Skip to content
WebAuditor360 AccountFree audit

Public methodology

Our audit engine: 40+ data sources and 87 checks, in a single report

This is how the engine that audits your website works inside. This page explains the five pillars and what each one checks, the source of every data point, how it is weighted, how we keep measurement noise from passing for a real change, and what we do when something cannot be measured. It is written so that any result in the report can be reproduced and challenged.

An audit, step by step

  1. URL validation. We resolve DNS and check that the IP is public: we never audit internal or private addresses. We follow redirects and, if port 443 accepts connections but does not answer while HTTP does, we audit over HTTP and flag it as an HTTPS failure.
  2. Parallel measurements. Our crawler (home page and up to 30 key pages), the real browser, the TLS probe and the queries to 40+ data sources all start at once. The browser is queued: when several reports run at once, each waits its turn and its time limit grows by however long it waited, so none ends up without cookies, accessibility or screenshots.
  3. Normalisation. Each source is converted into a common format with its status: measured, no data or error. A third-party error is never turned into a failure of your site.
  4. Rules. Fixed rules, identical for every website, turn the data into prioritised issues (critical, high, low or tip) and into the 87 checks across the five pillars.
  5. Plan. Issues are ranked by impact versus effort and, when we detect the platform (WordPress, Shopify, Wix, PrestaShop…), the steps say where to click in it.

The five pillars and what each one checks

Checks are numbered by pillar. These are the ones the engine measures in every report, and where the data comes from.

ATechnical checks 1-20

That Google can reach, read and index every page, and that it loads fast.

What is checked: HTTP status of the home page and the sample, robots.txt, noindex in the meta tag and the X-Robots-Tag header, canonical, XML sitemap (compressed ones too), broken internal and external links, redirect chains, orphan pages, URL structure, mobile friendliness, HTTPS, Core Web Vitals, mobile speed, JavaScript-dependent content, image weight, alt text, hreflang and structured data.

Where it comes from: Our own visit and a crawl of up to 30 pages, PageSpeed Insights (Lighthouse), the Chrome UX Report and the real browser.

BContent & authority checks 21-40

That each page clearly says what it is about, who is behind it and why to trust it.

What is checked: Titles and meta descriptions (length and duplicates), one H1 per page and heading hierarchy, sources linked from the text, corporate transparency (legal notice, tax ID, address), brand identity, in-text internal links, pages competing for the same thing, thin content, update dates and external popularity signals.

Where it comes from: Our own crawl, Common Crawl, Tranco, the Wayback Machine, Wikidata and VIES.

CAI visibility (GEO) checks 41-60

That AI search engines can get in, read your content and know who you are.

What is checked: robots.txt rules for OAI-SearchBot, GPTBot, Bingbot and Googlebot; CDN or firewall blocks that robots.txt does not show; crawlable text without JavaScript; main content marked up with <main> or <article>; company and brand data; authorship and dates; links to evidence and original data.

Where it comes from: The open ai.robots.txt list, visits under each bot’s name, unrendered HTML and Wikidata.

DAnswer engines (AEO) checks 61-80

That a specific answer can be lifted from your page without effort.

What is checked: Question-style headings, direct answers after each question, definitions and glossaries, lists, tables, FAQs, Schema.org types and valid JSON-LD, article and organisation markup, BreadcrumbList, sections with subheadings, dates, prices and units, and readability measured with INFLESZ and Fernández-Huerta. We also detect misleading markup, such as star ratings a company gives itself.

Where it comes from: Our own crawl and analysis of the JSON-LD on every page in the sample.

EExperience & conversion checks 81-100

That visitors can understand, trust and get in touch, without barriers or legal risk.

What is checked: Value proposition and calls to action, visible contact details, pricing, mobile navigation, forms and their validation (never submitted), privacy, cookie consent, legal information, security (headers, TLS, malware lists), contrast, keyboard use, form accessibility, mobile text size and conversion tracking.

Where it comes from: Real Chromium with axe-core, Mozilla HTTP Observatory, Google Safe Browsing, our TLS probe and the cookie scanner.

See my website’s 5 pillars

How the score is weighted

Each measured check is worth 100 if it passes, 50 if it needs work and 0 if it fails. Each pillar score is the average of its measured checks, and the Web Health Score is the average of all measured checks: they all weigh the same, so no pillar can hide another. Anything not measured neither adds nor subtracts.

A check that depends on a detected issue takes the worst state: a critical or high issue makes it fail, a low one makes it “needs work”, and a tip takes nothing away. The report evaluates 87 checks with fixed rules: 68 with objective data and 19 with rule-based estimates from the HTML and the crawl, which the report labels as such.

Levels: Excellent (90-100), Good (75-89), Needs work (50-74) and Poor (0-49). The verdict is separate: Critical if there is any critical issue; Needs attention with three or more high-priority issues or a score below 70; and Healthy otherwise. So a site with a good score but with Google blocked in robots.txt is not reported as healthy.

