How AI Images Changed Stock Photographer Earnings

Search for what AI has done to stock photography and you will find two stories, both wrong. One says the industry collapsed and photographers are finished. The other says nothing really changed. The published numbers say something more specific and more useful: the stock market split into two halves that are now moving in opposite directions, and which half your work sits in matters far more than how much of it you shoot. Here is what the filings, the library counts and the independent research actually show — and what a working contributor should do about it.

First, what actually happened to the money

The doom headlines usually claim the agencies themselves are dying. They are not — and it matters that we get this right, because the real problem is somewhere else entirely and you cannot fix a problem you have misdiagnosed.

Getty Images finished 2025 with revenue of $981.3 million, up 4.5% year over year. Shutterstock finished the same year at $989.9 million, up 6%, which the company described as record revenue. Getty paid out more than $220 million in contributor royalties in 2025. Those are not the numbers of an industry that has collapsed.

But the headline figure hides the story. Both companies report their stock library separately from everything else, and once you look at the segments, the picture changes completely.

Getty splits its business into Creative — the concept and lifestyle imagery that makes up most of what microstock contributors shoot — and Editorial, which is real events, news, sport and entertainment. In 2025 Creative grew 0.7%. Editorial grew 6.9%. In the first quarter of 2026 the gap opened wider: Creative fell 4.5% (and 8.0% on a currency-neutral basis, which strips out the flattering effect of a weaker dollar), while Editorial rose 11.0%.

Shutterstock tells the same story with less nuance because it does not split editorial out. Its Content segment — the core stock library — grew 4% across 2025 and then fell 12% in the first quarter of 2026, with total company revenue down 18% and a net loss of $47.6 million.

One more data point that is easy to miss: roughly 75% of Getty's 2025 revenue came from exclusive content. Three quarters of the money is being paid for images the buyer cannot get anywhere else. That is the whole thesis of this article in a single statistic.

The supply side: the shelves filled in three years

The demand side moved slowly. The supply side did not. Adobe Stock began accepting labelled AI-generated content in December 2022. By May 2023, AI images were 2.5% of its image library. By April 2025 — according to a count published by CineD and widely reported since — they were 47.85%.

The single most striking comparison in that data: photographers took roughly two decades to collectively upload about 313 million images to Adobe Stock. AI submissions matched that number in under three years, at a peak rate of around 29 million new images a month.

Adobe reacted in May 2025 by capping how much a contributor can submit — limits reportedly tied to past acceptance rate and sales performance — and contributors began reporting acceptance rates falling sharply, with similar content already in our collection becoming the most common rejection. The volume strategy that worked in microstock for fifteen years stopped working, and it stopped working for everyone, not only for AI submitters.

Why your per-file earnings fell even though the agencies grew

Here is the mechanism most of the panic articles miss. An agency's revenue and an individual contributor's revenue are two different numbers, and generative AI pushed them in different directions at the same time.

Think of the library as a shelf and licensing revenue as the money walking past it. If the money walking past grows 4% in a year while the number of items on the shelf grows by tens of percent, the average revenue per item falls — even though the agency had a record year and nobody cut your royalty rate. Nothing was taken from you directly. The denominator simply exploded.

This is arithmetic rather than a measurement, so treat the illustration below as exactly that. But it explains why a contributor whose portfolio and royalty percentage did not change can still watch monthly income slide for two years while reading press releases about record revenue.

Two consequences follow, and they point in the same direction. First, adding more near-identical files to the shelf no longer helps you; it actively hurts, because your own files start competing with each other for the same query. We wrote about that failure mode in detail in why near-identical shots need different keywords. Second, discoverability stopped being a nice-to-have. On a shelf of 763 million assets, an image that cannot be found does not exist.

Where a brief goes now

The segment numbers are not random. They describe a fork that now sits in front of every commercial image brief, and the fork explains both the Creative decline and the Editorial growth.

When a buyer needs a smiling team around a laptop, an abstract technology background, or a generic bowl of pasta on a wooden table, a model can now invent something adequate in seconds. Adequate is the operative word: the buyer was never emotionally invested in that image, they needed a rectangle that communicates "collaboration" and moves on. That work is gone, and it is not coming back.

When a buyer needs the actual stadium on the actual night, a named landmark that has to be recognisably itself, a real product with a real licence behind it, or a released person in a documented situation, a generated image is not a substitute — legally or factually. Someone had to be there with a camera. That is why Editorial is the fastest-growing line on Getty's income statement, and it is the clearest signal in the whole dataset about what to point a camera at.

What independent research says about creative income

The agency filings tell us about revenue at the top. For what happens to the people underneath, the best evidence comes from outside the stock industry entirely — because researchers can observe freelance marketplaces in a way they cannot observe microstock portfolios.

• Freelance image work fell measurably. A study by Hui, Reshef and Zhou, published in Organization Science, tracked a large online freelancing platform after the release of DALL·E 2 and Midjourney and found a 17% drop in job posts related to image creation, alongside a 21% drop in writing and coding work after ChatGPT. Affected freelancers saw both employment and earnings fall. Notably, top performers were hit disproportionately hard — being good was not, by itself, protection.
• Forecasts for adjacent creative sectors are in the same range.A 2024 economic study commissioned by CISAC, the international body for authors' societies, projected that 24% of music creators' and 21% of audiovisual creators' revenues are at risk of being lost to generative AI by 2028. Those are different industries, but the mechanism — cheap synthetic substitutes competing with licensed human work at the commodity end — is identical.

