Methodology, in public

Every audit check has a methodology page behind it.

Every check links to the page that explains what it measures, why it matters, how it is scored, what the confidence flag means, the primary sources, and what we changed last. A reader can trace any score we produce back to the academic paper, vendor documentation, or Google docs that justify it.

70 methodology pages published. More land every week as the check registry grows.

AI Visibility pillar

AI Visibility

API answer samples from ChatGPT, Claude, Perplexity, Gemini, Grok, AIO. Passage extractability. Schema for LLMs.

Check: ai-crawler-readability

AI crawler readability (raw HTML token ratio)

A page should expose at least 70 percent of its rendered DOM token count in the raw HTML, since GPTBot, ClaudeBot, and PerplexityBot do not execute JavaScript.

Read the methodology →

Check: schema-completeness-for-ai

Page-type structured data

Whether a page carries the structured data that matches what the page actually is, and whether that markup matches what visitors can see. Google requires no special schema for its AI features.

Read the methodology →

Check: ai-citation-presence

AI citation rate and measurement stability (the Variance Engine)

AI engines are non-deterministic: ask the same question five times and the answer moves. We poll the identical prompt N times per engine (one to two runs today), count how often a brand is actually cited, publish a 95 percent Wilson confidence interval anyone can recompute, and flag results that are too volatile to optimize yet.

Read the methodology →

Check: engine-divergence

Engine divergence: where the AI engines disagree about citing you

The AI engines do not agree with each other. We ask every connected engine the same question and surface exactly where they split, so "cited by Perplexity, invisible in ChatGPT" becomes a visible, verifiable fact rather than a guess.

Read the methodology →

Check: source-of-citation

Source of citation: which sources actually drive your AI mentions

When an AI engine cites you, it cites you through a source. We collect the real URLs each engine returned and group them by domain, so you can see exactly which sources are feeding your AI mentions, instead of guessing.

Read the methodology →

Check: citation-index-panel

The Citation Index panel: a standing record of who gets cited

We run our own standing, slow-cadence cross-engine measurement panel, asking a fixed set of questions across engines on a regular schedule, to build a longitudinal record of who gets cited for what in each vertical. The published Index comes later; this is the measurement substrate behind it.

Read the methodology →

Check: vertical-benchmarks

Vertical benchmarks: how you compare, without exposing anyone

We publish anonymized cohort benchmarks for each vertical, for example "the median brand in this category is cited by 2 of 6 engines," computed only from organizations that opt in, and only when a cohort is large enough that no single organization can ever be identified.

Read the methodology →

Check: entity-graph

Entity graph: can AI engines resolve your brand?

Before an AI engine can cite you confidently, it has to know who you are. We read your Organization schema and the identity links it points at, and report how well-anchored your brand is as an entity the knowledge graph can resolve.

Read the methodology →

Check: passage-extractability

Passages that stand on their own: our editorial guideline for quotable paragraphs

Our editorial guideline for AI-answer extraction, not a Google rule: we check whether each main paragraph makes sense when quoted alone, because Bing says content that stands on its own is more likely to be selected for grounding and citations. A heatmap of every paragraph against a 134 to 167 word window rides along as secondary evidence; Google says it has no preferred word count.

Read the methodology →

Check: ai-crawler-access

AI crawler access: can AI engines actually fetch you?

An AI engine that cannot fetch your page cannot cite it. For each major AI bot we check both what robots.txt allows and what your site actually serves when that bot knocks, because a CDN or firewall can block a crawler your robots.txt never mentioned.

Read the methodology →

Check: attorney-authorship-credentials

Attorney authorship and credentials

Checks that legal content shows who wrote or reviewed it, through an attorney byline or Person schema with credential properties.

Read the methodology →

Check: clinician-authorship-credentials

Clinician authorship and credentials

Checks that a medical site shows a named clinician with credentials, a medical-review line, or credentialed schema.

Read the methodology →

Check: clinical-citation-presence

Clinical citations to primary sources

On pages with clinical content, checks for links to primary medical sources such as NIH, CDC, FDA, PubMed, or peer-reviewed journals.

Read the methodology →

Check: ai-method-citations

Citations behind AI method claims

When an AI-agency page makes model, benchmark, or dataset claims, checks for links to the papers, datasets, or repositories behind them.

Read the methodology →

Check: people-entity-presence

People behind the business

Whether a visitor and an answer engine can see the people behind your business, and whether your structured data describes them.

