Live, scanning the open webAI Visibility Audit Platform
Methodology · The AI Visibility Score

How the AI Visibility Score works

AudFlo is an AI Visibility Audit Platform for founders. The AI Visibility Score is how it measures whether AI will recommend you: a number from 0 to 100, built from four weighted pillars and 32 Checks.

0–100
single visibility score
4
weighted pillars
32
diagnostic checks
0–100 ScoreOne number for how recommendable you are.
4 PillarsWeighted dimensions of AI readability.
32 ChecksEight diagnostic layers per pillar.
5 FactorsThe recommendation formula behind it all.
Executive summary

The short version

The AI Visibility Score is one number from 0 to 100 that reflects how easily AI engines like ChatGPT, Claude, Gemini, and Perplexity can understand, classify, and recommend your product. It comes from 32 diagnostic layers grouped into four weighted pillars that add up to 100%. Technical Visibility, worth 20%, asks whether AI engines can reach and read your site. Structural Understanding, worth 30%, asks whether they can tell what your product is and who it is for. Answer Selection, worth 30%, asks whether they will pick you when a buyer asks a category question. Authority Signals, worth 20%, asks whether your site looks trustworthy to people and machines. A higher score means a higher chance of being recommended when it counts.

What we measure

The four pillars

Each pillar measures a different dimension of AI engine readability. The pillars are weighted because not all signals matter equally, together they make up 100% of your AI Visibility Score.

i.

Technical Visibility

20% of score8 layers

What it measures

Whether AI engines can crawl, parse, and trust your site. The eight layers cover indexability, robots.txt configuration, sitemaps, schema markup, llms.txt, meta tags, page speed, and social meta.

Why it matters

If AI engines cannot reach your site or read its structure, none of the other signals matter. A site that is blocked from crawling or missing structural cues scores zero on this pillar regardless of content quality.

How it is scored

We run eight checks. Each one passes, partly passes, or fails. Added up, they form the score for this pillar, which counts for 20% of your total.

What fixes improve it

Generating an llms.txt file, adding JSON-LD schema to key pages, submitting a sitemap, and confirming robots.txt allows AI crawlers are the four highest-leverage fixes. See the technical layers checked on a real site in the sample audit.

ii.

Structural Understanding

30% of score8 layers

What it measures

Whether your content states clearly what your product is, who it serves, and what category it belongs to. The eight layers evaluate your primary headline, category statements, word count, reading level, jargon density, and paragraph structure.

Why it matters

AI engines classify products in one or two sentences before deciding whether to recommend them. Vague or jargon-heavy positioning means an AI engine cannot confidently describe your product to a buyer, so it recommends a competitor it can describe.

How it is scored

Each check looks at one thing and measures it against a clear standard. Added up, they count for 30% of your total, which ties this for the most important pillar.

What fixes improve it

Rewriting your homepage headline to name your category, your buyer, and your outcome in one sentence is the fastest win. A clear category statement like the definition block on the AudFlo homepage is the target pattern. Adding a product description in plain English and removing unexplained acronyms also raises this pillar quickly.

iii.

Answer Selection

30% of score8 layers

What it measures

Whether an AI engine will pick you when a buyer asks a category question. The eight layers measure FAQ presence and FAQPage schema, comparison content, use case pages, question-shaped headings, and topic coverage breadth.

Why it matters

AI engines answer questions by assembling the most complete, structured response they can find. Sites with published answers to common questions, and clear comparison content showing how they differ from alternatives, are far more likely to be cited or recommended.

How it is scored

Each check looks for clear answers to the questions buyers actually ask. A proper FAQ that AI can read and quote scores higher than the same answers buried in paragraphs. Added up, they count for 30% of your total.

What fixes improve it

Adding a FAQ block with FAQPage schema markup and publishing a comparison page naming your category alternatives are the two highest-leverage moves. The AudFlo comparison page is an example of comparison content structured to score well on this pillar.

iv.

Authority Signals

20% of score8 layers

What it measures

Whether your site signals trustworthiness to both people and AI engines. The eight layers cover testimonials, customer logos, founder presence, customer stories, Review schema markup, Person schema markup, and external authority signals.

Why it matters

AI engines weight trust signals when choosing between two sites with similar structural scores. A site with named testimonials, a credible founder page, and Review schema consistently outperforms one with generic "loved by thousands" copy.

How it is scored

Each check looks for real proof that people trust you. Testimonials with real names, a founder page, and reviews AI can actually read score higher than vague claims. Added up, they count for 20% of your total. We weight this lower than Structural Understanding and Answer Selection because trust takes longer to build than the other fixes.

What fixes improve it

Adding a founder page with credibility signals, wrapping existing testimonials in Review schema, and publishing at least one named customer story are the three fastest wins. The AudFlo founder page is an example of a founder credibility page built to score on this pillar.

The formula

The five recommendation factors

The four pillars are how AudFlo measures your score. They all serve one question from the AI Visibility Playground: when an AI engine decides whether to recommend you, how confident is it?

Recommendation Confidence = Accessibility × Classification × Evidence × Entity Confidence × Contextual Fit

These five factors multiply. If any one is near zero, the whole result collapses. Each factor is measured by one or more of the four pillars, and read from signals AudFlo already collects during a scan:

Accessibility. Can AI crawlers reach and read your site?

Measured by Technical Visibility. Read from robots.txt access, llms.txt, sitemap, schema presence, meta tags, and load speed.

Classification. Can the AI tell what category you are in and who you serve?

Measured by Structural Understanding. Read from the primary headline, title tag, meta description, and category statement.

Evidence. Can your claims be verified?

Measured by Authority Signals. Read from testimonial count and naming, customer stories, founder presence, and Review schema, graded against the Evidence Ladder.

