See your site the way AI bots see it.
Paste a URL and we check whether ChatGPT, Perplexity and Google's AI crawlers can reach the page, then score what they'd actually find once they do. Free, no signup, results in seconds.
Five things that decide whether AI can cite you
Each dimension below is scored independently, so a bad title doesn't hide a robots.txt problem and vice versa — the report tells you exactly which one to fix first. If you want the mechanism behind each check, the full LLM SEO guide covers each dimension in depth.
AI Crawler Access
30% of the scoreThis checks your robots.txt against the two bot tiers that actually decide whether you show up: search-tier crawlers (OAI-SearchBot, PerplexityBot, Googlebot, Bingbot, Applebot, Claude-SearchBot, DuckAssistBot) that feed AI search results, and assistant-tier fetchers (ChatGPT-User, Claude-User, Perplexity-User, meta-externalfetcher, MistralAI-User) that run when someone pastes your URL into a chat. Block either and the page is invisible to that surface no matter how good the content is — access is binary, content quality never gets a vote. Most sites that fail here didn't decide to block AI; a "Disallow: /*?" rule written for a different bot years ago catches these too. Open robots.txt and check each bot by name, not just the wildcard group. Training-only crawlers (GPTBot, ClaudeBot, Google-Extended, CCBot) are scored separately as a policy choice, not a fault — blocking them doesn't hurt this number.
Machine Readability
20% of the scoreWhether the page declares a language, a charset, a viewport, and exactly one H1. These aren't accessibility checkboxes — they're the signals a parser uses to segment a page before an LLM ever sees the words. Two H1s (or none) breaks the outline a summarizer relies on to tell "this is the actual topic" from "this is a sidebar," and a missing lang attribute can get a page misclassified for a language query it should have won. The fix is mechanical: one lang attribute on <html>, one meta charset declared early in <head>, one viewport meta, and exactly one H1 per page — if you have two, one of them should be an H2.
Structured Data
15% of the scoreWhether the page ships any JSON-LD at all, and what types. This is the most direct way to tell an AI system what a page is without making it guess from prose — Article, Product, FAQPage, Organization — and answer engines lean on schema to pull out entities and facts rather than re-parsing your paragraphs every time. Pages with zero JSON-LD score zero on this dimension regardless of how well-written the content is. Start with Organization and WebPage site-wide, then add the type that matches the page: Article for posts, Product for listings, FAQPage for anything with a Q&A section.
Content & Citability
20% of the scoreTitle length (15–60 characters), description length (50–160), body word count (300+), and whether the page has at least two headings. A summarizer looking for a citable passage skips pages that are too thin to have one, and a page with no heading outline gives it nothing to quote with attribution — it either paraphrases loosely or moves on to a competitor's page that made the quote easy to lift. If a page is under 300 words, that's usually the real problem, not the title or description. Write both within range, then make sure the page has at least two headings that frame distinct, quotable sections rather than one wall of text.
Trust Signals
15% of the scoreCanonical URL, Open Graph tags (at least two), and a Twitter Card. A missing canonical means an AI crawler that lands on a duplicate or parameterized URL has no signal pointing back to the real one, so it may cite or index the wrong copy. OG and Twitter metadata are what render when an AI assistant surfaces a card or preview for your link — without them it either shows nothing or falls back to whatever it can scrape, which usually looks broken. Add a self-referencing <link rel="canonical">, og:title, og:description and og:image, and a twitter:card meta tag.
JavaScript-rendered site? The server can only see so much.
This checker fetches your page the way a crawler does — no browser, no JavaScript execution. If your content only appears after client-side rendering, the four content dimensions can't be measured from here (we say so instead of guessing a score). The Tracetify Chrome extension runs the same scoring logic inside the rendered page, so it sees what a browser-based AI agent sees.
Questions, answered
Anything else — support@tracetify.com
Is the checker free?
Yes. No account, no email, no credit card. The limit is 20 checks per hour per IP address, and it exists to stop scripted scanners, not to push you toward a paywall.
What do you fetch, and do you store my page?
Two requests: your robots.txt and the single page you pasted, fetched as TracetifyBot with an 8-second timeout and a 2 MB cap. The HTML is parsed for the score and discarded. We keep the computed result in a cache for one hour, so checking the same site twice returns instantly.
Why does my site say “Needs browser rendering”?
Your page builds its content with JavaScript in the browser, and we fetch it like a crawler that does not run scripts. Scoring the empty shell would give you an F you do not deserve, so we score crawler access only and skip the four content dimensions. The Tracetify Chrome extension runs the same checks inside the rendered page and can score all five.
Does blocking GPTBot lower my score?
No. GPTBot, ClaudeBot, CCBot and the other training crawlers are reported but never scored. Whether to feed model training is your call, and we do not grade policy choices. The score only moves on search and assistant crawlers, the ones that decide whether AI products can cite you.
Why does my score differ from other SEO tools?
Different question. Most SEO graders measure ranking factors like page speed, backlinks and keyword usage. This checker measures whether an AI system can reach your page and read it once it gets there. A page can score 90 on a classic SEO audit and still block every AI crawler it has.
This free check covers crawler access on your own pages. The SEO agent goes further — how often your competitors get cited in ChatGPT and Google AI answers, and what to do about it.
ask the agent →