Disclosure checks that people do not ignore

The failure mode for any automated disclosure check is not missing things. It is training the creator to click past it. A tool that flags sixty trivial items per post ends up ignored on the three that carry warning risk. A pre-publish check has to earn its place on every run, which means being deliberate about what it blocks on, what it merely warns about, and what it refuses to judge at all. The design constraint is attention — and on a one-screen caption, that ratio matters more than in a long email.

Block on what is certainly wrong

Some constructions are facts, not opinions. A post you told the checker is a paid partnership or affiliate commission that contains no disclosure of that connection anywhere in the text is a fact about the paste: §255.1 makes the expectation of the audience the test, and "you said this is paid but nothing here says so" is not a style preference. An affiliate shortlink or "link in bio" with zero commission wording is the same class: the machine can see the link-shaped CTA and can see the absence of any sentence telling the reader it pays you. Block here, before publish, because these are the findings that escalate into warning-level consequences. The only honest uncertainty is how a given reviewer will weigh the pattern — not whether the gap exists in the text.

Warn on what is probably wrong

Patterns the guides agree on, but that sensible creators sometimes break on purpose, belong in the middle tier. A relationship label like "ambassador" or "collab" is not automatically deceptive — whether it also needs an explicit connection sentence depends on context the machine cannot see, and §255.2 turns on whether the connection is reasonably expected by the audience. A disclosure that sits late in a long caption is the classic warning: the concern is conspicuousness, and often the honest fix is one plain-text line moved to the top, not a killed post. A "#ad" buried in a hashtag block warns for the same reason — it may be perfectly legal and still invisible. Emoji-only labels ("🤝 with @brand") deserve a warning slot for a rarer reason: they may be understood by the creator's core audience and miss it entirely for everyone else. Warn, quote the sentence, let a human decide — context the machine cannot see (audience familiarity, the platform label you set) often justifies what the pattern alone condemns.

Leave alone what cannot be decided mechanically

Whether the material connection is real — paid, gifted, commissioned, employed, family — lives in your contracts and affiliate dashboard, outside the pasted text, and an honest checker refuses to pretend otherwise. EndorsePrecheck never connects to an account, a brand deal, or any platform's paid-partnership toggle; it lists verification as a labeled advisory and stops there. Whether your platform's built-in label is switched on is invisible from a text paste — so the toggle appears as a reminder, never as a finding. Whether a non-US regime adds requirements beyond Part 255 — EU/UK UCPD and ASA positions appear only as context advisories, never as rulings. And a tool that pretends otherwise produces confident noise, worse than silence: it burns the credibility you need for the checks that genuinely know things. This is also why an honest pre-check refuses to output a "compliance probability." Nobody can compute it — enforcement is discretionary — and pretending otherwise is the check becoming the pattern it flags.

A workload that stays alive

Creators who keep a disclosure check alive run it on a fixed, boring cadence: one scan per draft, before the first publish, never after the posting calendar is committed. The scan is fast precisely because it is mechanical — locate the disclosure tokens, measure the first one against the platform's fold, match the published-factor vocabulary — and it takes seconds and produces the same ranks every time. What the cadence protects is the creator's attention: if the block tier is already cleared by the time a human reads the draft, the human spends their effort on the judgment calls the tool deliberately refused to make. That division — machine on patterns and placement, human on evidence and intent — is the arrangement that holds up in practice. The reverse (creators re-reading the guides at midnight, tools pronouncing on their contracts) produces exhausted marketers and untrustworthy scores.

What to measure instead of compliance theatre

A single pass/fail verdict is the enemy: it hides the one finding that matters behind fifty that do not, and invites bargaining ("I always put #ad somewhere"). Named, tiered findings survive disagreement — someone can argue with a rank, but nobody can argue with a quoted sentence and a named factor. If you must report a number, report the count of open HIGH findings, because that is the only figure with an obvious action attached: zero, or the post does not go live. Dashboards and "brand-safety scores" sound like governance and function as decoration; the fix checklist, closed line by line, is the actual control.

Tiering is the product

What creators actually want from a disclosure check is not more findings; it is a trustworthy ordering of them. Severity tiers — a missing or contradicted connection first, clarity and placement warnings second, tone noise last — are what keep the check on the desk instead of on the mute list. EndorsePrecheck separates these tiers rather than collapsing them into one verdict, because a single undifferentiated signal is what gets muted. And every tier stays reviewable: quoted evidence, named factor, suggested edit — nothing silently rewritten, nothing posted on your behalf.

Decision-support only: findings quote your copy and the published Endorsement Guides. You make the call before anything goes live. Heuristic pre-check — not an endorsement-disclosure certification or legal advice; no material-relationship verification, account connection, platform-policy measurement, or enforcement prediction performed.