What your number actually means

A measurement of whether AI answer engines can reach your website, read it, and resolve your brand as a specific entity. Every line of it was tested against your live site during the run — nothing is looked up from a database or inferred from a third-party crawl.

It is not a measure of whether ChatGPT, Gemini or Perplexity currently recommend you. That is a different question with a different answer, and it needs a set of buyer questions run repeatedly across those engines — because answer engines are not deterministic, and a single run cannot support a claim about movement. We keep the two numbers apart deliberately. A vendor who merges them into one number is quoting you noise.

Two separate things, because they disagree more often than people expect. First, what your robots.txt gives each AI crawler permission to do. Second, what your server actually does when 24 real AI crawler user-agents knock on the door — which is a different question, and the one that decides the outcome.

A crawler that cannot fetch your page cannot cite you, no matter how good the page is. Blocking OAI-SearchBot removes you from ChatGPT's answers. Blocking PerplexityBot removes you from Perplexity's. There is no partial credit and no workaround: this is binary.

A low score here usually means you are invisible to at least one major answer engine right now, today. In our own portfolio we found sites blocking AI crawlers by accident — a Cloudflare toggle switched on years ago by somebody who has since left.

robots.txt names every AI crawler you want and allows it explicitly, private paths stay disallowed, and the live probe returns HTTP 200 for every crawler that matters.

Whether your content exists in the HTML that is actually served — before any JavaScript runs — and whether it is shaped so an answer can be lifted out of it: real headings, a direct answer near the top, a title and description that say something, and enough words on the page to constitute a source.

AI crawlers do not run your JavaScript. Almost none of them do, and the ones that sometimes do will not wait. A React or Vue site that renders client-side serves an empty shell to every crawler, which is why a beautiful site can score badly here and a plain HTML page can score perfectly.

You get fetched and then skipped. The crawler came, found nothing quotable, and moved to whoever answered the question in the first paragraph.

Server-rendered or statically generated HTML, one clear H1, a 40–60 word direct answer near the top of every important page, descriptive headings, and a title and meta description written for a human.

Your JSON-LD: whether an Organization or LocalBusiness block exists and is valid, whether FAQ, Service, Product, Article and Breadcrumb types are used where they apply, and whether the entity carries the identifiers that let a model tell you apart from a similarly-named company.

Models resolve entities, not strings. Without structured data your brand name is text that has to be guessed at from context — and when another company shares your name or category, the guess frequently goes the other way. We found exactly that on our own domain: a near-identical brand name in the same country and category, and the model was quoting their facts about us.

You get half-credited, misattributed, or merged with somebody else. The answer names a company; it is not reliably yours.

Valid Organization or LocalBusiness JSON-LD on every page, sameAs links to your real profiles, a consistent legal name, FAQPage markup on pages that answer questions, and Breadcrumb markup for structure.

Sitemap presence and health, whether the sitemap is declared in robots.txt, canonical consistency, whether www and non-www agree, whether your 404 handling is honest, whether feeds exist, and whether IndexNow is set up so new pages get picked up in hours rather than weeks.

Answer engines do not crawl the open web from scratch on every question. They read indexes. Bing's index feeds ChatGPT Search and Microsoft Copilot; Google's feeds AI Overviews and Gemini grounding. Absence from an index is absence from everything built on top of it.

Slow discovery at best, permanent invisibility at worst. Most Indian sites we audit have never been submitted to Bing at all — six of our own properties had Bing Webmaster Tools accounts with zero sitemaps in them.

A clean sitemap declared in robots.txt, consistent canonicals, one host that redirects to the other, Bing Webmaster Tools verified, and IndexNow keys live.

Whether a real organisation is visible behind the site: a named entity, a physical address, contact routes, an about page with substance, authorship, and dates that show the content is maintained.

When an engine picks three sources out of a hundred candidates, it favours the ones that look accountable. An anonymous page with no address, no author and no dates loses to a page that has all three, even when the content is comparable.

You stay a source of last resort. You will be read, occasionally quoted, and rarely recommended by name.

A real about page, a postal address, a named team, working contact routes, visible publication and update dates, and consistent identity across your own profiles.