AI Visibility Audit — ChatGPT, Claude, Perplexity
Your site may already be invisible to AI — and you would have no way of knowing.
In 2026, generative engines answer billions of questions by retrieving and citing web pages in real time. Whether your page gets cited depends on two things: can the engine reach it, and can it use what it finds.
Most site owners have never checked either.
The trap nobody sees
robots.txt controls which crawlers can fetch your pages. The problem: blocking a training crawler does not block the retrieval crawler from the same vendor.
- Blocking
ClaudeBotdoes not blockClaude-SearchBot - Blocking
GPTBotdoes not blockOAI-SearchBot - Each user-agent token needs its own directive
Hosting defaults, CDN bot rules, and security plugins routinely block retrieval crawlers without the site owner ever noticing. The result: the page is invisible to AI answers, and nothing in the site's analytics shows it.
This agent checks all six retrieval crawlers individually.
What it does
You give it a URL. It returns two things:
1. An engine access table — whether each retrieval crawler is permitted to fetch the page:
OAI-SearchBot/ChatGPT-User(OpenAI)Claude-SearchBot/Claude-User(Anthropic)PerplexityBot/Perplexity-User(Perplexity)
2. A citability assessment — whether what the crawler finds is structured for extraction:
- Structured data — JSON-LD with citation-relevant types (Article, FAQPage, Product, Organization)
- Author attribution — who wrote this, who publishes it
- Date signals — datePublished, dateModified
- Heading structure — H1/H2 hierarchy engines use to locate answers
- Meta description — the ready-made summary engines consume
- Server-rendered content — whether the text is in the HTML or only appears after JavaScript
- Canonical URL — whether the authoritative version is declared
What you get
A downloadable report with a visibility risk score, the engine access table, and findings ranked by impact. Each finding includes what is missing, why it matters for AI citation, and the exact fix.
Training crawlers you are blocking are listed separately and never counted as a defect — opting out of training is a legitimate rights decision, and this tool respects it.
Who this is for
- SaaS founders who want their product pages cited when someone asks "best tool for X"
- Content teams who write authoritative material and want it to appear in AI answers
- SEO professionals adding GEO to their practice
- Agencies auditing client sites for AI readiness
What this is NOT
This tool does not query the AI engines. It cannot tell you whether ChatGPT currently cites you for a specific question — no tool can, since generated answers vary between runs.
What it establishes is whether the engines can reach your page and whether what they find is structured for extraction. Those are necessary conditions, not guarantees.
On llms.txt
The scanner reports whether /llms.txt exists at your domain root. As of 2026, there is no clear evidence it materially affects AI retrieval. The report notes its presence without overselling it.
Built by ZALTEN — diagnostic agents that find every gap before it finds you.
