Brand compliance checker: catch risky claims before you publish

Most marketing copy gets into trouble the same handful of ways: a superlative nobody can substantiate, a guarantee that reads as a contract, a health benefit a food product isn’t allowed to claim, a “was £99” that was never £99. This checker reads your copy against 15 claim rules, flags each one with a severity, and tells you what would make it defensible.

New to the term? What is brand compliance? covers the five areas and who owns each, and marketing claims compliance goes into what substantiation regulators expect, market by market.

What it checks — the 15 rules

Rules are data, not code, and carry a version (currently v1.2). Severities below are the defaults; the jurisdiction resolver can only raise them, never lower them.

Code	Category	Level	Catches
`GTE`	Guarantee / promise	Critical	Language that can create a contractual or warranty obligation.
`HLTH`	Health / efficacy (regulated)	Critical	Any health, medical or efficacy claim.
`FIN`	Financial promotion	Critical	Investment, returns or credit claims.
`SUB`	Superlative / substantiation	High	A superlative or best-in-category claim stated as fact.
`CMP`	Comparative claim	High	Any comparison to competitors or to a prior state.
`PRF`	Performance / quantified result	High	Any quantified outcome, speed, saving or timeframe.
`ABS`	Absolute language	High	Absolutes that a single exception makes false.
`IMPL`	Misleading by overall impression	High	Copy that is literally true but leaves a misleading net impression.
`PRC`	Price / offer transparency	Review	Headline prices and offers that can mislead by omission.
`GRN`	Environmental / green	Review	Sustainability and eco claims.
`ING`	Ingredient / composition	Review	Free-from, organic, vegan, hypoallergenic and allergen-adjacent claims.
`DSC`	Disclosure trigger	Review	Mechanics needing visible terms or auto-renewal disclosure.
`TST`	Testimonial / endorsement	Review	Social proof, reviews, ratings, influencer language.
`URG`	Scarcity & urgency	Review	Urgency or scarcity pressure that may be false.
`IP`	Third-party IP / trademarks	Review	Competitor names and marks, third-party imagery, quoted reviews, music, likeness.

Two of these — IP and IMPL — carry no keyword pattern at all. They are judgement rules, which is exactly why a find-and-replace list of banned words cannot do this job.

How a flag becomes a verdict

Every flag lands on one of three levels, and the report’s verdict is decided by the worst one present — not by an average, and not by the model’s own summary.

Critical → hold. Don’t publish this as written.
High → needs review. Someone has to sign off or substantiate it.
Review → review terms. Usually a missing qualifier, not a missing basis.
Nothing fired → clear, against these rules and these markets.

The score is then clamped so it can never contradict the verdict — a report holding a critical flag cannot come back reassuringly green — and the same A–E band scale the brand voice audit uses is applied on top.

Footnotes and qualifiers never mitigate a flag here. If the headline creates the impression, small print underneath is not the cure — it can only make things worse, never better.

Markets matter

The same sentence is fine in one regime and a problem in another, so you tell the checker where the copy runs — Global, the EU, the US, the UK, Ireland, or any combination of individual countries. Jurisdiction is where the copy is seen, not where you are.

Where regimes disagree the strictest one wins: a run covering several markets is judged on the union, because copy that is clean in one market and not another is not clean. Global is the default for that reason — it flags anything that would fail anywhere. The ruleset currently carries authored per-regime overrides for the EU, Ireland and the UK; other markets are checked against the base rules.

Two registers you control

Out of the box the checker is conservative, which is right for a first run and wrong as a permanent state — a claim you have already substantiated shouldn’t be re-flagged every quarter, and a phrase your legal team has banned should be caught even though no regulator lists it.

Approved claims — things you have cleared

Record the claim, the market, the basis and an expiry (substantiation goes stale, so expiry is mandatory).
A match annotates and downgrades the flag. It never removes it — the reviewer still sees it, with your reference attached.
Critical claims stay critical. An approval cannot clear a guarantee, a health claim or a financial promotion.
An approval narrower than the run is held back and counted, so a UK-only clearance visibly doesn’t apply to a global check rather than silently failing.

Never-say — things you have banned

Your own prohibited phrases, matched literally against the submitted copy — no model call, no credit.
Exact, phrase (tolerating inserted words, which is how word bans get evaded) and broad matching with stored synonyms.
House flags move the score and verdict like any other, but are capped below hold — your house style is not a regulator.

Try the keyword pass free, no account

The free highlighter demo runs the ruleset’s own patterns over pasted text and marks up what they catch — like this: “The best mattress in Europe, with a lifetime guarantee and clinically proven results.” No sign-up, no credits, nothing stored.

It is deliberately half the engine: keywords only, no judgement rules, no jurisdiction resolver, no score, no registers. It says so on the page rather than letting a clean result read as a clean bill of health. There are also 21 worked examples across greenwashing, guarantees, price, comparison, ingredients, health, DEI, gambling, alcohol and lending — each with the bad version, the codes it trips and a rewrite.

Running it on real volume

One piece: paste text, give a URL, or upload a .docx — 1 credit.
A catalog: upload a CSV and check up to 100 rows per run, at 1 credit per 25 rows. See catalog audit for the Shopify / WooCommerce / Magento / Wix export path.
Where you write: the same check runs from Google Docs, Slack, the Chrome extension — and the WordPress plugin can block a publish, not just warn about it.
Sharing a result: save a report to its own link when you need to hand a finding to a client or a legal reviewer. Reports are not stored unless you press save.

Brand compliance checker FAQ

What is a brand compliance checker?

A tool that reads marketing copy and reports where it breaks a rule — either a rule the law sets (a claim that needs substantiating, a disclosure that has to be visible) or a rule you set (a phrase your brand has banned, a claim legal has already cleared). This one checks the claims and language half of brand compliance; visual identity needs a design review, and no text checker can substitute for one.

How is this different from running the copy past a chatbot?

The verdict is not the model’s. A model reads the copy and proposes flags; the severity floor, the score clamp and the final verdict are then computed in code from a versioned rule catalogue — so a critical finding cannot be talked down into a note, and two runs on the same copy give the same answer. Your own registers are matched literally against the text, with no model involved at all.

Can it replace legal review?

No, and it is built on the assumption that it will not. It is a review aid: it finds the lines worth a lawyer’s attention, states the legal question each one raises, and keeps a record of what has already been cleared so the same argument is not had twice. The decision stays with a qualified person.

Which markets does it cover?

The rule catalogue carries jurisdiction handling for the EU, Ireland and the UK, and merges to the strictest applicable regime rather than averaging across markets. Copy that runs in more than one place is therefore judged by the tightest one, which is the safe direction to fail in.

Is there a free version?

Yes — the keyword highlighter runs the ruleset’s own patterns on pasted text, free, with no account and nothing stored. It is half the engine and says so: the judgement rules carry no keywords, so a pattern pass is structurally blind to them. A full check is 1 credit for a page, text or a .docx.

Does approving a claim make the flag disappear?

No. A matching approval annotates the flag and downgrades it to review — the claim is still shown as having been made, with the note that it was cleared and by what. Critical and prohibited findings can never be approved away, approvals carry a mandatory expiry date, and an approval scoped to one market cannot soften a check that covers more.

Check my copy Try the free highlighter See a sample report

This is an automated drafting aid, not legal advice, and it is not a substitute for a qualified reviewer. Flagged lines are candidates for human review; a clear verdict means nothing in the ruleset fired for the markets you selected, not that copy is lawful. Rule codes and severities reflect ruleset v1.2; credit costs and row caps are current as of this page’s last update, with the pricing page authoritative.