Methodology

How we work out each score

What goes into the AI Visibility Score and the Trust Score, how often we check, and how sure we are of each number. Below those, how each of our other scores and audits is worked out.

Last updated 09 October 2026See a sample report

AI Visibility Score

The AI Visibility Score is the share of relevant AI answers that name or cite you. We work it out for each product in each market, then roll it up for the brand. For every answer we record three things:

Mentioned

The answer names your brand.

Cited

The answer links to one of your pages as a source.

Position

Where you appear in the answer, compared with competitors.

Prompt sampling

We build a prompt panel for each product and market: the questions your buyers actually ask an AI engine before they choose. Each prompt is tagged to a product, a market and a buying stage.

  • 6 prompts per product and market by default
  • Prompts reviewed with you when the panel is set up
  • Competitors set for each product, not just for the domain

The five engines

We track ChatGPT, Claude, Gemini, Perplexity and DeepSeek. Each prompt runs on every engine, and we store the full answer with its citations, so you can read exactly what each engine said.

CChatGPTCClaudeGGeminiPPerplexityDDeepSeek

Each engine is queried through its official API with web search turned on, so answers reflect what a buyer would see. The free audit uses four engines; Claude is included on paid plans.

How often we check

Each prompt runs [weekly] on every engine. Scores show the latest run and the trend across the last [12] weeks.

Confidence

AI engines don't give the same answer twice. Each score carries a confidence level based on how many prompts and runs sit behind it, so a score built on a handful of answers is marked as such.

High

[Threshold]

Medium

[Threshold]

Low

[Threshold]

Trust Score

The Trust Score estimates how much a stranger, human or AI, would trust your site on first look. It combines signals across identity, expertise, legal pages, reputation and structured data, scored out of 100.

How it's scored: the Trust Score is the share of applicable trust signals your site passes. We check 76 signals in seven groups: expertise and content; legal and compliance; identity and contact; reputation; schema and technical; trust and transparency; brand and culture. Signals that don't apply to your type of business are left out of the score. The free audit checks your homepage only, against 54 of the 76 signals.

AI search

More from the same AI answers, and the on-site checks that make you easier to quote.

Citations, sentiment and other brands

Every answer stored for the AI Visibility Score is read three more ways: which sources it cites, how it describes you, and which other brands it names.

Citation sources

Each cited domain is classified into one of ten categories, such as review platforms, forums or publishers, and marked as yours, a competitor's or a third party's.

Sentiment

Each answer that names you or a competitor is classed as positive, neutral or negative.

Other brands

Brands named in answers that aren't on your competitor list are collected, so you can add or dismiss them.

How it's scored: Citation rate is the share of answers that cite a domain. Citation share is a domain's share of all citations in scope. Competitor gap rate is the share of answers where a source backs a competitor and not you.

  • Position is the order in which brands are first mentioned in an answer, not a ranking
  • Gemini reports source domains rather than full page links
  • DeepSeek cites sources only when it searches the web for an answer

AEO score

Answer Engine Optimization (AEO) measures how well each page is structured to be used as a direct answer: a featured snippet, a voice answer or an AI answer. It is rule-based; no AI model decides the score.

Structure and content

Question-style headings, short direct answers, lists and tables, and internal links.

Schema and canonical

FAQ and other structured data, and a correct canonical setup.

Search signals

The queries each page already ranks for in Search Console, results-page features and page speed.

How it's scored: each page starts at 100 and loses points for every issue found, with more serious issues costing more. The site score summarizes the audited pages. A separate site-level check flags AI crawlers blocked in robots.txt.

  • Needs Google Search Console connected, because several checks use the queries each page ranks for
  • Checks access for GPTBot, ClaudeBot, PerplexityBot and GoogleOther

GEO score

Generative Engine Optimization (GEO) measures whether your site is built so AI answer engines can read, trust and cite it. It is an on-page and crawl-access audit. Whether engines actually name you is measured separately, by the AI Visibility Score.

Site checks

AI crawler access in robots.txt, and whether you publish an llms.txt file.

Page checks

18 rule-based checks per page: schema, author, content depth, question headings, citations, tables, lists and heading structure.

Content review

An AI review of your site's content against a set of quality judgments.

How it's scored: page and site scores each start at 100 and lose points for every issue found. The GEO score combines the site and page results. The AI content review raises issues to fix but doesn't change the score.

  • The free audit runs page checks on up to 10 pages
  • Full audits need Google Search Console connected
Website and trust

How we audit the site itself, beyond the Trust Score above.

Technical SEO

The technical audit crawls your site the way a search engine does, as AdvioraBot, and checks crawlability, indexability, site architecture, HTTPS and security, structured data, mobile rendering and page speed.

Crawl

A sample of pages by default, or your sitemap, or the full site. [Confirm crawl limits per plan.]

Rendering and speed

Mobile rendering in a headless browser, plus Google PageSpeed Insights for mobile and desktop.

Severity

Every issue is graded Critical, High, Medium, Low or Info.

How it's scored: there is no single technical score. Issues are counted by severity, so the most damaging ones are fixed first, and each issue maps to a recommended fix.

  • Page speed combines lab measurements (LCP, CLS, TTFB and others) with PageSpeed Insights
  • No AI model is involved in the technical audit

Email domain score

The email domain audit reads your domain's public DNS records. Because it reads public records only, it can't tell whether a particular email passed authentication.

Authentication

SPF, DKIM and DMARC, combined into one score out of 100.

Transport security

MTA-STS and TLS-RPT, scored separately.

Brand and reputation

A BIMI record check and blocklist checks.

How it's scored: [Weighting of SPF, DKIM and DMARC to confirm before publishing.]

