How the AI Visibility Score works
Surfaced AI is an AI Visibility Audit Platform for founders. The AI Visibility Score is how it measures whether AI engines can understand and recommend you: a number from 0 to 100, built from four weighted pillars and 32 individual checks.
0–100
Single score
4
Weighted pillars
32
Checks
Executive summary
The short version
The AI Visibility Score reflects how easily AI engines like ChatGPT, Claude, Gemini, and Perplexity can access, classify, and recommend your product. It's built from 32 checks grouped into four weighted pillars that add up to 100%: Technical Visibility (20%) asks whether AI engines can reach and read your site at all. Structural Understanding (30%) asks whether they can tell what your product is and who it's for. Answer Selection (30%) asks whether your content is shaped the way AI engines assemble answers. Authority Signals (20%) asks whether your site gives them a reason to trust what you claim. A higher score means a higher chance of being understood — and recommended — when it counts.
What we measure
The four pillars
Each pillar measures a different dimension of AI-engine readability. They're weighted because not every signal matters equally — together they make up 100% of the AI Visibility Score.
i.
Technical Visibility
What it measures
Whether AI crawlers can reach, load, and parse your site at all: HTTPS, robots.txt access, a sitemap, structured data, an llms.txt file, title and meta tags, and a mobile viewport tag.
Why it matters
If a crawler is blocked or the page fails to parse, nothing else on this list matters — a site scores zero on every other pillar if it was never readable in the first place.
How it's scored
Eight pass/fail checks. Each one is worth an equal share of this pillar's 20 points.
What fixes improve it
Adding an llms.txt file, confirming robots.txt allows crawling, adding JSON-LD schema, and setting a proper meta description are the four fastest wins here.
ii.
Structural Understanding
What it measures
Whether your homepage copy tells a machine reader what you are and who you are for: a descriptive H1, a real meta description, enough body copy to classify from, a clear category statement, and plain, low-jargon language.
Why it matters
An AI engine has to classify your product in a sentence or two before it will ever recommend you. Vague or jargon-heavy homepages give it nothing to classify.
How it's scored
Eight pass/fail checks, together worth 30% of the total — the single largest pillar.
What fixes improve it
Rewriting your H1 to name your category and audience in one line, and adding a plain-English sentence describing what you do, are the highest-leverage changes on this pillar.
iii.
Answer Selection
What it measures
Whether your content is shaped the way AI engines assemble answers: FAQ content and FAQ schema, comparison content, use-case coverage, question-style headings, topic breadth, and internal linking.
Why it matters
AI engines favor sites that already answer the question being asked, in a format they can lift directly. A site with no FAQ and no comparison content gives the engine nothing to quote.
How it's scored
Eight pass/fail checks, tied with Structural Understanding for the largest share at 30%.
What fixes improve it
Publishing an FAQ block with FAQPage schema, and a plain comparison page naming your alternatives, are usually the fastest way to move this pillar.
iv.
Authority Signals
What it measures
Whether your site gives an AI engine a reason to trust you: testimonials, customer logos, a founder or about page, case studies, review markup, and links to real external profiles.
Why it matters
Between two similarly well-structured sites, AI engines lean toward the one with visible, checkable proof. Unverified claims about yourself carry the least weight.
How it's scored
Eight pass/fail checks, worth 20% of the total — weighted lower than the two content pillars because trust signals take longer to build than a copy edit.
What fixes improve it
Adding one named customer testimonial and linking a real founder or about page are the two fastest wins on this pillar.
Try the weighting
See how the weights shape a score
Drag the sliders to reweight the pillars and watch the same set of results produce a different score.
Example site · checks passed
Weights are normalised to 100% total, so the score stays comparable.
Resulting score
55
out of 100
Shipped weighting
Why these weights
Drag a slider to see how much the weighting matters. Structural Understanding and Answer Selection carry the most because an engine has to understand and be able to quote you before trust signals change anything.
Inside Authority Signals
How we weigh proof
Not all proof is equal. Within Authority Signals, we treat stronger, harder-to-fake evidence as more valuable than a generic claim — the same tiering applies however the proof was produced, hand-written or generated.
Independent, verifiable data
Original research or numbers only you have — the hardest signal to fake and the easiest for an AI engine to cite with confidence.
Third-party coverage you don't control
Independent mentions, reviews, or press you didn't write yourself.
Named customer case studies
A real customer, a real outcome, attributable and checkable.
Named testimonials
A named person vouching for you — still something you publish yourself.
Unverified claims about yourself
"Loved by thousands" with nothing behind it. The weakest signal we check for.
How the number reads
The four ranks
Your score maps to one of four bands.
Strong
80 and above
Your site gives AI engines what they need to understand, classify, and cite you with confidence.
Moderate
65–79
Close. A handful of targeted fixes should move you into Strong territory.
Developing
45–64
Real gaps remain. AI engines can probably find you, but struggle to classify or trust you.
Weak
Below 45
AI engines likely can't tell what you offer yet. Most sites start here before their first fix.
What the score doesn't tell you
The AI Visibility Score measures structural and discoverability signals. It can't measure brand reputation, product quality, or word of mouth. A great product with a poorly structured site can still score low; a mediocre product with a well-structured site can still score high. Discoverability is necessary, not sufficient.
How we built this
We built the checklist by studying what ChatGPT, Claude, Gemini, and Perplexity actually need to see before recommending a product, then translated that into checks a script can run automatically — some borrowed from established SEO practice, others specific to how AI engines read a page (FAQ schema, llms.txt, question-shaped headings).
Methodology updates
AI engines change, and so will this methodology. Material changes to pillar weights or checks will be versioned and announced here.
FAQ
Common questions
Technical Visibility and Authority Signals are each worth 20%. Structural Understanding and Answer Selection are each worth 30%, because clear positioning and answer-shaped content matter most to whether an AI engine recommends you.
Every check in the current methodology is binary — it either passes or it doesn't. We chose binary checks over partial credit so the score stays simple to reproduce and explain.
Yes. As AI engines change what they reward — for example, wider adoption of llms.txt or FAQ schema — we expect to add, remove, or reweight checks. Material changes will be versioned on this page.
Moderate (65–79) means most of the foundational signals are in place and a few targeted fixes would close the gap. Developing (45–64) means there are still real structural or content gaps beyond quick fixes.
Trust signals matter, but they take longer to build than a copy or schema fix, and they matter less if an AI engine cannot yet classify what you do in the first place. We weight the pillars that unblock basic understanding more heavily.
No. It measures discoverability signals only — how legible your site is to AI engines today. A great product with a poorly structured site can still score low; a mediocre product with a well-structured site can still score high. Discoverability is necessary, not sufficient.