Entity Confidence. Does the AI recognize your brand as a real, distinct entity?

Measured by Structural Understanding + Authority Signals. Read from name and category consistency across the headline, title, and metadata, plus Organization and SoftwareApplication schema and external mentions.

Contextual Fit. Are you the best match for the exact prompt?

Measured by Answer Selection. Read from FAQ coverage and FAQPage schema, comparison and use-case pages, and question-shaped headings that match real prompts.

Two factors worth naming

Entity Confidence is whether an AI can tell your brand apart from similarly named products. AudFlo reads it from how consistently your name and category appear across your headline, title, and metadata, plus your Organization and SoftwareApplication schema, and it lives inside Structural Understanding and Authority Signals. Contextual Fit is whether you are the best match for the exact prompt a buyer types. AudFlo reads it from your FAQ, comparison, and use-case coverage, and it lives inside Answer Selection.

Entity Confidence

Who are you?
What category are you in?
Are mentions of you connected to the same entity?
Can an AI tell you apart from similar names?
Can an AI recommend you confidently?

The full breakdown is in the Entity Confidence framework.

Inside Authority Signals

The Evidence Ladder

The Evidence factor is not simply pass or fail. AudFlo grades the proof on your site against five tiers inside the Authority Signals pillar, because AI engines do not trust every signal equally. A vague claim is weak. A named customer story is stronger. Original data is strongest.

Stronger, harder to fake

Tier 1Original Research
Tier 2Third-Party Validation
Tier 3Case Studies
Tier 4Testimonials
Tier 5Marketing Claims

Weaker, easy to fake

Tier 1Original research, benchmarks, and proprietary data

The hardest evidence to copy and the easiest to cite. A figure only you have is one an AI must attribute to you.

AudFlo check: Original data or research present?

Tier 2Independent mentions, reviews, and coverage you do not own

Independent sources carry more weight than anything you say about yourself, because an AI can corroborate them.

AudFlo check: Independent third-party mentions present?

Tier 3Named customer stories with a measurable outcome

Proof from a real, named customer, with a result an AI can repeat back to a user.

AudFlo check: Named case studies present?

Tier 4Named quotes that vouch for you

A named person vouching for you helps verify trust, but it is still something you publish about yourself.

AudFlo check: Named testimonials present?

Tier 5Unverified claims you make about yourself

The weakest signal, because nothing here can be checked against an independent source.

AudFlo check: Unsupported claims detected?

Want to see this applied? The sample audit shows all four pillars scored on a real site, with the blockers found and the fixes generated.

View sample audit →
How the number reads

The four ranks

Your score maps to one of four ranks.

Recommended

80 and above

AI engines will reliably surface you when relevant queries come in. You are doing the work other founders skip.

Approaching recommendable

65 to 79

Close. A handful of targeted fixes (usually FAQ schema, comparison content, or testimonial markup) move you into Recommended territory.

Below recommendable

45 to 64

Real gaps. AI engines can find you but struggle to classify you or to confidently recommend you over competitors with stronger signals.

Invisible

below 45

AI engines either cannot reach your site or cannot understand what it offers. This is where most founders start. It is also where the biggest wins are.

What the score does not tell you

AudFlo measures structural and discoverability signals. It cannot measure brand reputation, product quality, network effects, or word of mouth. A great product with a poorly structured website still scores low. A mediocre product with a perfectly optimized website still scores high. The score reflects how discoverable your product is to AI engines, not how good your product is. Discoverability is necessary, not sufficient.

How we built this

We built AudFlo by studying how ChatGPT, Claude, Gemini, and Perplexity actually decide what to recommend when users ask category questions. Some signals match traditional SEO best practices. Many are new and specific to AI engines: FAQ schema, Person markup, the llms.txt specification, structured comparison content. The weights are based on observed AI engine behavior in 2026 and informed by published research on AI search ranking.

Methodology updates

AI engines change. The methodology updates when major shifts occur, for example, the llms.txt specification became part of the Technical Visibility pillar when major AI engines began reading it. Significant changes are versioned and announced.

Current version

1.0

Last updated

June 2026

FAQ

Common questions


How are the four pillars weighted?+
Technical Visibility is 20 percent of the score. Structural Understanding is 30 percent. Answer Selection is 30 percent. Authority Signals is 20 percent. The four pillars add up to 100 percent of the AI Visibility Score.
Where do the weights come from?+
The weights are based on observed AI engine behavior in 2026 and informed by published research on AI search ranking. We tested AudFlo's signals against actual ChatGPT, Claude, Gemini, and Perplexity recommendations across hundreds of categories before locking the weights.
Will the methodology change over time?+
Yes. AI engines evolve. When major shifts occur (for example, the introduction of the llms.txt standard), we update the methodology and version it. The current version is 1.0, last updated June 2026.
What is the difference between Recommended and Approaching recommendable?+
Recommended (80 plus) means AI engines reliably surface you for relevant queries. Approaching recommendable (65 to 79) means a few targeted fixes will move you into Recommended. The fixes are usually FAQ schema, comparison content, or testimonial markup.
Why is Authority Signals only 20 percent?+
Authority is harder to fix in a single Fix Pack cycle. We weight it lower so a small business with limited social proof is not unfairly penalized. The structural pillars (Technical and Structural Understanding) carry more weight because they are also more directly fixable.
Does the AI Visibility Score predict revenue?+
No. AudFlo measures discoverability, not greatness. A high score makes you findable and recommendable. Whether AI engine recommendations convert to revenue depends on your product, your category, and your conversion flow.

Ready to see where you stand?

Run a free scan and see your own score across all four pillars in about two minutes.

See your own score →