Checkup methodology

One public URL. Observable evidence. Clear limits.

How Overhyped checks one public URL, calculates Search and Agent scores, handles optional review items, and keeps mobile Lighthouse evidence independently scoped.

The short answer

The checkup reports what its fetch can verify—not whether a website will rank or be cited.

One submitted public URL coordinates Search visibility, Agent readiness, and Website performance checks. Every result includes an observed value and status. Optional or contextual practices stay in Review and do not reduce the Search or Agent score.

Report inventory

Three evidence sections inside one report.

33

Search visibility checks

Page metadata, HTML, crawl and index signals, links, structured data, trust evidence, and other observable page signals.

16

Agent readiness checks

Public HTML, crawler-policy evidence, discovery resources, and optional agent-friendly conventions.

Up to 9

Mobile Lighthouse items

Four Lighthouse categories and five mobile lab metrics when Google PageSpeed Insights completes the Performance stage.

Performance is independent. A missing API key, quota limit, timeout, or upstream PageSpeed failure makes only that stage unavailable; Search and Agent evidence still returns.

What is collected

A bounded fetch, not a full-site crawl.

Submitted page

The final public HTML response, status, headers, metadata, visible initial text, headings, links, images, directives, and structured data for one URL.

Supporting public files

Bounded requests for robots.txt, a discovered or likely sitemap, llms.txt, llms-full.txt, and an optional Markdown response for the submitted page.

Provider-specific access

Root-policy evidence for selected search and agent crawlers. A named crawler group is optional; wildcard rules may govern access instead.

Mobile performance

Google PageSpeed Insights mobile Lighthouse categories and metrics, kept separate from exact-URL and origin-level Chrome UX Report field data.

Scoring contract

Only objective Search and Agent issues lower their scores.

Each Search and Agent stage starts at 100. Critical findings deduct 10 points and Needs work findings deduct 2 points. The score never falls below zero. Lighthouse uses Google's returned category and metric scores instead of this deduction model.

Critical10 points

An objective issue that can materially block or undermine the checked evidence.

Needs work2 points

An objective weakness that the single-page fetch can verify.

Review0 points

An optional, intent-dependent, contextual, or manually verified item.

Passed0 points

The checked public evidence met the defined condition at the recorded time.

A score of 100 is not proof of complete SEO or AI visibility.

It means the objective checks in that stage found no score-impacting failure at that time. Review items, private data, site-wide evidence, rankings, indexing, citations, and business outcomes remain outside that claim.

Optional review items

llms.txt, Markdown, and crawler declarations are context—not Google requirements.

The checkup can inspect llms.txt, llms-full.txt, Markdown delivery, structured discovery links, and named crawler policies because some publishers or providers may use them. Missing optional conventions stay in Review and do not reduce the score.

For Google's AI experiences in Search, conventional SEO remains foundational. Google says no new machine-readable files, AI text files, special markup, or additional technical requirements are needed for AI Overviews or AI Mode. Read the current Google Search Central guidance.

What it cannot prove

Eligibility and observable evidence are not outcomes.

  • It does not crawl every page or verify complete internal-link, duplicate-content, backlink, or broken-link coverage.
  • It does not read private Search Console, analytics, CMS, conversion, server-log, or business data.
  • It cannot prove current indexing, rankings, AI mentions, citations, recommendation, referral traffic, leads, or revenue.
  • Lighthouse is controlled mobile lab evidence. It is not a complete accessibility audit or a substitute for real-user field data.
  • Fetched page text is untrusted evidence. A recommendation still needs context and verification before implementation.

Read the consolidated Overhyped evidence limitations or inspect the representative Website Checkup report.

Canonical sources

Use one page for each important public claim.