How we score AI visibility.

The free scan reads your site the way a machine does.

The free score is a technical audit, done in about a minute. We fetch your homepage and check what an AI assistant’s crawler would find: your SEO basics (title, meta description, H1, canonical, language), your structured data (JSON‑LD — does the site say, in machine-readable form, that you’re an organisation or local business), security headers, and your robots and sitemap setup.

Crucially, we also check whether you’re blocking the AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google‑Extended — often left switched off by accident. The free scan doesn’t ask any assistant anything; it measures whether there’s anything for an assistant to find, quote and cite when it does come looking.

The ongoing measurement asks the questions your customers ask.

For clients on a plan, we go further — continuously. We put real, buying-intent questions to ChatGPT, Claude, Gemini and Perplexity — the kind a prospective customer would type — and read the verbatim answer that comes back: are you named, are you recommended, is a competitor named instead, and where does the assistant get its information.

No single assistant speaks for the whole market, so we combine what they say into one 0–100 score with a ± range, because language models are probabilistic — ask the same question twice and the wording can differ. Rather than hide that behind a single tidy number, we surface it. Every score traces back to answers you can read for yourself.

We show what actually moves the score.

Three things reliably raise how assistants answer for you. We measure each one and tell you where you stand on it — no jargon, no black box.

• Being named

  The assistant mentions your brand by name in its answer — not a competitor, and not a generic category.

• Being cited

  It draws on your own pages as a source and links or refers back to them when it answers.

• Structured content

  Clear, well-structured pages that state who you are, what you do, and for whom — so a model can quote them cleanly.

We propose fixes — then re-measure to prove it.

A score on its own changes nothing. So we turn it into a short list of concrete pages and fixes, ranked by the difference we expect them to make. You review every one: nothing goes live without your approval.

Once approved changes publish, we run the same measurement again and re-score. You see the movement against the same baseline — so improvement is something we demonstrate, not something we assert.