Compliance checks

Your website makes promises. It says you are carbon-neutral, that you never test on animals, that everything is handmade in Britain, that you never use the word "guaranteed". Then a new page ships, a product description is copied from a supplier, and quietly the site starts contradicting its own rules. Compliance checking reads your pages against a policy you write, and flags the contradictions.

What a compliance profile is

A compliance profile is your rulebook, written in plain English. It has two parts:

• A name, so you can tell your policies apart (for example "Sustainability claims" or "Regulated finance wording").
• The policy text: the rules themselves, typed in as ordinary prose. This is where you state your ethos, your sustainability commitments, the claims you must never make, and the tone you expect. There is no fixed template. Whatever you write is what the judge holds your pages to.

A profile is reusable: write it once and assign it to as many domains as you like. If you only need rules for a single site, you can instead type rules straight onto that domain rather than creating a shared profile.

What the results mean

A completed check ends on one of two verdicts:

• Clear. No contradictions were found on the pages checked.
• Issues. One or more pages contain text that contradicts your policy. Each issue is listed with the page it is on, the exact quoted sentence, the rule it conflicts with, and an explanation.

Each issue also carries a severity, which is the judge's read of how clear-cut the contradiction is:

• High. The page text directly contradicts an explicit statement in your policy.
• Medium. It likely contradicts the policy but takes some interpretation.
• Low. A tension worth a human look, but not a clear contradiction.

If a check cannot gather enough readable text to judge (no pages fetched, or very little content across them), it finishes as Not enough content rather than guessing, and the credits are refunded.

Dismissing a flag you disagree with

The judge is a strong first reader, not the final word. If a flag is a false positive, or a claim your legal team has cleared, you can dismiss it with an optional note. Dismissing is an overlay: it hides the flag from your rollups without rewriting the report, and it survives re-runs, so the same wording will not keep re-flagging. You can restore a dismissed flag at any time. A site whose only remaining flags are all dismissed rolls up as clear.

What you need before you start

• A monitored website to check. Add the site first if you have not (see Add your first website).
• Your rules, written down. A short, clear policy is worth more than a long vague one. You can paste in an existing ethos or claims document.
• Scan credits. Compliance draws on the same scan-credit pool as security scans and the other page audits. It costs one credit per ten pages checked, rounded up, so a run of up to ten pages is a single credit.

How a check runs

When you run a check, TLDTrack first works out which pages to read. It discovers your pages from your sitemap where you have one, or by a shallow crawl of your homepage otherwise, and you pick the pages you want covered (up to 50). It then fetches those pages, extracts their readable text, and sends it to the AI judge in batches to compare against your policy.

• A small run (up to eight pages) is judged instantly, in the same request, so you see the verdict straight away.
• A larger run is queued and processed by a background worker within a few minutes; the page updates itself when it is done.
• The exact set of pages is snapshotted with the check, so later edits to your page selection do not change what a past report covered.

The judge only ever flags a contradiction it can back with a verbatim quote from the page. If it proposes a flag whose quote does not actually appear on the page, that flag is dropped rather than shown, so you are not chasing phantoms.

How alerting works

When a check completes with a verdict of issues and at least one active, non-dismissed high-severity flag, TLDTrack emails the account owner and your notification contacts, and sends a push notification to your team. The alert names the site, and the report lists the offending pages and quotes. Lower-severity-only results are recorded for you to review but do not raise an alert, so the notification is reserved for genuine, clear-cut problems.

How to set it up, step by step

1. 1
   Open Compliance Checks from your dashboard and create a policy: give it a name and paste or type your rules in plain English.
2. 2
   Assign the policy to the domains it applies to (or add inline rules on a single domain instead).
3. 3
   Discover the site's pages, then choose which ones to cover (up to 50). High-value pages such as product, claims and about pages are the ones that matter most.
4. 4
   Run the check. A small run returns a verdict immediately; a larger one finishes in the background within a few minutes.
5. 5
   Review the flags. Each one shows the page, the quoted sentence, the rule it conflicts with and an explanation. Fix the copy, or dismiss a flag you disagree with, adding a note for your team.
6. 6
   Re-run after edits, or leave it and run again whenever your site or your policy changes.

Where to set this up

Good to know

• Compliance shares the scan-credit pool with security scans, accessibility and SEO audits. One credit covers up to ten pages.
• There is no fixed checklist. The judge holds your pages only to the policy you actually wrote.
• Every flag is grounded in a real quote from the page; unquotable flags are dropped.
• Runs are on-demand. Trigger one after a content change, a policy update, or before a campaign goes live.
• Dismissals persist across re-runs, so cleared wording will not keep flagging.