Put the segment data and the research side by side and a consistent range emerges: roughly a fifth to a quarter of commodity creative income is exposed. Not everything. Not nothing. And concentrated almost entirely in work where the buyer cannot tell, and does not care, who made the image.

The courts are not going to fix this for you

Many contributors are waiting for a legal outcome that resets the board. It is worth being realistic about how that is going.

On 4 November 2025 the English High Court handed down judgment in Getty Images (US) Inc & ors v Stability AI Ltd— the first UK ruling on copyright and generative AI training. Getty largely lost. The court rejected the secondary copyright infringement claim, holding that the Stable Diffusion model is not itself an "infringing copy", so importing or distributing it in the UK did not infringe. Getty won only an extremely limited trade mark finding relating to watermarks reproduced by early model versions.

Crucially, Getty accepted during the proceedings that the training itself did not happen in the UK, so the central question — whether training a model on copyrighted images is infringement — was never decided. Other cases in other jurisdictions continue. But nothing about this ruling suggests a quick judicial rescue, and planning a career around one is not a plan.

Who is still getting paid, and for what

Alongside licensing, a second revenue line has appeared: agencies selling access to their catalogues as AI training data, and passing a share back to contributors. It is real money, and it is worth understanding rather than dismissing.

• Shutterstock's Contributor Fund.When Shutterstock licenses catalogue data for model training, participating contributors share a payout set at a 20% average corporate royalty rate of the revenue Shutterstock receives. It shows up as "Data Licensing" in the earnings summary. This line grew 16% across 2025 to $203.3 million of company revenue — though it fell sharply in Q1 2026 on deal timing, which is a reminder that it is lumpy, not dependable.
• Getty's own generator. Getty built a model with Bria trained exclusively on licensed content, sold as commercially safe with indemnification, and says it compensates contributors on a recurring basis for their creative visuals being used in training.
• Getty and iStock do not accept generated images from contributors. You also may not use AI-generated material as a visual reference for a submitted file, or as a prop or background element within one, because you must own the copyright in everything in the frame. Breaching this can terminate the contributor agreement. Adobe Stock takes the opposite approach and accepts generated content if it is labelled. Know which platform you are on.

One structural change worth noting: the proposed Getty–Shutterstock merger, announced in January 2025, did not happen. After the UK Competition and Markets Authority conditioned clearance on selling Shutterstock's editorial business, Getty's board resolved in June 2026 not to proceed and terminated the agreement. Contributors who had been bracing for a single combined agency rewriting royalty terms can put that particular worry down.

What to actually do about it

Everything above points in the same direction, and it is a narrower direction than "shoot better photos". Six moves, in the order we would make them.

1. Stop competing where the model wins. If a competent prompt produces something a buyer would accept in place of your image, that image is a poor investment of a shooting day. Isolated objects on white, abstract concept backgrounds, and generic staged lifestyle are the clearest examples.
2. Shoot what has to be verifiable. Real events. Named places that must be recognisably themselves. Real products and real interiors. Released people in situations a buyer needs documented rather than imagined. This is the only part of the market where the published growth rate is comfortably positive.
3. Go for specificity over generality."Woman drinking coffee" competes with a million files and a text prompt. "Woman drinking coffee at a window seat in a Belgrade tram in winter" competes with almost nothing, and it is the query a real buyer with a real brief actually types.
4. Treat exclusivity as a commercial decision, not a loyalty one.Three quarters of Getty's revenue comes from exclusive content. If your work is genuinely differentiated, exclusivity is where the higher rates are — our Getty vs iStock royalty comparison lays out the actual percentages so you can do the maths on your own portfolio.
5. Abandon volume as a strategy.Upload caps and "similar content" rejections have made mass submission counterproductive. Fewer, more distinct files with genuinely different metadata now beat large near-identical sets.
6. Take metadata as seriously as the shot. On a library approaching a billion assets, search placement is distribution. Correct keyword order, agency-correct vocabulary and titles that match how buyers phrase briefs are no longer housekeeping — see how many keywords to use and the Adobe Stock contributor guide for the specifics.

What we would not do

We would not pivot a photography business into generating AI stock. The supply chart above is the reason: that shelf filled to nearly half a library in under three years, upload caps arrived, and the marginal AI image now competes with hundreds of millions of others made the same way, at zero marginal cost, by anyone with a subscription. Whatever advantage existed there in 2023 has been competed away. The scarce asset — the one the numbers keep rewarding — is being somewhere real with a camera and a signed release.

Sources & further reading

• Getty Images — Fourth Quarter and Full Year 2025 Results (revenue by segment)
• Getty Images — First Quarter 2026 Results
• Shutterstock — Full Year 2025 and Fourth Quarter Results
• Shutterstock — First Quarter 2026 Results
• CineD — AI Flood on Adobe Stock: Nearly Half of All Images Now AI-Generated (also covered by PetaPixel)
• Hui, Reshef & Zhou — The Short-Term Effects of Generative Artificial Intelligence on Employment, Organization Science
• CISAC — Global economic study on generative AI and creators' revenues
• Getty Images v Stability AI [2025] EWHC 2863 (Ch) — analysis
• Shutterstock — Data Licensing and the Contributor Fund
• Adobe Stock — Generative AI content guidelines for contributors