This question comes up on almost every first call, and it deserves a straight answer rather than a marketing one: it depends, and the requirements are actively changing, so anyone who tells you with total confidence that there's one rule is skipping the part where they should say "check with a lawyer in your category."
What's actually settled
Platforms are increasingly labelling AI-generated and AI-edited content they detect, across Meta, TikTok and others. That labelling by itself isn't a penalty. Consumers seeing a small "AI info" tag under a video isn't the same as a brand being flagged for deception, and in practice most audiences care far more about whether the product looks and performs as claimed than whether the image behind it was generated.
What consistently does cause real trouble is misleading claims: a supplement image implying a result the product can't produce, a fabric shown behaving in a way the real material doesn't, a "before and after" that overstates what's achievable. That's a substantiation problem that exists independently of AI. It was true when brands retouched photos in Photoshop, and it's true now.
The categories where it actually matters
Regulated or high-scrutiny categories, skincare, supplements, health claims, financial products, carry a different bar. If your product makes a claim about efficacy, safety or results, that claim needs to be substantiated regardless of how the imagery was produced, and disclosure requirements in those categories are worth a direct conversation with your own legal or regulatory advisor, not a general answer from a creative studio.
For most DTC and ecommerce categories outside that zone, apparel, homeware, general consumer goods, the practical risk is reputational rather than regulatory: a brand caught hiding AI production after the fact reads worse than a brand that was upfront about it from the start.
The working position
Be honest about the process without leading with it. State plainly, if asked or where relevant, that imagery is AI-produced. Don't build a campaign around concealing it, and don't build one around announcing it as the headline either, since neither extreme is what the audience actually cares about. What they care about is whether the product does what the ad says it does.
That's the same standard good advertising has always been held to. AI production doesn't lower it and doesn't raise it. It just means the substantiation question arrives earlier in the process, at the brief stage, instead of after a shoot when reshooting is expensive.
If a specific claim in your category needs sign-off, get that sign-off before production starts. That's true of every claim in every ad, whatever produced the image behind it.