Responsible AI disclosure

The “Generated by AI” mark.

Images and videos that FeedGraph creates with AI carry a small mark in the bottom-right corner. This page explains what it means, why it is there, and what it does and does not appear on.

Generated by AIFeedGraph
Light
Generated by AIFeedGraph
Mid-tone
Generated by AIFeedGraph
Dark
Placement

Bottom-right corner, inset from the edge, on every AI-generated image and video.

Treatment

Embossed and transparent — no panel, no brand colour, no fill of its own. It is sized to the asset, so it stays small on a large creative and readable on a small one.

Scope

What carries the mark

The mark means AI made these pixels. Applying it to anything else would make it meaningless, so original photography is never marked.

Marked
  • Product images generated by AI from your existing product photography.
  • Product videos generated by AI.
  • Images where AI created or replaced a meaningful part of the picture — a generated background, or an AI-extracted product cutout.
  • Every size and crop of those assets, at the point they are produced.
Not marked
  • Your own product photography, exactly as you uploaded or synced it.
  • Images edited only by ordinary, non-AI processing — resizing, cropping, format conversion, or the local background remover.
  • Templates and layouts composed from your own images without AI.
  • Anything else you upload to FeedGraph.

What the mark means

“Generated by AI” means exactly that: a model created these pixels, or materially changed them. It is a statement about how the asset was made, not a judgement about quality or accuracy.

“FeedGraph” is the attribution — it says which tool did it, so anyone looking at the asset knows where to ask.

Why we apply it

People are entitled to know when what they are looking at was made by a machine. Marking our own output is the simplest, most direct way to say so — it travels with the asset wherever it ends up, without depending on a caption or a platform setting.

It also protects the merchant. Advertising platforms and regulators increasingly expect synthetic media to be disclosed. An asset that already declares itself is one less thing to reconcile later.

Where it appears

Bottom-right, inset from the edge, at a size proportional to the asset. It sits in the same place on a square catalogue image, a 9:16 story, and a wide banner.

The treatment is embossed rather than printed: a light edge and a dark edge with nothing filled in between. That way it reads on a white studio backdrop and on a dark lifestyle shot without ever becoming a solid block sitting on top of the creative.

It is part of the file

The mark is composited into the pixels when the asset is produced, before it is saved. It is not an overlay or a layer that can be switched off in the interface, and it is not added at display time.

Alongside it, the file carries a machine-readable declaration in its own metadata — the IPTC digital source type that identifies synthetic media, which platforms and verification tools can read directly.

Some models, including Google's, also embed their own invisible watermark. We deliberately avoid re-encoding in ways that would strip it.

What cannot be changed

Whether the mark appears at all. Every image and video FeedGraph generates carries it, and there is no setting, plan or export path that removes it.

The mark is applied by the same code for every workspace, so two customers cannot end up with different levels of disclosure on the same kind of asset.

Assets made before a change

The mark is burned in at the moment of generation, so an asset keeps the wording and placement it was made with. Changing the settings affects what is generated next, not what has already been produced.

FeedGraph also declares AI-written product data at the field level, and can attach an AI declaration to feed exports. Those are separate from this mark, which covers images and video only.

Last updated 2026-08-24. See also AI governance in Security.