Your AI Visibility score is a number between 0 and 100. It tells you how well an AI system (ChatGPT, Perplexity, Google AI Overviews, a browsing agent) can find, parse, trust, and cite your website.
It's a separate score from your SEO Health score. The two overlap, but they're not the same thing. A site can rank well in traditional search while being effectively invisible to AI. A site can have structured data and still fail entity disambiguation. The AI Visibility score isolates the signals that specifically matter to AI-assisted discovery, so you know where to focus.
This article explains how we calculate it, check by check.
How does the score work?
The score is weighted. Not every check carries the same importance, because not every failure has the same impact on whether an AI will find, trust, and cite you.
Each check is assigned a weight from 1 to 5. That weight is multiplied by 10 to give a maximum point value per check. A passing check earns all its points. A warning earns 60% (with some checks getting softer or harder treatment depending on the nature of the signal). A failing check earns zero.
Your final score is calculated as:
earned points / total possible points × 100Checks that aren't present in a scan result (because the check doesn't apply, or the data isn't available) don't count toward the total. You're not penalised for things we didn't measure.
As of June 2026, the score draws from eleven checks, grouped across four categories:
Structured Data
AI systems don't read your page the way a human does. They parse structured signals (JSON-LD, schema graphs, named entities) to decide whether your content is machine-readable, whether it can be attributed, and whether it deserves to be cited. Structured data is where that process starts.
What does Schema Validation check, and why does it matter for AI? Weight 4 / 5
The most heavily weighted check in the structured data group, and the most consequential for AI readiness overall.
We parse every JSON-LD block on your page and check for structural errors (missing required properties, invalid nesting, type mismatches), missing recommended properties that reduce richness, and duplicate schema types. Duplicate types usually mean a third-party script is injecting its own schema on top of yours.
Beyond pass/fail, we calculate an ai_ready signal for schema. A schema can be technically valid but still not AI-ready. To earn the full score here, your schema needs to include signals that help AI systems identify your content as a trustworthy, citable source: FAQPage or QAPage schema for question-and-answer content, speakable markup for voice and AI summary extraction, and sameAs authority links on your primary entity blocks.
A passing schema without the ai_ready signal scores at 80% rather than 100%. The remaining 20% reflects the difference between schema that won't break anything and schema that actively signals quality to AI systems.
What is Schema Graph Linking and does it affect AI visibility? Weight 2 / 5
Schema blocks can reference each other using @id values to form a connected graph. A WebPage can declare its publisher. A BlogPosting can reference its author. An Organization can link to its WebSite.
When these connections are present, AI systems get a clearer picture of the relationships between entities on your site. When they're missing, each block is read as a standalone, unconnected declaration, which is less useful for entity resolution.
This check looks for cross-references between your schema blocks. It's a refinement signal rather than a foundational one, which is why the weight is low, but it does contribute meaningfully to entity graph quality.
Does your page title need to match your schema name? Weight 2 / 5
This check compares your <title> tag to the name property in any WebPage schema on the same page. If they're inconsistent, it introduces ambiguity: the page says one thing in HTML and another in structured data. AI systems correlate these signals to verify page identity, and a mismatch is a minor trust failure.
Low stakes on its own, but an easy win. It's also the kind of inconsistency that often creeps in when schema is generated separately from the page template.
Crawler Access
Before any AI system can process your content, it has to be able to reach it. This group assesses whether you're inadvertently blocking AI crawlers, and whether you've provided the explicit access signals that AI systems are starting to look for.
Is your site blocking AI crawlers in robots.txt?
Weight 5 / 5
The highest-weighted check in the entire AI Visibility score.
We fetch your robots.txt and check whether any of the major AI crawler user agents are explicitly blocked. The crawlers we check include:
- GPTBot (OpenAI)
- ClaudeBot (Anthropic)
- PerplexityBot
- Google-Extended
- Applebot-Extended
- Bytespider
A block here is absolute. If GPTBot is disallowed, OpenAI's systems cannot index your content, and your AI Visibility score will reflect that as a zero for this check regardless of everything else you've done.
We also flag Crawl-delay directives targeting AI crawlers, which don't block access but do slow it down. That's an informational flag rather than a failure.
What is llms.txt and does your site need one?
Weight 3 / 5
llms.txt is an emerging standard. It's a plain-text file at your domain root that tells AI systems what your site is about, which pages are most important, and how your content is licensed for AI use.
It's not yet universally supported, so we don't penalise hard for its absence. A missing llms.txt is an informational flag rather than a failure. It's an easy opportunity most sites haven't taken yet.
