AI disclosure label generator
Amazon reads a metadata tag and renders its own disclosure. Google, Meta and TikTok do not — there the disclosure has to be visible on the creative itself. This tool draws that label into the image, positioned to survive each platform's cropping, and writes the contains-synthetic-performer XMP tag on the way out.
Drop images here
JPG · PNG — nothing is uploaded, the label is drawn in your browser
Placement
Conservative default — not an official figure. Google allows labels burned into creatives but only says to keep them out of the very corners and edges, with no published margin. 6% is a conservative default — also turn off Asset Enhancements and Adaptive Layouts, which can crop the label.
Preview
Why a visible label, and not just metadata
Since July 2026 Google has allowed advertisers to add text or visual labels directly inside image and video ad creatives without breaching its policies on text overlays and watermarks. For campaigns targeting the EU, India and New York, Google also renders a visible overlay on assets marked as AI created or edited. Google states plainly that using its label setting “doesn't guarantee compliance with specific regulations” — the platform toggle is a tool, not a safe harbour, which is the case for burning your own label in as well.
Turn off the enhancements that can crop your label. Google names Asset Enhancements (Display and Performance Max) and Adaptive Layouts (Demand Gen) as features that may trim content during rendering. Labels sitting near a corner or edge are the first thing to go.
About the margins in the presets
Google's published guidance is qualitative — keep labels out of the very corners and edges so they are not trimmed — and it does not state a numeric margin. The presets here are therefore conservative defaults, each labelled as such in the panel above, and every one of them is adjustable. Nothing in this tool invents an official-sounding number that platform documentation does not actually contain. Check the result on a real device before a large flight.
What the label should say
No platform has published required wording, so the constraint is comprehension rather than compliance: a shopper has to understand it at a glance. “Synthetic performer” is the statutory term and almost nobody outside the industry knows it.
| Wording | When it fits |
|---|---|
AI-generated person | Short enough for feed and story placements; states exactly what triggered the requirement |
This ad features an AI-generated person | Where there is room, and where a full sentence reads less like a watermark |
AI-generated content | Broader claim — fits TikTok's policy, which covers fully AI-generated imagery whether or not a person appears |
Created with AI | The softest form; use it where the AI element is scenery rather than a human figure |
Whichever you pick, keep it consistent across a campaign. A disclosure that changes wording between placements reads as decoration rather than a statement.
When not to burn a label in
- Amazon listing images. Amazon reads the metadata tag and renders its own disclosure, and main images have to keep a pure white background with the product filling most of the frame. A burned-in label works against both. Use the metadata route there.
- Masters you will re-crop later. Burn the label into each output size, not into the source — a label placed for 1:1 will sit in the wrong place, or off the canvas, after a 9:16 crop.
- Creative with no synthetic person in it. Over-disclosing is not free: labelled content measurably costs consumer trust, so applying it to images that never needed it spends that for nothing.
Will a burned-in label affect ad review?
Google states that labels added for AI disclosure will not be treated as violations of its policies on text overlays and watermarks. TikTok rejects ads for improper use of disclaimers, which is the opposite failure — declaring something the creative does not contain. Meta applies its own label independently of whatever you burn in.
The one mechanical risk is cropping, not review: automated enhancements can trim the edges of a creative, and a label near an edge is the first thing to go. Google names Asset Enhancements in Display and Performance Max, and Adaptive Layouts in Demand Gen. Turn those off for campaigns carrying a disclosure.
Platform-by-platform
| Platform | How disclosure works |
|---|---|
| Google Ads / YouTube | Set the AI label in Google Ads, DV360, CM360, Merchant Center or Ads Editor, and/or burn one in. Labels Google adds automatically cannot be overwritten. |
| Meta | Meta applies its own AI info label to content made with its tools and tries to detect third-party AI. Detection by Meta and your own disclosure duty are not the same thing. |
| TikTok | Toggle “This ad contains AI-generated content” in Ads Manager. For Spark Ads the original post needs its own disclosure. Note that duplicating a campaign resets the toggle, even when the ad is reused. |
| Amazon | Metadata, not pixels: write contains-synthetic-performer into XMP and Amazon renders the disclosure. A burned-in label is usually unnecessary and main images must keep a pure white background. |
This page summarises publicly available platform documentation and is not legal advice.