Website Tracking, Consent, Schema and AI-Crawler Audit

Paste in a list of pages (usually home pages) and get, for each one, a ranked fix list covering the things that quietly break marketing reports and local search:

• Tracking: Google Tag Manager containers, GA4 / Google tag / Universal Analytics IDs, Google Ads, Meta, TikTok, LinkedIn, Pinterest and Bing pixels; the same container or property installed twice (inflated pageviews), legacy Universal Analytics left on.
• Consent: which consent manager is present (15 vendors recognised), whether a Google Consent Mode default is set, and whether it loads before the tags.
• Structured data: JSON-LD that does not parse, missing @context/@type, incomplete LocalBusiness and Organization data (address, phone, hours, logo, sameAs).
• AI crawlers: what robots.txt really lets GPTBot, ClaudeBot, PerplexityBot, Google-Extended and others fetch (evaluated the way RFC 9309 specifies), llms.txt, and noindex/nosnippet left over from a staging launch.
• SEO basics: title and description length, one <h1>, canonical, hreflang, lang, image alt text, Open Graph, and an optional spot-check of internal links.

Every finding carries its evidence (the matched snippet or URL) and a plain-language fix. Same input, same output: there is no AI in the audit, so results are repeatable and you can diff them month to month.

Who it is for

Agencies and in-house marketers who audit many small-business sites and want a first pass in minutes; consultants who need an evidence-backed list to send a client; anyone checking a site after a migration or a redesign.

What it cannot see (read this)

The audit reads the HTML the server returns, plus robots.txt and llms.txt. It does not run a browser. Tags that a tag manager, consent manager or website builder injects with JavaScript do not appear in that HTML. In our own tests on small-business sites, a first sample of 12 (used to fix detection bugs) found a tag or pixel on 8, and a fresh sample of 11 found one on 6 of the 9 sites the audit could read (2 sites blocked or timed out the request and are reported as such). Cross-checking the misses in a real browser, most had no analytics hosts at all, but one (a Wix site in the first sample and a Tealium-managed site in the second) loaded Google Analytics only at runtime. The report names the website builder or tag manager when it can recognise one. For that reason every "not found" result is marked verify and worded as "not found in the static HTML": a lead to check in a browser, not a verdict. A browser-based check is planned.

Nothing here is legal advice. Consent findings are technical observations; they do not say a site complies or fails to comply with any law.

Input

• Pages to audit: one URL per page, up to 100 per run. http and https only; private and internal addresses are refused.
• Internal links to spot-check: 0-25 same-site links per page, checked for errors (default 10; 0 turns it off).

Output

One dataset row per page: url, ok, counts of high/medium/low findings, topFix, the full fixList, every finding with evidence and confidence (confirmed or verify), and an inventory of what was detected. A page that blocks the audit (a 403 or error page) is returned as partial with only its robots.txt and llms.txt evaluated, and is not charged. A page that could not be reached at all is reported with an error and not charged.

Pricing

Pay per audited page (see the Pricing tab). You are never charged for a page that was blocked, unreachable or refused.

Good to know

The Actor identifies itself as SiteSignalAudit/0.1 and fetches one page, robots.txt and llms.txt per site (plus your chosen link sample), so it is gentle on the sites it audits.