When the file exists, we check the quality. A bare-bones llms.txt with just a title and a link list earns partial credit. A well-structured file with a description, meaningful section organisation, and an llms-full.txt companion earns the full score.
The ai_ready signal maps to a qualitative assessment of whether the file actually communicates useful intent to an AI system, or whether it's just there for the sake of it.
Content & Semantics
This group assesses whether AI systems can actually read and use your content once they've accessed it. Schema tells AI what your content claims to be. Content and semantics determine whether the raw page structure backs that up.
What is Content Extractability and why does it matter? Weight 3 / 5
An AI system crawling your page needs to extract meaningful prose. This check evaluates the structural characteristics that affect whether that extraction is clean and reliable.
We look at three things:
- Heading structure — H1 through H3 used correctly, with a logical hierarchy.
- Paragraph density — whether content is in proper
<p>tags rather than raw text nodes or<div>soup. - Q&A patterns — whether FAQ-style content uses structurally appropriate markup that maps to
FAQPageschema, if present.
A page that scores well here is one where an AI agent can reliably identify the main content, extract it in a coherent sequence, and attribute it to the right page. A page that scores poorly is typically one where content is buried in custom components, where heading hierarchy is broken, or where text lives in markup that's hard to parse.
How do HTML landmark elements affect AI content parsing? Weight 2 / 5
This check assesses whether your page uses the HTML5 landmark elements correctly and in the right places: <main>, <nav>, <header>, <footer>, <article>, <section>, <aside>.
These aren't just accessibility features. They're content zone signals. An AI agent parsing your page uses landmark structure to determine where the primary content is, where the navigation is, and what to ignore. A page with no <main> element, or with a <main> that contains navigation, provides a less reliable extraction surface.
We also check for correct use of <time> elements with machine-readable datetime attributes (important for publication dates and event schemas), and <figure> / <figcaption> for image attribution.
What is Interactive Semantics, and how does it affect AI agents? Weight 2 / 5
The newest addition to the AI Visibility score, and the one that reflects where things are heading most directly.
AI agents don't just read pages. Increasingly they interact with them: form submission, navigation, modal dialogs, disclosure panels. For an agent to interact correctly with your UI, it needs ARIA attributes and semantic roles to be implemented accurately.
We run nine sub-checks within this category, each targeting a distinct class of interactive element:
| Sub-check | What we look for | Severity |
|---|---|---|
| Form field labels | <input> and <textarea> associated with a <label> or aria-labelledby |
High |
| Button semantics | <a> tags misused as buttons — breaks agent interaction models |
High |
| Nav region labels | Multiple <nav> elements distinguished with aria-label |
Medium |
| Icon-only links | Links containing only an icon or SVG have an accessible name via aria-label |
Medium |
| Disclosure state | Expand/collapse toggles declare state with aria-expanded |
Low |
| Dialog roles | Overlay elements use role="dialog" and aria-modal |
Low |
| Live regions | Dynamic content areas that update without page load use aria-live |
Low |
| Article labelling | <article> elements have an aria-labelledby reference |
Low |
| Form labelling | <form> elements have accessible names |
Medium |
Each sub-check surfaces as a separate action item in the dashboard because each one is a distinct fix. Form field labels and button semantics are weighted higher because they're the most common interaction points for AI agents.
Why are some checks excluded from the AI Visibility score?
Several checks that appear in your SEO Health score don't contribute to AI Visibility: HTTPS redirect, mixed content, broken links. These are foundational technical health checks and they matter, but they're not AI-specific signals. We keep the AI Visibility score focused on the checks where the question is specifically "can an AI system access, understand, and trust this content?" rather than "is this site technically healthy?"
How do I improve my AI Visibility score?
Every check in the AI Visibility score is actionable. The Top Priorities list in your dashboard surfaces each one as a plain-English action item, grouped by impact level, with a detail panel that explains the specific issue and how to fix it.
If you want to see how your site currently scores, run a free scan. No account required.
Questions about how a specific check is evaluated, or how to interpret a result for a client site? Get in touch and we'll take a look.
By Tom Freeman · Co-Founder & Lead Developer
Full-stack developer specialising in high-performance web applications and automated monitoring.
Do Open Graph tags affect AI visibility? Weight 1 / 5
The lowest-weighted check in the AI Visibility score. Open Graph and Twitter Card tags don't directly affect AI crawler access, but they do affect how your content is previewed and attributed in AI-generated link summaries, and they're a lightweight signal of content metadata quality.
Missing social tags are a warning rather than a failure. Easy to add, worth having, but not going to move your AI Visibility score materially on its own.