Read the methodology →

Check: author-credentials-generic

Author credentials

For the people a page names, whether it shows why to trust them: a role, a license or certification, or years of experience.

Read the methodology →

Check: organization-name-consistency

Consistent business name

Whether your Organization markup, WebSite markup and site name call the business by one name.

Read the methodology →

Check: sameas-verified

Declared profiles exist

Whether the sameAs profile URLs in your structured data load, and whether they point back to your business.

Read the methodology →

Check: ai-answer-snippet-controls

Snippet and archive controls that limit AI answers

We read every robots meta tag (robots, googlebot, bingbot) and X-Robots-Tag directive plus data-nosnippet coverage, and show the effect Google and Bing document for each one on AI answers.

Read the methodology →

Check: definition-statements

A plain sentence saying what the business is

Our editorial guideline, not a Google rule: near the top of the page, one plain sentence should say what the business is ("Acme Law is an estate planning firm in Baltimore"). We also collect definitions under "What is" headings.

Read the methodology →

Check: question-headings-answered

Question headings answered at once

Our editorial guideline, not a Google rule: when a heading asks a question, the first sentence under it should answer it. We flag deferrals, fragments, run-ons and answers that never mention the topic.

Read the methodology →

Check: author-entity

Article authors: visible byline and author markup that agree

On article pages only, we check that a visible byline and the structured data author exist, name the same person, and follow Google's author markup best practices. Homepages and service pages are not measured here.

Read the methodology →

Check: content-freshness

Article freshness date

On article pages only, we read every published or updated date signal and report the newest one and its age at the time of the audit. Homepages and service pages do not need a date and are not measured.

Read the methodology →

Check: faq-schema

FAQ markup and what it does today

Information only. Since August 2023 Google shows FAQ rich results only for well-known, authoritative government and health sites. FAQ markup still describes your questions in a machine-readable way, but do not expect FAQ rich results.

Read the methodology →

Check: howto-schema

HowTo markup and what it does today

Information only. Google retired HowTo rich results in September 2023. Clear, numbered, visible steps are what matter; HowTo markup is optional.

Read the methodology →

Check: llms-txt-noop

llms.txt: why we do not score it

We never score llms.txt and never recommend adding one. Google says you do not need AI text files to appear in its AI features, and no major answer engine has said llms.txt affects citations.

Read the methodology →

Check: citation-worthiness

Quotable facts on the page

Our editorial guideline, not a Google rule: we count the facts an answer engine could quote (numeric statements, attributed quotations, your own data), with phone numbers, dates and prices removed first.

Read the methodology →

Technical pillar

Technical

Crawl, render, schema validation, Core Web Vitals, render-parity. What AI crawlers see.

Check: js-render-parity

JS render-parity diff (raw HTML vs rendered DOM)

The raw HTML returned by a non-JS fetch should contain at least 70 percent of the content that a JS-rendered DOM contains, because AI crawlers do not execute JavaScript.

Read the methodology →

Check: schema-validation

JSON-LD validity (schema.org spec conformance)

Every JSON-LD block on the page should declare @context (https://schema.org), declare a @type, and parse as valid JSON.

Read the methodology →

Check: core-web-vitals

Core Web Vitals (LCP, INP, CLS)

LCP under 2.5 seconds, INP under 200 milliseconds, and CLS under 0.1, Google's public thresholds at the 75th percentile of real-user traffic. We report both CrUX real-user field data and Lighthouse lab data, clearly separated, and only the real-user field data affects the score.

Read the methodology →

Check: date-signal-consistency

Date signals that agree

A measured fact about your markup: the published and updated dates in your structured data, meta tags and visible byline should agree with each other and never point to the future.

Read the methodology →

Check: outbound-broken-links

Outbound links in your content that no longer load

A measured fact we fetch: links from your page copy to other sites, citations first, that return not found or point to a domain that no longer exists.

Read the methodology →

Check: bing-crawl-access

Bing crawl access: can Bing read your site?

We read your robots.txt the way Bing does (a bingbot group, else msnbot, else the catch-all group) and check that bingbot may fetch your homepage and the audited page, that no bingbot-only noindex is set, and that any Crawl-delay is within what Bing recommends.

Read the methodology →

Check: bing-webmaster-verification

Bing Webmaster Tools verification

We look for the two verification methods visible from outside (a msvalidate.01 meta tag or a BingSiteAuth.xml file). Finding one is a pass. Not finding one proves nothing, so it is advisory guidance, never a score.