Search and content

Where keyword data comes from, and how we judge overlap and content gaps.

Keyword data and intent

Keyword metrics (search volume, cost per click, competition and seasonality) come from licensed search data and are refreshed regularly. Your own queries come from Google Search Console and Google Ads.

Metrics

Volume, cost per click and competition, with an optional live snapshot of the results page.

Intent

Each query is classified into a buying stage: Awareness, Consideration, Decision or Trust-check.

Topics

Queries are grouped by topic. Brand terms are removed before customer segments are built.

How it's classified: an AI model labels each query with its intent and topic.

Cannibalization Risk Index

When two of your pages compete for the same searches, they split clicks and rankings. Adviora finds those pairs from Search Console overlap and a semantic comparison of the keywords each page ranks for, then confirms likely pairs with live search data.

Window

Overlap uses the last 30 days of Search Console data.

Which page to keep

The page with more clicks over the last 90 days.

Recommended action

Merge and redirect, canonical, noindex, differentiate, or expand the existing page.

How it's scored: the Cannibalization Risk Index (0–100) combines five factors: how much the pages overlap in search, how traffic is split between them, how similar their engagement is, how much their rankings swap, and how similar their content is. Only pairs above a minimum risk level are reported, and the highest are marked high risk.

  • The pre-publish check uses the same comparison to tell you whether an existing page already covers a keyword
  • Risk levels will be refined as data grows

Content intelligence and buyer personas

Content intelligence and buyer personas are built for each product and location from your Search Console, Analytics and keyword data.

Trust deficit

Observed, latent and effective trust deficit scores show where missing proof holds content back. The Effective Trust Deficit Score (ETDS) drives the recommendations.

Industry thresholds

Each industry profile, such as education, healthcare, finance, SaaS or ecommerce, has its own trust threshold.

Personas

Intent split, emotional drivers, devices and the questions each audience asks, with a data coverage score showing how much evidence sits behind them.

How it's scored: [How the trust deficit scores are calculated. To confirm with the Intelligence team.]

  • Needs Google Search Console and Analytics connected
Market and customers

How we compare you with competitors and find the customers most ready to buy.

Competitor scores

Each competitor is audited across 16 dimensions, from backlinks, E-E-A-T, AEO and GEO to traffic, hiring, reviews, ads and news, plus a short AI summary of positioning, growth, threat and opportunity.

Six scores

GEO, AEO, E-E-A-T, backlinks, traffic and hiring, each scored with the same methods we use for your site.

Overall

One score that combines the six.

Head-to-head

Each competitor compared with you, metric by metric.

How often: competitors are re-audited weekly, and Change Monitor compares consecutive audits. Recent Updates checks the pages you choose every 3, 6, 12 or 24 hours.

  • Keyword data shows the opportunities a competitor targets, not live rankings

Share of voice

Share of voice compares how often you and your competitors are mentioned, and how much engagement those mentions get, across Facebook, Instagram, YouTube, TikTok, Reddit, X, news and Quora.

How it's scored: mention share = your mentions ÷ all mentions of you and your competitors. Engagement share = your engagement ÷ all engagement on those mentions. Both cover the last 30 days, with a week-by-week comparison.

  • Public posts are collected for each brand on each platform
  • News counts mentions only, because news coverage has no engagement data
  • No social account connection is needed

Customer discovery (PDRS)

Customer discovery groups search, ad and traffic signals into customer segments and scores each one with the Prospect Discovery Readiness Score (PDRS), from 0 to 100.

Signals

Search Console, Google Ads, Analytics and keyword data, with brand terms removed.

Segments

Grouped by topic, country and intent. A segment needs enough signals and search volume to be reliable.

Plans

An AI-written activation plan for each segment, and a short summary of who, where, why, what and how.

How it's scored: PDRS combines five factors: intent strength, fit, trust readiness, competitive displacement and channel efficiency. Hot is 75 or more, Warm is 55 to 74, and Cold is below 55.

  • Scores refresh each time customer discovery runs
Ads and action

How we grade a Google Ads account, and how findings become a ranked list of fixes.

Google Ads benchmark score

The Google Ads audit reads the last 90 days of a connected account and checks it against 81 benchmarks in 15 areas, from conversion tracking and bidding to Performance Max and Shopping.

Evidence

Each result says what it is based on: your account data, official guidance, AI-assisted review or manual review.

No guessing

If a check can't be worked out from your data, the result says so. It is never marked as a pass.

Analysis

An AI model writes the plain-language analysis from the results. It doesn't change the score.

How it's scored: each benchmark passes, warns or fails. Benchmarks sit in tiers, and the most important carry more weight. The result is a score out of 100 with a maturity band. Benchmarks that don't apply, such as Shopping checks without Shopping campaigns, are left out.

Recommendation priority

The growth roadmap turns the latest results from the E-E-A-T, AEO, GEO, technical SEO, backlinks, content intelligence and persona audits into one ranked list. Each issue is matched to a benchmark rule; no AI model decides the priority.

Approval

Each item carries an approval level: execution, strategy lead or business owner.

Outcome check

30 days after a fix is executed, the related score is measured again, starting with E-E-A-T items.

Without Google

If Google isn't connected, the roadmap uses E-E-A-T, technical SEO and backlinks only.

How it's scored: priority combines each issue's expected impact, urgency and effort, and the most serious issues always come first. Each item is marked Critical, High, Medium or Low.

Limits

A single AI answer is a sample, not a verdict. Engines change their models and sources without notice, so scores can move for reasons outside your site. We show the trend and the confidence next to every score for this reason, and we never promise that an engine will cite you.

See your own scores

Run the free audit. It only needs an email address.

Start free audit