Which stock agencies accept AI-generated video?

AI video policies side by side

Articles on this topic often disagree. The table below gives each agency’s position as its own help page states it on 11 October 2026, with the date that page displays.

Does Shutterstock accept AI-generated videos?

No. Shutterstock’s policy page states: “Shutterstock will not allow AI-generated content to be submitted by contributors for licensing on our platform.” It gives two reasons. The first is ownership: contributors must prove they own what they submit, and “AI-generated content ownership cannot be assigned to an individual.” The second is pay: Shutterstock often cannot tell which model produced a piece of content, so it cannot confirm that every artist involved is compensated.

The rule rests on the contributor terms: Shutterstock cites sections 13d and 13f of its Terms of Service, which require proven IP ownership of everything submitted. In review, such clips are rejected as Non-Licensable Content. Shutterstock’s list of non-licensable subjects includes “Content that appears to have been created with AI-generative technology.” The wording targets content that looks AI-generated. The policy page names no exception for AI-assisted editing.

The same page points to Shutterstock’s own AI generator, which it says was trained on datasets licensed from Shutterstock. Contributors still cannot submit AI-generated content. The rest of the Shutterstock rules are on the Shutterstock video requirements page.

Adobe Stock: accepted with a label

Adobe’s generative AI FAQ is direct: “Adobe Stock accepts generative AI images, vectors, and videos that meet current submission standards.” The conditions for video come from three Adobe pages, all updated on 11 June 2026:

• Check “Created using generative AI tools” when you submit. Adobe describes it as “Required for all content created with generative AI software.”
• If the clip shows a person or property that looks real but is invented, also check “People and Property are fictional”.
• A clip that depicts or is based on a real, identifiable person needs a model release; real property needs a property release.
• The clip meets the same technical requirements as camera video: 5 to 60 seconds, 3,900 MB at most, a listed resolution and frame rate.
• “Don’t replicate other artists’ styles,” and do not use prompts tied to real-world intellectual property, or prompts that refer to people, places or property unless you have the legal right to.
• You need all the rights required for commercial licensing, so review the terms of the generative AI tool you used before submitting.
• Keep names of artists, real people, fictional characters and government agencies, references to copyrighted works and descriptions of real news events out of prompts, titles and keywords.
• “Avoid any embedded text, signature watermarks, or model inconsistencies.”
• Generative AI content cannot be submitted to the Illustrative Editorial collection.
• Submit at most three iterations of similar assets; more can be treated as spam.

Adobe’s refusal page lists four quality problems specific to generative AI:

• anatomical inaccuracies or malformed objects;
• surreal or distorted results;
• shadows, depth or lighting that do not fit the subject;
• render artifacts.

Adobe’s guidelines do not ask you to write “AI” in the title or keywords: the checkbox carries the label. They ask for the content type that matches how the asset looks, and for titles and keywords that describe it.

The label works in both directions. Adobe may refuse AI content submitted without it, and camera footage submitted with it. All other Adobe rules are on the Adobe Stock video requirements page.

Pond5: not accepted

Pond5’s help page, dated 5 November 2024, opens with: “At this time, Pond5 does not allow the submission of AI-generated content in our Marketplace.” Its reasoning matches Shutterstock’s. Pond5’s Contributor Agreement requires you to own everything you upload, authorship of AI content “cannot be attributed to an individual person,” and Pond5 cannot ensure that every artist involved in generating it is paid.

The page also describes Pond5 as a Shutterstock family brand: Shutterstock announced its acquisition of Pond5 on 11 May 2022.

Getty Images and iStock: not accepted

Getty’s public “Preparing your content” page, which covers both Getty Images and iStock, says: “Getty Images does not accept submissions created using AI generative models (e.g., Stable Diffusion, Dall-E 2, MidJourney, etc.).” The next sentence sets the limits of the ban: “This does not prevent the submission of 3d renders and does not impact the use of digital editing tools (e.g., Photoshop, Illustrator, etc.) with respect to modifying and creating imagery.”

The page carries no date. Getty’s detailed video rules sit behind the contributor login.

What counts as AI-generated

Only Adobe and Getty say in writing where editing ends and generation begins, and Adobe goes into the most detail. Its generative AI guidelines ask for the label when generative tools “changed, augmented, or added a new primary subject”: new people, animals or objects, or a recolored main subject. Extending the background, removing distracting objects or people and recoloring the background do not need the label. Adobe’s examples come from Photoshop and Illustrator, and the same checkbox applies to video.

3D work is a separate case from prompt-generated video. Getty names 3D renders as allowed, and Shutterstock’s VR and 3D article accepts 3D animations built from 3D environments or models, as long as they are meant to be viewed in 2D.

Upscaling is a separate risk, whatever the tool. Adobe states that “Upscaled videos (e.g., from HD to 4K) will be rejected,” and Shutterstock tells contributors not to modify a clip’s resolution or up-res it. An AI upscale from 1080p to 4K runs into those rules even at agencies whose AI policy never mentions it.

If you make AI video

• Among these four agencies, send AI-generated clips to Adobe Stock only. Keep them in a separate folder from your camera footage, so a batch upload never carries them to Shutterstock or Pond5.
• Check the generative AI box on every AI clip, and on no camera clip.
• Leave artist names, style references, real people, characters and brands out of prompts, titles and keywords.
• Get the same releases you would for camera footage whenever a real person or property is depicted.
• Watch each clip frame by frame for anatomical errors, malformed objects, mismatched lighting and render artifacts: Adobe names each of them as a refusal reason.
• Remove embedded text and watermarks, and submit no more than three variations of a similar clip.

For the other reasons agencies send clips back, see why stock footage gets rejected.