%22%2F%3E%3Cstop%20offset%3D%22100%25%22%20stop-color%3D%22hsl(112%2C%2055%25%2C%2040%25)%22%2F%3E%3C%2FlinearGradient%3E%3C%2Fdefs%3E%3Crect%20fill%3D%22url(%23g)%22%20width%3D%221%22%20height%3D%221%22%2F%3E%3C%2Fsvg%3E)
If you've searched for stock photos recently, you've probably encountered AI-generated images without realizing it. Maybe several. Maybe most of them.
The major stock sites have spent the last two years quietly adjusting their AI policies. Some banned AI submissions while building their own generators. Some flooded their libraries with AI content. Some did both at the same time. The result is a landscape where the average buyer has no idea what they're getting.
Here's what's actually happening at each major stock site, what it means for you, and why we think the entire stock photo middle ground is evaporating.
The site-by-site breakdown
Shutterstock: ban uploads, sell their own
Shutterstock partnered with OpenAI in 2022, licensing its library for DALL-E training. They now offer an AI image generator built directly into the platform.
But here's the twist: third-party AI uploads are banned. Shutterstock's content policy says they can't verify model sources or ensure contributing artists are compensated, so contributors can't upload AI-generated images at all.
They'll happily sell you AI images from their own tool, though, at $15-29/month for 100 generations. Each comes with $10,000 in legal indemnification. The contributors whose photos trained the model get a cut from the Contributor Fund, but how much per artist? Unclear.
Adobe Stock: the AI flood
Adobe is the outlier. They're the only major stock site that actively accepts third-party AI submissions, and the results are staggering.
As of April 2025, 47.85% of Adobe Stock's image library was AI-generated: 313 million AI images sitting alongside 342 million photographs. Nearly half the library. Adobe's response? They quietly defaulted search results to "Exclude Generative AI" in January 2025, meaning buyers have to opt in to see AI content.
They require Content Credentials metadata and have the best provenance labeling infrastructure in the industry. But contributors report their acceptance rates have plummeted from 90% to below 20%. Adobe is quietly throttling AI volume while publicly maintaining the open-door policy.
The catch: Adobe built the best labeling system and the best AI tool (Firefly), then got overwhelmed by AI submissions anyway. The 47% stat tells you everything about where stock photography is heading.
Getty Images / iStock: banned it, then built their own
Getty banned AI-generated submissions in January 2023 and sued Stability AI, claiming 12 million+ Getty photos were copied to train Stable Diffusion.
Then they launched Generative AI by Getty Images, powered by NVIDIA and trained on their own library. Getty offers $50,000 indemnification per generated image; iStock offers $10,000.
The message to contributors: your photos are too valuable to let other AI companies use, but perfectly fine for us to feed into a generator.
And that lawsuit against Stability AI? The UK High Court largely sided with Stability AI in November 2025, ruling that model weights are not "infringing copies" and rejecting Getty's central copyright claim. The US case drags on.
Alamy: principled ban
Alamy bans AI-generated submissions entirely, classified as "unsuitable material," uploading them may lead to contract termination. They allow limited AI retouching (denoising) on real photos but prohibit generative fill or dramatic AI alterations.
They participate in the Fair Diffusion Program for ethical AI training, and may explore a separate, clearly-labeled AI collection in the future. But for now: no AI in the library.
Stocksy: the most consistent stance
Stocksy prohibits AI-generated content, and they mean it. Their ban extends to AI-generated props within images (framed art, magazine covers in the background) and they caution contributors against using generative AI editing tools at all.
As an artist-owned cooperative, Stocksy's contributors have actual governance power. Their 2026 visual trends report emphasizes human-created, authentic content. If you want to pay premium for guaranteed human photography, Stocksy is the honest version of that deal.
Dreamstime: AI in a box
Dreamstime accepts AI images but only in a dedicated "Illustration and Clipart > Generative AI" category. Tag an AI image as regular photography? Rejected. AI-generated realistic human faces? Rejected, because you can't provide a model release for someone who doesn't exist.
The category separation is smart but limited. Once downloaded and placed on a website, nobody sees the Dreamstime category tag.
123RF, Depositphotos, Pond5
123RF accepts AI submissions in a dedicated category and offers its own AI generator with $25,000 indemnification per image.