The report also groups issues into nine areas, each with its own score (speed, Google & search, cookies & privacy, security, accessibility, email & domain, reputation, technical quality and carbon footprint). Each area starts at 100 and deducts by issue priority: 35 for a critical one, 14 for a high one, 4 for a low one and 0 for a tip. It shows where the work is concentrated; the overall score comes from the checks.

Site authority (0-10) and the TLS grade (A+ to F) are shown as data and do not change the 0-100 score. Authority combines Common Crawl (40%), Tranco (30%), domain age (20%) and Wikidata (10%); if a source is missing its weight is redistributed, and with fewer than two we give no figure. It is not Moz’s “Domain Authority” or the SSL Labs grade, and the report says so.

How we keep measurement noise out

Some measurements are stable (a DNS record, a header, a cookie) and others are not. PageSpeed’s lab test can differ by several points on the same site in two back-to-back runs, because it depends on the load on Google’s servers at that moment. Treating that as a real change would be misleading.

Rigour in every data point: what happens when a source fails

The rule is simple: never pretend. If a data point could not be measured, the check shows as “not measured”, does not count towards the score and the report says why. Some common cases:

Data sources

Every issue in the report names its source. If a source fails or has no data for your site, the report carries on without it and says so.

SourceWhat for
Google PageSpeed Insights v5Lighthouse on mobile and desktop: performance, accessibility, best practices and SEO.
Chrome UX Report (CrUX) & HistoryReal Chrome user data (28 days) and its weekly trend.
Google Safe Browsing v4Whether Google flags the site as dangerous (malware, phishing).
Mozilla HTTP Observatory v2Security header grade.
Google DNS & Cloudflare DNSA, AAAA, NS, MX, CAA, SPF, DMARC, DKIM, MTA-STS, TLS-RPT, BIMI, DNSSEC.
DNSBLSpamhaus, SpamCop, Barracuda and other spam blocklists.
RDAPRegistrar, creation date, expiry and locks of the domain.
IP RDAP & RIPEstatWho hosts the site, network, ASN and country; RPKI and abuse contact.
Shodan InternetDBOpen ports and known vulnerabilities for the IP.
TLSCertificate, issuer, expiry, TLS versions, HTTP/2.
Certificate Transparency (crt.sh, Cert Spotter)Issued certificates and public subdomains.
hstspreload.orgWhether the domain is on the HSTS preload list.
TrancoDomain popularity.
Wayback MachineSite age and home page changes.
Common CrawlPresence in the large public archive used by many AI models.
Wikidata & WikipediaWhether your brand exists as an official entity.
Have I Been PwnedPublic data breaches for the domain.
Nu HTML Checker & W3C CSS ValidatorHTML and CSS code errors.
Website Carbon & Green Web FoundationCO₂ per visit and whether the hosting is green.
endoflife.date, OSV.dev, Retire.jsUnsupported software and vulnerable libraries.
WordPress.orgRemoved, abandoned or outdated plugins.
Open Cookie Database, EasyPrivacyWho sets each cookie and why.
Phishing.Database, Block List Project, Spamhaus DROPOpen phishing, scam and hijacked-network lists.
VIES (European Commission)VAT number in the Spanish legal notice (Spanish sites only).
ai.robots.txtWhich AI crawlers your robots.txt blocks or allows.
Chromium + axe-coreReal browser: cookies before consent, accessibility, technologies and screenshots.
Our own visit and crawlRedirects, headers, robots.txt, sitemap, llms.txt, Open Graph, JSON-LD and a sample of up to 30 pages. We never submit forms.
Lighthouse (Apache-2.0)Fallback when PageSpeed Insights fails: mobile performance with our own headless Chromium, in a separate process.
Common Crawl Web GraphDomain position by harmonic centrality among some 100 million domains (we keep the top 10 million).
Our own TLS probeTLS versions, ciphers, key, chain, OCSP and HSTS, graded A+ to F.

AI, transparently

The free report uses no AI: the score, the issues and the summary come from fixed rules applied to measured data. In the Pro plan, an AI (Claude, by Anthropic) only writes the summary, the action plan and the weekly digest, and answers “Does AI recommend you?”. It only receives the measured data of the audited site, never your email. Before showing its text, we extract every figure it quotes and look it up in that data: if any is missing, the text is discarded and the rule-based one is used. The score is never set by AI.

An audit that respects your website

Who is behind it? See About.

Audit my website

Plan Pro · $19.90/month

Let someone keep an eye on your website

The free report is a snapshot. Pro is the movie: automatic weekly audits, alerts, “Does AI recommend you?”, competitor benchmarks and branded PDFs. $19.90/month, taxes included, cancel any time.

See the plan · $19.90/month