and answer detailed questions about what is on your site, all of this under the guise of a regular user and from an undeclared IP address.

Then, in October, Perplexity began publishing ads encouraging users to cheat on their academic projects with their Comet browser. These ads entirely dropped the mask and said that “Comet wrote a whole assignment for me. #worksmarternotharder” and that the browser could help you “Ace your way through school.”

Since then, more ads have appeared, all with a similar message about encouraging academic dishonesty.

Getty Images has been a stalwart of the image licensing space for over 30 years. Is this really the company that they want to be hitching their wagon to? Especially since Perplexity is lagging far behind OpenAI, Microsoft and Google in the AI search race. 

This makes Getty’s choice to partner with Perplexity seem even more odd. That, in turn, may explain why its shareholders were not keen on the deal.

Bottom Line

Ultimately, this is an unnecessary deal with an untrustworthy partner. It will not significantly help Getty, and it’s likely an attempt to launder Perplexity’s reputation using Getty’s name.

However, that’s not going to work. This kind of tarnish doesn’t come off; it just taints everything that it touches. I come away from this thinking less of Getty in a way that I don’t with the New York Times, Reddit and other companies that have signed deals with AI companies. 

That’s because those companies at least gained something significant from the deals. They extracted major concessions from the AI companies they partnered with and 

The deal doesn’t benefit Getty to any great extent, and the recent reporting on Perplexity has painted it as the least trustworthy of the major AI companies. It’s a rare deal where everyone loses.

Though I sincerely hope I am wrong about this deal, all the indications point to it harming Getty far more than helping it.