The same file that must be declared as AI-modified on one platform may need nothing at all on the next, and only counts as disclosed on the third if a human ticked a box in an ad manager months ago. This is the register that works out the answer platform by platform, records what you actually declared, and keeps the evidence behind it.
Buy Now — $19One-time payment · works offline · nothing leaves your browser
The platforms are not asking the same question, so one answer cannot serve all three. These are the three shapes of rule this register models.
Some platforms treat any AI-generated or AI-modified image, video or audio as disclosable, including real footage edited past a minor correction. Under this rule an AI cleanup pass on real footage is in scope, which is the case teams most often miss.
Others attach the duty to photorealistic or synthetic media showing people. A fully generated product shot with no human in it can sit outside this rule entirely — the same file that the broad rule captures.
Elsewhere the mandatory duty is scoped to a category, with a voluntary label available for everything else. Ticking that voluntary label is a decision worth recording, because it is the difference between choosing not to and forgetting.
One record per creative, filled in before it goes live.
How AI was used, how far the delivered asset was changed, whether a real or synthetic person appears, and whether it is a political ad. Every platform rule reads from these four facts, so you record them once and the answers follow.
Tick the platforms this creative runs on and each one gets its own outcome with the reason attached: required and missing, declaration missing behind an automatic label, labelled but thin on evidence, compliant, or genuinely not required.
Platform policy in this area changes every few weeks, so the rule table is yours to edit, with a checked date and a source against each entry. Anything you have not marked verified is flagged everywhere it is used, including in the printed pack.
State rules do not live inside platform policy. Tag the markets a creative is targeted at and provenance or synthetic-performer expectations are checked separately, so a creative can pass every platform toggle and still surface a gap.
Six creatives from one advertiser. The register bands them worst-first, so the ads that need a decision are always at the top.
| Creative | AI use | Per-platform outcome | Band |
|---|---|---|---|
| Spring Launch — 15s vertical cutGenerated presenter · live | AI-generated video | Broad-rule platform: label required, none applied Narrow-rule platform: label required, none applied | Label required |
| Range refresh — stock b-rollVendor chased twice, no answer | Not yet confirmed | No rule can be applied until somebody confirms how AI was used | Unconfirmed |
| Founder film — 30s cutdownCredentials read by the platform | AI-edited video | Broad-rule platform: auto-label only, declaration missing | Declaration missing |
| Ballot measure explainerPolitical · synthetic narrator | AI-generated voice | Category-rule platform: declared, evidence thin | Evidence thin |
| Spring Launch — static keyvisualTargeted at a provenance state | AI-generated image | Broad-rule platform: compliant Jurisdiction: no content credentials on the delivered file | Jurisdiction gap |
| Retail partner co-op adFull provenance on file | AI-generated image | Broad-rule platform: compliant Narrow-rule platform: no label required | Compliant |
Every guide to this ends the same way: keep a record of what you declared, per platform, before the ad runs. None of them gives you the record. Your options are an enterprise marketing-compliance suite sold to agencies, or a spreadsheet with one AI column that cannot tell three different rules apart. This is the thing in between, for $19, once.
Buy it, open it, start logging. No account, no subscription, no seat count.
No, and it says so on every screen and in the printed pack. It applies the rules you have recorded to the facts you have recorded, and shows you which creatives have a thin record. What that means for your campaign is a question for your own counsel.
That is exactly why the rule table is yours rather than baked in. Each platform ships as an editable entry with a checked date and a source line, and anything you have not marked verified is flagged in the register and in the printed pack, so nobody mistakes a shipped starting point for current research. Re-checking a policy is a two-minute edit, not a new version of the app.
Four things a column cannot do: it gives the same creative a different answer per platform, it separates an automatic platform label from your own declaration, it keeps “unconfirmed” as its own state rather than an empty cell, and it refuses to score a record that contradicts itself instead of quietly picking one side.
No. Those answer one jurisdiction’s question about an asset. This one is about divergence: the same creative running in three ad managers that disagree about what counts as AI content and who has to say so.
Nowhere. It is a single HTML file that stores everything in your own browser. You will be typing in client names, vendors and prompt locations, so that matters. Export a JSON backup or a CSV whenever you want a copy or need to hand the register to someone else.