AI Visibility

AI Visibility answers a question that didn’t exist a few years ago: can ChatGPT, Claude, Perplexity, and Google’s AI actually see, read, and cite your website? It runs your site through a per-bot access matrix, tests what your pages look like to a crawler that doesn’t run JavaScript, checks for an llms.txt, and rolls it all into an AI Readiness Score with a prioritised list of fixes. It doesn’t moralise about AI: it measures your visibility to it and explains the trade-offs honestly.

The search-to-AI shift

A growing share of the questions people used to type into Google now go to an AI assistant instead. When someone asks ChatGPT for a recommendation, or Perplexity for a comparison, or clicks through Google’s AI overview, the “search results” are an answer, and being in that answer is a new discipline sitting alongside classic SEO. This is generative engine optimization (GEO): making your site easy for AI systems to crawl, understand, and quote. It overlaps SEO but adds three things SEO never had to worry about: whether you allow the AI crawlers, whether your content is extractable without a browser, and whether your identity is stated in a machine-readable way.

How AI crawlers differ from Googlebot

Googlebot is patient and sophisticated: it renders pages, runs JavaScript, and waits for content to appear. Most AI crawlers are not. They typically fetch your raw HTML and read what’s there, no rendering, no waiting. That single difference is behind most “why can’t AI see my site?” surprises, and it’s why this tool tests your pages the way those crawlers actually experience them rather than the way your browser does.

Training, fetch, and search: three kinds of bot

The most important thing to understand about AI crawlers is that they are not one category, and blocking them means very different things. AI Visibility groups them by purpose and scores them accordingly:

• Training bots (GPTBot, ClaudeBot, Google-Extended, CCBot…) gather content to train future models. Blocking them is a legitimate content-rights decision, so on this tool, blocking a training bot does not lower your score. It’s shown in the matrix as a choice, neutrally, not a failure.
• Assistant-fetch bots (ChatGPT-User, Claude-User, Perplexity-User…) fetch a page in real time because a user just asked about it. Block these and AI tools can’t open or cite you at the exact moment a customer is looking.
• AI-search bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot…) plus the classic search crawlers Googlebot and Bingbot build the indexes behind AI answers and overviews. Block these and you’re absent from AI search entirely.

The score measures visibility to AI users: the fetch and search bots plus your content’s readability, not your stance on training. That distinction is deliberate: this tool informs and lets you decide, rather than fear-mongering or preaching.

llms.txt, briefly

llms.txt is a proposed convention: a markdown file at /llms.txt that summarises your site and points to its key pages for AI systems: a sitemap written for language models rather than crawlers. Adoption is early and it’s optional, but it costs about ten minutes and gives assistants a clean, curated map of what you’re about. AI Visibility checks whether yours exists, returns real content (not an HTML error page), starts with a proper heading, and links to your important pages, and if it’s missing, it hands you a starting point.

SSR vs. CSR: the invisible-shell problem

Here’s the failure that catches the most modern sites. If your page is a client-side-rendered (CSR) single-page app (React, Vue, and friends) the HTML that ships is often an almost-empty shell, and the real content only appears after JavaScript runs in a browser. Human visitors never notice. But an AI crawler that reads raw HTML sees the empty shell: no headline, no copy, nothing to cite. The “What AI sees” panel on this tool shows you that exact view: your page as a non-JavaScript crawler receives it. If it comes back thin, that’s your finding. The fix is server-side rendering (SSR) or prerendering, so the meaningful content is in the HTML before any JavaScript runs. Structured data helps here too: the same JSON-LD that earns a Google rich result gives an AI an unambiguous statement of who you are and what you offer.

For AI agents

This tool exposes a plain-text endpoint designed for language models and agents:

GET https://nitrosites.com/tools/ai-visibility/scan.php?url=<target>&format=agent

It returns a token-cheap digest of the same analysis shown on this page. Soft rate limits apply; results are cached briefly. No key or login required. (Yes, the AI-visibility tool has an agent-friendly endpoint. The recursion is the point.)

Frequently asked questions

What is GEO (generative engine optimization)?

Making your site easy for AI systems (ChatGPT, Claude, Perplexity, Google's AI results) to crawl, understand and cite, so you're in the answer when customers ask AI instead of searching. It overlaps SEO but adds AI-crawler access, extractable content and machine-readable identity.

Should I block AI crawlers?

Split the question in two. Training bots (like GPTBot) collect data for future models: blocking them is a legitimate content-rights choice and doesn't affect your score here. Assistant and AI-search bots (like ChatGPT-User or PerplexityBot) fetch pages to answer a user right now: block those and AI tools can't recommend or cite you.

What is llms.txt?

A proposed convention: a markdown file at /llms.txt summarising your site and its key pages for AI systems, like a sitemap written for language models. Adoption is early, it costs ten minutes, and this checker validates yours if present.

Why can't AI see my JavaScript site?

Most AI crawlers read your raw HTML and don't execute JavaScript. If your content only appears after a framework renders it client-side, AI systems see an empty shell: the "What AI sees" panel on this tool shows you exactly that view. Server-side rendering or prerendering fixes it.

Does schema markup help with AI visibility?

Yes, structured data is one of the clearest machine-readable signals of who you are, what you sell and what your content answers. It's a shared win: the same JSON-LD serves Google rich results and AI comprehension.

How is the AI Readiness Score calculated?

Six weighted pillars: AI crawler access (25), content extractability (25), machine-readable identity (20), metadata (10), llms.txt (10) and crawl infrastructure (10), each with itemised checks and evidence shown in the report.