ProvenanceGuard
← Go back

← All guides

How to Label AI-Generated Images Before You Publish Them

By the Ingenious AI team · Published: November 14, 2026 · Reviewed: November 14, 2026 · 8 min read

The short answer

Label in the pixels, not only beside them. Publish the visibly labelled copy, whose custody strip names the content as AI-generated and registered, keep the caption wording consistent with the declaration you sealed, and register the fingerprint first so a stripped copy can still be traced back to the record.

Seal an image free → Recover a reposted image Pricing

A caption is a sentence next to an image, and the internet is very good at separating sentences from images; the label that survives is the one the audience cannot scroll past without the image itself.

There is also a tone question. A custody strip that looks like part of the work's presentation, clearly worded and consistently used, reads as professionalism rather than apology. Teams that label every AI asset the same way, from the first post rather than after a complaint, find the label becomes part of their brand's trust signal. The creators who struggle are usually the ones who labelled inconsistently, so that an unlabelled repost looks like concealment next to their labelled originals.

What is an AI content label? a disclosure attached to AI-generated content that identifies it as AI-generated; in its durable form it is rendered into the image's pixels, so it is seen wherever the image is seen and survives reposting.

The problem behind labelling AI images

A visible label does the communicating that metadata cannot. Most people who see an image will never open its properties, and most platforms will never show them a credentials panel. A label in the pixels, naming the content as AI-generated, is read by exactly the audience the disclosure rules exist for, on exactly the surface where they encounter it. ProvenanceGuard's labelled copy carries that strip as part of the image, so the disclosure travels with the pixels instead of depending on a caption that can be separated from them.

Labels have their own failure modes, and a serious workflow plans for them. A label placed along an edge can be cropped out; a label rendered too small becomes illegible after a platform resize; a screenshot of the centre of an image may exclude it entirely. That is why the visible label is one layer among several: the survival re-checks specifically test whether the label remains detectable after simulated platform transforms, and the registry fingerprint remains as the fallback when the label is gone.

How ProvenanceGuard handles it

What sealing does not do is just as clearly bounded. It does not certify that a disclosure statement is legally sufficient, it does not turn the assertion into a C2PA manifest, and it does not store the uploaded original: the registry keeps the sealed and labelled copies and the record, so verification and downloads keep working, while the original bytes are not retained. The seal is evidence of a declaration made at a time, about a specific file, and that precision is what makes it useful later.

Sealing is the moment a record is created. ProvenanceGuard takes the creator and disclosure the uploader declares, the SHA-256 of the original bytes, and the registration time, and signs that assertion with HMAC-SHA256 under the product signing key. The signed assertion is embedded into the file as XMP, producing the sealed copy. Because the signature covers the assertion's contents, altering the creator, the disclosure or the hash inside the file afterwards invalidates the signature, and the verification page shows that plainly.

What survives distribution, and what does not

Alongside the sealed copy, sealing produces a visibly labelled copy: the same image with a parchment custody strip in the pixels naming it as AI-generated content whose provenance is registered with ProvenanceGuard. The distinction matters. Embedded assertions are machine-readable and fragile in transport; a visible label is human-readable and survives anything that keeps the strip in frame, including a screenshot. Publishing the labelled copy where the audience actually sees the image, and keeping the sealed copy as the record, uses each layer for what it is good at.

The practical reading of these dates is about preparation rather than panic. Duties that have already applied cannot be met retroactively for assets published unmarked, but a workflow adopted now, applied consistently to everything published from here on, with records kept per asset, is a defensible position that improves with every week it runs. The grace deadline in December 2026 is specifically about marking for pre-existing systems, which makes conformance testing of the marking itself, not just the policy around it, the urgent piece of work.

A routine that survives busy weeks

The legal backdrop in the product's own evidence wording is specific. The EU AI Act's Article 50 transparency duties have applied since 2 August 2026, and the marking grace period for systems already on the market before that date ends 2 December 2026. California's AI Transparency Act, SB 942, is operative from 2 August 2026. New York's synthetic performer law, in GBL section 396-b and effective 9 June 2026, covers disclosure where commercial advertising features a synthetic performer. The evidence packs carry disclosure wording for exactly these three jurisdictions.

None of that makes a provenance record a legal verdict, and ProvenanceGuard's packs say so themselves: evidence aid, not legal advice. What a record does is give counsel, a compliance team or a regulator something concrete to work from: the file, its hashes, the tool that sealed it, the timestamps, the declared disclosure and the wording used. Organisations that can produce that pack in minutes answer questions very differently from organisations reconstructing a publication history from chat logs and memory.

Five actions to take

  1. Seal each image before the labelled copy is generated, never after publishing
  2. Use one label wording everywhere so an unlabelled copy is visibly the exception
  3. Re-check one published asset through the Recovery lookup so you know the stripped-copy path works before you need it
  4. Seal the final export of the asset, with the creator and disclosure declared in your own accurate words
  5. Download and keep both copies the seal produces: the sealed copy as the record, the labelled copy for publication

A worked example: A creator launching an AI-assisted print series

She sealed all 18 prints, published labelled copies, and when a marketplace repost dropped her caption entirely, the strip in the pixels still disclosed the work and the recovery lookup matched the repost to her registry record at 171 of 192 fingerprint bits agreeing.

The lesson generalises beyond this case. Records made at publication time are cheap, calm and complete; records reconstructed under pressure are expensive, partial and contested. Everything in this guide is in service of moving work from the second category into the first.

FAQ

Does a hashtag or caption disclosure count as labelling?

It discloses at the point of publication, which is worth doing, but it is separable from the image and usually gone by the second repost. Pixel-level labelling plus a registry record is the version that survives distribution.

Can anyone verify a record, or only the owner?

Anyone. Verification pages are public by design, the single-hash lookup is a public API, and an exhibit ID or a file hash is enough. A record only its owner could check would not settle a dispute with a stranger.

Is ProvenanceGuard's assertion a C2PA manifest?

No. It is ProvenanceGuard's own signed assertion, embedded as XMP and HMAC-signed, and labelled that way everywhere. The inspection detects genuine C2PA markers when a file carries them, but the two records are never conflated.

What formats and sizes are supported?

JPEG and PNG, up to 25 MB per file. Those are the formats the inspection, XMP embedding and fingerprinting are built and tested against; other formats are refused rather than approximated.

Seal an image free →