Are AI-generated images allowed in ads?

The platform rules, stated plainly

Meta permits AI-generated creative in general advertising, and requires advertisers to disclose AI-generated or digitally altered content in ads about social issues, elections and politics. It also applies its own AI labels to some content independently of what you declare.

Google Ads follows a similar shape: generated creative is allowed, and election advertising carries disclosure requirements for synthetic content. YouTube requires disclosure of realistic synthetic or altered content at upload and shows viewers a label.

So for a local business running ordinary product and service ads, generated creative is permitted. The obligations bite hardest around politics, elections and anything realistic enough to be mistaken for a recording of real events.

The rule that applies whatever the platform says

India's consumer protection rules on misleading advertising do not care how an image was produced. An image implying a place, a person, a result or a product that does not exist is a misleading claim, and a generated image makes exactly that kind of implication easier to produce by accident.

The test we teach is simple and it keeps you clear of nearly all of this: if a reasonable person would assume the image is a photograph of something real, either use a photograph or label it clearly.

Four images to never generate

These are the ones that cost real money, in refunds, removed ads or lost trust.

• Your premises — a generated classroom, showroom or clinic that does not match what a visitor walks into.
• Your team or customers — a synthetic face presented as a real student, staff member or client is a fabricated endorsement.
• Results and dashboards — a generated screenshot of numbers you did not achieve is a fabricated claim, not a mockup.
• Real people's likenesses — celebrities, politicians, competitors. Personality rights apply in India and platforms remove it.

Where generated images genuinely help

Backgrounds and abstract creative, where nothing is being claimed. Concept visuals for a campaign you are pitching. Variant testing where the subject is generic — six background treatments for the same offer, to find which one holds attention.

Also product photography clean-up, which is the most underrated use: swapping a cluttered background behind a real photograph of a real product keeps the claim honest while making the creative usable.

What to do before the ad goes live

Check the creative at the size it will actually appear. Most generated-image tells — hands, text, logos, background geometry — survive at full size and become obvious in a feed at 320 pixels wide, which is worse than the reverse.

Then check whether anything in the frame makes a claim. If it does, either it is true and photographable, or it does not run.