Read the methodology →

Check: indexnow-readiness

IndexNow: notify Bing when pages change

Guidance only. IndexNow lets a site tell participating search engines the moment a page is added, updated or deleted. Nobody outside your site can see whether you use it, so we never score it.

Read the methodology →

Check: structured-data-visible-match

Structured data that matches the visible page

We compare FAQ questions and answers, review text, author names and organization names in your structured data with the text a visitor can see, and flag markup that describes text that is not on the page.

Read the methodology →

Check: root-response-access

Page loads as content: did our crawler get your real page?

We check whether the page you asked us to audit answered our crawler with a readable web page, or with an error, a refusal, a rate limit, something that is not a web page, or a bot check. When it did not, we grade nothing about your content and tell you what the server sent instead.

Read the methodology →

Foundational pillar

Foundational

Keyword universe, content depth, on-page execution. The base every other pillar sits on.

Check: h1-presence

H1 presence and uniqueness

Every indexable page should declare exactly one non-empty H1 element distinct from the document title.

Read the methodology →

Check: title-quality

Title presence and our title length guideline

A missing title is a measured gap. The 30 to 60 character band is our editorial guideline, not a Google rule: Google says there is no title length limit and truncates long titles as needed, while Bing says overly short titles may reduce indexing reliability and eligibility for grounding and citations.

Read the methodology →

Check: meta-description

Meta description (length and CTR signal)

A meta description in the 50 to 200 character range helps Google select the snippet text and gives AI engines a clean topic summary.

Read the methodology →

Check: thin-content

Main content sufficiency

Flags pages that are empty, made mostly of the shared site template, or that never cover the topic their own heading promises. There is no word-count target, because Google has none.

Read the methodology →

Check: keyword-in-title

Topic alignment between title and page

Our editorial guideline for AI-answer extraction, not a Google rule: a check that the title names the topic the page declares in its own headings, or, on the homepage with Search Console connected, the queries the site is actually shown for. Bing asks sites to align titles, headings and content intent.

Read the methodology →

Check: reading-level

Reading ease: our editorial readability guideline

Our editorial guideline for AI-answer extraction, not a Google rule: we compute the Flesch reading ease of the page copy and compare it with our 50 to 70 guideline. No engine publishes a readability score; Bing asks for content that is easy to understand without external context.

Read the methodology →

Check: attorney-responsible-contact

Responsible lawyer or firm and contact details

Checks that a law firm page names a responsible lawyer or firm together with at least one way to reach them, the baseline in ABA Model Rule 7.2(d).

Read the methodology →

Check: attorney-jurisdiction-disclosure

Bar admission jurisdictions named

Looks for wording that tells visitors where the firm’s lawyers are admitted to practice, such as "licensed in Maryland and Virginia" or a state bar reference.

Read the methodology →

Check: attorney-client-relationship-notice

Attorney-client relationship notice

On pages that invite inquiries, checks for a notice that contacting the firm does not by itself create an attorney-client relationship.

Read the methodology →

Check: prior-results-disclaimer

Prior-results disclaimer next to case results

When a page shows verdicts, settlements, or case results, checks for wording that past results do not guarantee a similar outcome.

Read the methodology →

Check: privacy-practices-notice-link

Notice of Privacy Practices on the website

Checks for a link to a Notice of Privacy Practices, which HIPAA requires a covered entity to post prominently on a website about its services.

Read the methodology →

Check: medical-outcome-claims

Guaranteed health-outcome claims

Screens medical pages for absolute outcome promises such as "guaranteed cure" or "100% success" that need substantiation.

Read the methodology →

Check: cpa-credential-signals

CPA credential signals

When a site uses the CPA title, checks that it is tied to named licensed CPAs, a firm permit or board reference, or a license lookup.

Read the methodology →

Check: tax-outcome-claims

Guaranteed tax-outcome claims

Screens tax and accounting pages for guaranteed refunds, savings, or debt-elimination promises.

Read the methodology →

Check: fair-housing-notice

Equal Housing Opportunity notice

Checks for the Equal Housing Opportunity logotype, statement, or slogan, which HUD guidance treats as alternatives.

Read the methodology →

Check: brokerage-license-disclosure

Brokerage identification in advertising

Checks that a real-estate site identifies the supervising brokerage or license, as state advertising rules require.

