Skip to main content
Copyright

AI Assisted or AI Modified? Choosing Between the Two Content Labels

In short

Two AI content labels, AI ASSISTED and AI MODIFIED, are easy to mix up. Here is the plain difference, with worked examples, so you can pick the right one for a post, photo, audio or video.

AI Assisted or AI Modified? Choosing Between the Two Content Labels — Swiss Trust Layer

Lena runs social media for a small shop in Bern. She made two things this week. For the first, she wrote short notes and let an AI tool draft most of the text. For the second, she took a product photo the shop already owned and used an AI tool to clean up the background. Both used AI. Both now need a label under the ProLitteris Content Label Terms. But they do not get the same label. The right choice starts with one question. Is this a photo, audio or video that was changed, or is it something else?

Two labels, two different jobs

AI ASSISTED and AI MODIFIED describe different ways AI was involved, so they are not interchangeable.

Both labels come from the Content Label Terms version 1.2, dated 21 September 2026, published by ProLitteris. A label says how AI helped make a piece of content. It does not say the content is good. It does not say who the author is. It only describes the AI role, and that is the whole point of picking the correct one.

What AI ASSISTED means

AI ASSISTED is for hybrid content where the AI part is a defining contribution.

The terms define AI ASSISTED as hybrid content with a defining AI contribution. Hybrid means part human, part AI. "Defining" means the AI part is central to the result, not a tiny touch at the edges. A short example: you write a rough outline and an AI tool writes most of the article from it. Another: you lay out a poster and an AI tool creates the main picture on it. In each case a person and an AI both shaped the work, and the AI share is large enough to matter.

What AI MODIFIED means

AI MODIFIED is for a photograph, audio recording or video recording that is built on existing content.

The terms call AI MODIFIED a variant of AI ASSISTED. It applies to content based on existing content, and only for photographs, audio recordings and video recordings. So you start from a file that already exists, and an AI tool changes it. Examples: an existing photo whose background you extend, or an existing recording whose background noise you remove with AI. Section 4.3 of the terms says AI MODIFIED applies where its requirements are met. If those requirements are not met, the label does not apply.

The written-text trap

AI MODIFIED never covers written text, even when AI changed an existing text.

This is where people pick the wrong label. The terms state that AI-modified written texts are not covered by the AI MODIFIED label. So if you take an old article and have an AI tool rewrite it, you do not use AI MODIFIED. That label is only for photographs, audio recordings and video recordings. For written text, look at AI ASSISTED instead, and use it only if the content is hybrid with a defining AI contribution.

One label per content, per version

Each piece of content, in each version, carries a single label.

The terms say one label applies per content and per version. You do not put both labels on the same file. If you make a new version later, you label that new version on its own. So the task is simple to state. Look at the version in front of you and choose the one label that fits it.

Pick the label: worked examples

These everyday cases show which label fits and why.

  • An existing product photo, background cleaned up by AI: a photograph built on existing content, so AI MODIFIED, if its requirements are met.
  • A post drafted mostly by an AI tool from your notes: not a photo, audio or video, so not AI MODIFIED; it can be AI ASSISTED if the AI part is defining.
  • An old article rewritten by AI: a written text, so AI MODIFIED does not apply; consider AI ASSISTED.
  • A video clip where AI changed existing footage: a video recording built on existing content, so AI MODIFIED, if its requirements are met.
  • A recorded talk cleaned and re-balanced by AI from the original audio: an audio recording built on existing content, so AI MODIFIED, if its requirements are met.

A label is not a claim about who wrote it

A label describes how AI was involved, not who the author is.

Be clear about what a label does. It tells readers how AI took part in making the content. It does not prove who created the work, and it does not settle a copyright question on its own. Keep the label for its real job, which is describing the AI role. Treat questions of authorship and rights as separate matters.

Where sealing fits in

Sealing records which labelled version existed at a given time.

A label and a seal answer different questions. The label says how AI helped. Sealing gives the file a trusted RFC 3161 timestamp and a cryptographic fingerprint, which is a short code that stands for the file, plus a public link anyone can open without an account. Only that fingerprint is sent to the timestamp authority. So if you label a version AI ASSISTED and seal it, you hold dated proof of which labelled version existed, and when. A seal does not show who wrote the work, and it does not change which label is correct.

Before you publish, settle the label with one question. Is it a photo, audio or video built on existing content? If yes, AI MODIFIED may fit. If it is a written text, or the AI part is a defining contribution to a hybrid work, look at AI ASSISTED. Pick one label for the version in front of you, then seal that version so its date is fixed. Start at swisstrustlayer.com.

Protect your work with Swiss Trust Layer AG

Seal your intellectual property with a court-proof e-Seal backed by Swisscom Trust Services.

Book a Free Demo