Depositphotos (now part of Vista/Canva) bans external AI uploads but offers its own generator powered by Bria.ai. Generated images stay exclusive to the customer and aren't added to the main library.
Pond5 (owned by Shutterstock) bans AI-generated content entirely, same reasoning as its parent company.
EyeEm: the cautionary tale
EyeEm shut down in January 2026. Before it died, Freepik (which acquired it from bankruptcy) updated the terms to license all 160 million user photos for AI training, with a 30-day opt-out window while deletions took 180 days. The entire library was captured for AI training. Community built, community extracted from, community abandoned.
The pattern
Here's what every stock site is telling you, whether they say it directly or not:
| Site | Accepts AI uploads? | Sells AI images? | Has own AI tool? |
|---|---|---|---|
| Shutterstock | No | Via own tool | Yes |
| Adobe Stock | Yes | Yes (47%+ of library) | Yes (Firefly) |
| Getty/iStock | No | Via own tool | Yes |
| Alamy | No | No | No |
| Stocksy | No | No | No |
| Dreamstime | Yes (restricted) | Yes | No |
| 123RF | Yes (dedicated category) | Yes | Yes |
| Depositphotos | No | Via own tool | Yes |
| Pond5 | No | No | No |
The dominant strategy: ban outside AI, build your own. The platform keeps the margins. Contributors get a fund. Buyers get AI images at stock photo prices.
The uncomfortable economics
Stock sites are selling AI-generated images at the same prices as human-shot photography. A Shutterstock subscription costs the same whether you download a photo someone spent hours composing and editing, or an image that took a GPU three seconds to generate.
The production cost difference is staggering. A professional stock photographer invests in equipment, travel, models, locations, editing time. An AI image costs fractions of a cent in compute. But the buyer pays the same.
Meanwhile, contributor earnings are dropping, with 62% reporting stagnant income and a 14% decline in average earnings per image. The platforms are replacing their most expensive input (photographer labor) with their cheapest one (compute), while keeping buyer prices stable. That's not an efficiency gain passed to customers. It's pure margin extraction.
Why pay $29/month for a subscription that includes AI images when you can generate them yourself? Or use a curated AI library like OKslop for nothing?
The stock photo middle ground is evaporating
Here's the positive framing that nobody in the stock photo industry wants to hear: stock photography was always a temporary commodity. It was a practical solution to a real problem: most people can't afford a photographer for every image they need. Stock filled the gap between "no image" and "custom photography."
But stock photos never fully delivered on what images are actually for: human expression and connection. Nobody looks at a stock photo of businesspeople high-fiving and feels something. Stock was functional. It was wallpaper. It served a purpose, but let's not romanticize it.
AI-generated images are the next evolution of that same commodity. Faster, cheaper, more customizable. If what you needed was wallpaper, AI wallpaper is strictly better than stock wallpaper. It's cheaper, you can describe exactly what you want, and nobody pretends it's art.
And nobody should be locking up the visual understanding that millions of photographers collectively contributed to. AI models learned to see from the world's images. Stock companies treating that collective visual knowledge as proprietary, while cutting the photographers out of the economics, is the real story here.
What actually makes sense now
For commodity images, use AI directly. Blog headers, presentation backgrounds, social media visuals, placeholders: just use AI. Generate it yourself, or use a curated library like OKslop where the licensing is simple, the images are free, and the AI-generated nature is disclosed upfront (it's in our name). Don't pay stock photo middleman prices for images that cost pennies to generate.
For images that matter, hire a real artist. If you need a photo that communicates something specific and human (a brand story, a documentary moment, an editorial illustration), work with a photographer or illustrator directly. Pay them fairly. Get something that can't be generated.
The gap between "commodity visual" and "meaningful image" is wider than ever. Stock photography sits awkwardly in between: too expensive for what AI can now do for free, too generic for what only a human can make. The sites quietly filling their libraries with AI while charging photography prices are accelerating their own irrelevance.
The future is simpler: slop for the stuff that doesn't need to be precious, and real artists for the stuff that does.
For more context on the stock photo landscape, see The State of Stock Photography in 2026. For how AI image licensing actually works, see AI Image Licensing Explained.