Read the methodology →

Check: fair-housing-language-review

Preference or limitation language in listing copy

Screens real-estate copy for explicit preference or limitation phrases, such as "adults only" or "no children", that the Fair Housing Act bars in advertising.

Read the methodology →

Check: financing-cost-disclosure-cues

Cost-of-capital context near advertised rates

When a financing site advertises rates or factor rates, checks for total-cost or APR context so advertised numbers do not contradict offer-time disclosures.

Read the methodology →

Check: financing-approval-claims

Guaranteed approval or funding claims

Screens financing pages for guaranteed-approval or unconditional-funding promises that underwriting may not honor.

Read the methodology →

Check: financing-licensure-disclosure

Licensing or registration reference

Checks whether a financing site references the licenses or registrations it holds, which depend on the product and the states served.

Read the methodology →

Check: ai-performance-claims

AI performance claims hygiene

Screens AI-agency pages for guaranteed rankings or citations and absolute accuracy claims that need substantiation.

Read the methodology →

Check: trust-pages-presence

About, Contact and Privacy pages

Whether a visitor can reach an About page, a Contact page and a privacy policy from your homepage, and whether those pages load and say something.

Read the methodology →

Check: primary-source-citations

Primary source citations

Whether the statistics your page states link to a source in the same paragraph, and whether any of that sourcing is primary.

Read the methodology →

Check: skimmability

Skimmability: can a reader find the point fast?

Our writing methodology, not a Google ranking rule: we look at how the main content is broken up, by headings, lists, tables and paragraph length, and flag long walls of unbroken prose.

Read the methodology →

Check: filler-and-unsupported-claims

Filler and unsupported claims

Our writing methodology, not a Google ranking rule: we flag stock phrases that add words without adding information, and superlatives with no number, source or named award beside them.

Read the methodology →

Check: in-page-repetition

In-page repetition and template text

A measured fact about the page: sentences repeated within it, and, on multi-page audits, how much of its text appears word for word on other pages we crawled.

Read the methodology →

Check: buyer-question-coverage

Buyer question coverage

Our usefulness methodology, not a Google rule: for your industry, we list the questions buyers usually ask before they call, and check whether the pages we read answer each one in plain words.

Read the methodology →

Check: dated-statistics

Statistics with a date and a source: our sourcing methodology

Our sourcing methodology, not a Google ranking rule: we list every number your page states as a fact and check whether each one says when it was true and where it comes from.

Read the methodology →

Check: stale-year-references

Past years presented as current

A measured fact about your page: we look for titles, headings and copy that present a year that has already passed as current, such as "2024 guide" in 2026.

Read the methodology →

Methodology changelog

Changes

  1. Score model v3, trust, writing, data and Bing checks

    • New score model (v3). Trust, writing, usefulness and citation-readiness findings are our methodology, not engine rules, so together they can lower a pillar only up to a fixed limit. Measured defects (a blocked crawler, a broken citation, missing markup) have no limit.
    • A new finding kind, advisory, covers things nobody can see from outside your site, such as an IndexNow key or Bing verification by DNS. Advisory items explain what to check yourself; they never score, never count toward coverage and never enter your action plan.
    • The overall readiness score now appears only when every pillar that applies to your industry has enough measured checks, and it always averages the same pillars. A check that newly starts measuring can no longer raise the overall on its own.
    • Scores from the previous model (v2) and this one are not compared. Re-audits and weekly digests show "the scoring model changed" instead of a score change.
    • Pages that answer with a bot check, an error or something other than a web page are no longer graded as your content. Content checks on those pages show as not measured, with what the server sent.

    23 new checks

  2. Local checks apply to businesses with a location or service area

    • Local SEO is for businesses that serve customers at a location or in a service area. A general audit now runs the local checks only when the pages we read show one: a street address or a city and state, a Maps link, LocalBusiness or opening-hours schema, locations or service-area pages, or a city or state in the title or headings.
    • When none appear, an online-only site sees each local check marked not applicable with that reason. The local pillar then has no score, it is left out of coverage and of the overall readiness score, and it never enters the action plan. Accountants, lawyers, medical practices and real estate audits always run the local checks.
    • The LocalBusiness schema check now accepts every LocalBusiness subtype in the schema.org vocabulary, not a short list, so a page that declares MedicalClinic or HVACBusiness is no longer told it has no LocalBusiness schema.