Useful checks. Clear limits.

Website SEO checker and AI readiness scan: what it checks

Aureviq’s free scan is a rules-based website SEO checker and AI readiness checker. It reads your public pages (up to 40 by default), applies 29 fixed rules in four areas and shows the evidence for each result. It is an SEO audit tool for the basics, not a way to see where you rank.

How missing information is handled

Each area is scored out of a fixed total (30, 25, 25 and 20). A check the scan could not confirm, or that does not apply to your business, is left out of both sides of that sum instead of counting as a pass or a zero. If an entire area cannot be measured, the overall score is shown as unavailable.

What it does not measure

The score is a readiness score, not an SEO score checker for rankings. It does not see Google’s index, search volume or traffic, page speed, backlinks or competitors, and it does not run JavaScript, so content added by scripts may be missed. It is not a security audit or a certification that a business is legitimate. A GEO audit in this sense means checking whether your information is clear and accessible; it cannot tell you whether an AI tool mentions you. That has to be observed: the question asked, the engine used, the date, the answer and the sources it cited. Aureviq’s AI-answer monitoring, part of Website Growth, records those observations.

The first pilot—not a certification

We checked eight non-customer websites across eight industries. A separate browser-based reference review was saved before running the production scanner code. Codex performed both stages using different methods; no outside auditor or second reviewer certified the results.

We evaluated 76 of 80 planned site/check pairs, covering just 10 of the scanner’s 29 checks. All eight crawls were partial. This pilot does not validate the overall score, the other 19 checks, business credibility or AI/Google ranking predictions.

How the score works—and what it cannot tell you

• Our current standard scan reads up to 40 public pages, to a depth of three links, within a 30-second crawl budget. A slow or larger site may have fewer pages checked. It reads downloaded HTML, not every interactive state, and does not discover unlinked pages through a sitemap.
• The 29 programmed checks feed four weighted areas: website structure (30 points), business information and proof (25), discoverability (25), and contact/policy basics (20). These weights are product choices, not validated predictions of rankings.
• Unconfirmed and inapplicable checks are excluded from each area’s denominator. An entirely unchecked area makes the overall score unavailable; unknown evidence is not a zero. Limited coverage can still produce a high score.
• A title-length or single-heading rule is an editorial convention. Being outside it is not evidence of a Google penalty. Google says its AI search features do not require special structured data. Read Google’s guidance.
• Presence is not truth: we do not authenticate testimonials, check business licensing, confirm a phone belongs to the business, test email delivery or perform a security audit. “Collection confidence” describes crawl coverage, not measured accuracy.
• Detection update 1.3.0 recognises business-type variants, quotes with attached names and structured work stories. Client quotes do not count as staff biographies. Repeated quotes count once; affiliated feedback is not labelled independent. Unusual layouts, image-only text and JavaScript-only content can still be missed.
• Collection safety update 1.3.1 excludes known shopping action queries and applies wildcard robots restrictions to paths and queries. Sitemap and social-link verification validate each redirect destination; blocked or unresolved verification remains not checked. Scoring weights are unchanged, but collected pages, applicable checks and coverage confidence can change. Compare websites using the same collector version and settings.
• Evidence update 1.3.3 rejects empty, pending and browser-challenge responses, validates sitemap content, and excludes ambiguous numeric/date phrases from address evidence. Report cards show assessed and unconfirmed point totals alongside the scaled scores. This is a detection correction, not independent validation of the entire score.
• The pilot below used the earlier scoring model and eight-page limit. Its accuracy numbers are historical, not validation of version 1.3.3. Scores from different versions or different page samples should not be presented as business improvement.
• The score does not measure actual Google traffic or AI recommendations. Connected Search Console results and separately recorded AI-answer observations are different measurements.