Which stock agencies still accept AI submissions
Shutterstock and Getty refuse contributor AI outright. Adobe Stock accepts it with a declaration — here is the one viable channel and what it actually pays.
The stock-submission question has a shorter answer than most people expect. Of the three agencies that matter for a contributor, two say no and one says yes with a checkbox. That is the entire landscape, and knowing it saves a month of preparing a portfolio for a door that is closed.
The three positions
Shutterstock: no. Their contributor policy is unambiguous. The content policy update on AI-generated content states that "Shutterstock will not allow AI-generated content to be submitted by contributors for licensing on our platform." This is not a disclosure requirement with a gate attached. It is a refusal at the submission stage.
Getty: no. Getty also refuses AI submissions, and its position extends further than people assume — AI-generated material used as a background or a prop inside an otherwise conventional shot is caught too. That is a meaningful trap for photographers who have started using generative fill or AI-extended backgrounds as a production convenience rather than as the subject of the image.
Adobe Stock: yes, with a declaration. Adobe accepts generative content provided you tick "Created using generative AI tools" at submission and hold the rights to what you are submitting. The Adobe Stock generative AI FAQ sets out the requirement. It is a real, open channel — not a grudging exception.
Net position: Adobe Stock is the one viable major channel for AI stock as a contributor. Everything else in this post follows from that.
What Adobe actually pays
Adobe's royalty details put the contributor rate at 33% on non-video assets and 35% on video, flat for all contributors. There is no tier ladder to climb — a first-week contributor and a ten-year one are on the same rate.
That flatness is the important structural fact, and it changes the strategy in two ways.
The first is that there is no reward for tenure, so there is no penalty for starting now and no advantage to a competitor who started in 2019. The second is less comfortable: the percentage applies to the net price of whatever licence was actually used, and most downloads arrive through a subscription rather than a full-price single sale. Adobe's worked example on that page takes a plan covering ten monthly photo licences and divides it out to a royalty just under a dollar per licensed photo. That figure falls as the buyer's plan gets larger, because the same 33% is applied to a smaller net price per asset. A stock portfolio on Adobe is a volume business with a long tail, not a per-image business. Nobody should build a monthly revenue plan around a hundred assets.
| Shutterstock | Getty | Adobe Stock | |
|---|---|---|---|
| Contributor AI accepted | No | No | Yes |
| AI as background or prop | No | No | Yes, still declare |
| Declaration required | n/a | n/a | Yes, at submission |
| Contributor royalty | n/a | n/a | 33% non-video, 35% video |
| Rate varies by tenure | n/a | n/a | No, flat |
Stripping metadata is a ban risk, not a workaround
The reasoning goes: Shutterstock detects AI from the file, so remove what it detects from the file. It is wrong on the mechanics and wrong on the consequences.
No agency publishes its intake detection stack, so treat any confident description of one — including this paragraph — as inference. What is documented is the label layer. C2PA content credentials, the IPTC DigitalSourceType field, and the XMP and EXIF tags that generation tools write are open standards, and Shutterstock is among the companies that adopted DigitalSourceType for signalling generated content. All of those live in metadata and all of them can be stripped.
What cannot be stripped is the picture. Intake review at every agency includes a human looking at the image, and the tells that get generative work rejected — hands, signage, reflections, repeated faces — are visible without reading a single field. So the strip-and-submit approach fails on its own terms. But the bigger problem is what you have done by trying. Removing a disclosure signal in order to pass a policy check is not an ambiguous act. It converts a rejected submission, which costs you nothing but time, into a misrepresentation, which is a contributor-account matter. A rejection is recoverable. An account closure takes the whole portfolio with it, including anything you shot on a camera.
There is also a straightforward reason not to fight the metadata even where it is allowed: on the agency that does accept AI, the declaration is a checkbox, not a disadvantage. You are not clearing a hurdle by hiding provenance, you are giving up the one channel that works. The mechanics of these labels, and where they travel between platforms, are covered in AI content detection and labels and in the synthetic media disclosure entry.
What actually sells on the one open channel
Adobe being open does not make Adobe easy. A flat royalty on a subscription-heavy marketplace rewards a specific kind of asset, and it is not the one most people generate first.
Generate for a search query, not for a portfolio. Stock buyers arrive with a job. "Four people in a small meeting room, one standing at a whiteboard, natural light, copy space on the right" is a search. "Cinematic portrait, dramatic lighting, 8k" is not a search, it is a prompt, and there are already ten thousand of them.
Copy space is a feature. A designer needs somewhere to put type. An image that fills every corner with detail is harder to place than one with a clean third.
Cover the variants. The same setup at 16:9, 4:5 and 1:1, and with the subject on the left and on the right, multiplies the queries one build can answer. This is where batch generation earns its keep rather than one-off hero images, and running a whole set from a single brief is exactly what the text-to-image tool is for.
Fix the tells before submitting. Hands, text on signage, reflections, and repeated faces in a crowd are the recurring rejection causes and none of them require a regeneration — a targeted edit is faster and cheaper than another full pass. Keyword and description quality matter more here than on any social platform, because the entire discovery mechanism is a text search; alt text, filenames and license markup applies directly.
Is this the right channel at all?
Worth asking honestly, because the answer for many people is no. Adobe Stock is a legitimate open door and a low per-asset business. Three alternatives use the same production capability and pay differently.
Direct licensing to businesses. A local agency that needs forty on-brand images for a campaign is not shopping a stock library. The output is comparable, the rate is not, and there is no royalty split.
Selling the method rather than the assets. Prompt packs, workflow bundles and templates are sold to other builders and are not subject to any agency's submission policy. For anyone already generating at volume, this is often the better use of the same hours.
Replacing a stock budget rather than supplying one. The buyer side of this shift has moved faster than the contributor side — replacing stock footage budgets with AI and the death of stock photography both cover where that demand went. For contributors specifically, stock footage creators and photographers set out how the production side is changing.
None of that means skip Adobe. It means treat it as one channel with a known rate rather than as the destination, and size the time you put into it accordingly.
FAQ
Can I submit AI work to Shutterstock or Getty if I edit it heavily afterwards?
Both policies are written to catch AI-originated material regardless of subsequent editing, and Getty's extends to AI used as a background or prop within a larger composition. There is no documented threshold of manual editing that converts an AI-generated asset into an acceptable submission at either agency. If the origin is generative, treat both as closed and do not try to argue the percentage.
Does declaring AI on Adobe Stock hurt how the asset performs?
Adobe's requirement is that you declare it and hold the rights. Adobe does not publish per-asset performance data broken out by declaration status, so anyone telling you declared assets rank lower is guessing. What is documented is the requirement itself and the flat royalty — both apply whether the asset is declared or not.
What does "hold the rights" mean if a model generated the image?
It means you must have the right to license what you are submitting, which depends on the terms of the generator you used and on what went into the image. Two things reliably break it: a recognisable person without a release, and a recognisable brand, logo or trademarked product in frame. A generated image can contain either without you intending it, so review before submitting rather than after.
Is it worth submitting video rather than stills?
The rate is slightly higher at 35%, and the supply is thinner than it is for stills, which is the more interesting half. The constraint is that stock video buyers want usable clips — steady framing, clean loops, no drifting artifacts across the shot — and temporal consistency is where generated video is most likely to fail an intake review. Test a small batch through the whole intake process before committing production time to it.