Is AI stock footage ethical?

Mostly yes for illustration, and no when it pretends to be documentation. The line is not how the footage was made, it is what it claims to be. Below is where we draw that line, written by a library that generates its own clips, including the parts we have not solved.

We have an obvious interest here, so treat this the way you would treat any vendor writing about its own category. The useful test is whether the rules below are specific enough to be broken, and whether we admit anything that costs us something.

The question that actually matters: illustration or evidence

Nobody objects to a matte painting in a film, a stock photo of a generic office, or an illustrated diagram. Those are understood as representations. The objection to synthetic media arrives when it occupies the space where evidence is expected: a news frame, a product demonstration, a place you are being invited to believe in.

B-roll used as atmosphere sits firmly on the illustration side. Waves under a voiceover about slowing down are not a claim that a particular ocean was filmed on a particular Tuesday. That is why we think this category is defensible while other synthetic media categories are not.

What we refuse to make

These are binding rules in our production config, not aspirations. Each one costs us search traffic we could otherwise chase.

  • No real places. We never name or imply a real location. No Santorini sunset, no recognisable landmark lookalikes. Our scenes are types of place, not places. This costs us every location keyword in the category.
  • No wildlife, ever. Fake animals distort what people believe real species look like, and unlike a generic beach there is a factual world being misrepresented.
  • Nothing news-adjacent. No disasters, no crowds, nothing that could be lifted into a context where it reads as documentation of an event.
  • No faces, no identifiable people. To be precise rather than flattering: the gym set does contain hands and feet, framed on the equipment. What the library has nowhere is a face or anyone recognisable, because synthetic likeness raises questions we would rather not answer badly.

Where we have not solved it

Three honest gaps, listed because a page like this is worthless without them.

  • Disclosure does not travel. Every page here says the clips are AI-generated, but once a file is downloaded the label stays behind. Embedding provenance in the file itself, so it survives download and re-upload, is the right fix. We investigated it and did not ship it: signing with an untrusted certificate produces a claim that verification tools flag as coming from an unknown source, which is arguably worse than no claim at all. It stays on our list.
  • Training data. We use third-party models and cannot audit what they were trained on. That is a real limitation of our position, not something our own rules can fix.
  • Displaced work. Cheap atmosphere footage does take some jobs from people who used to shoot it. We would rather pay videographers for base footage and extend it with their consent than pretend the question does not exist.

How to use AI b-roll responsibly

  • Do not caption it as a real place. The moment a generated coastline is labelled with a real place name, you have made a claim the footage cannot support.
  • Keep it out of factual claims. Atmosphere behind narration is fine. Footage standing in as evidence of something that happened is not.
  • Disclose in contexts where provenance matters. Journalism, documentary and anything that could be read as a record deserve a line in the credits. Most marketing does not need one.
  • Check your licence. Ours allows commercial use and forbids resale of the raw clips. Read the terms.
Is AI-generated stock footage ethical to use in client work?+

For most commercial uses, yes, provided two things are true: the footage is disclosed as AI-generated by whoever supplies it, and it is not passing itself off as documentary evidence of a real place, event or person. The ethical weight sits on what the footage claims to be, not on how it was made. A generated beach used as atmosphere behind a voiceover is illustration. The same clip captioned as a specific coastline is a false claim.

Does AI stock footage take work from videographers?+

In the generic-atmosphere category, it competes with existing stock libraries more than with working shooters, because that market was already served by enormous catalogues of cheap or free clips. Where it genuinely displaces people is low-budget commissioned work: the small brand that would once have paid someone a day rate for filler shots. We think that is a real cost and not one to wave away, and it is part of why we are interested in paying videographers for base footage rather than only generating.

What about the training data behind the models?+

This is the least resolved question in the field and we will not pretend otherwise. The models we use were trained on large datasets whose contents are not fully published, and the legal position varies by jurisdiction and is still moving. We cannot audit that from the outside. What we control is our own output: no real locations, no wildlife, no faces, no imitation of a named creator's style.

How do you disclose that clips are AI-generated?+

Every collection page and every scene page says so in plain text, and the licence page repeats it. The honest gap: that disclosure lives on our site, so once an editor downloads a clip it travels without the label. Embedding provenance in the file itself is the fix, and we have investigated it rather than shipped it.

Do you use AI footage of real places?+

No, and this is a rule rather than a preference. We never name or imply a real location, so nothing in the library is a synthetic version of somewhere you could visit. We also make no wildlife, because fake animals distort what people believe real species look like, and nothing that could pass as documentation of a news event.

See what the rules produce

66 scenes across 8 collections, every one AI-generated, hand-reviewed and free to download. No real places, no wildlife, no faces. Browse the library, compare it against other free b-roll sites, or read what b-roll is.