AI search visibility
When an assistant answers for your category, does your name come up?
People increasingly ask ChatGPT, Perplexity, Gemini and Copilot for a recommendation instead of opening ten tabs. Those answers are assembled from pages the models can actually reach and parse — and a site can be perfectly ranked on Google while being invisible to every one of them.
Sample report · 3 September 2026
northfield-outfitters.com
- Visibility
- 38
- Conversion
- 76
- Security
- 42
- Compliance
- 86
- Speed
- 62
- Usability & Design
- 84
Being crawled is not the same as being cited
Search rankings and answer-engine citations are decided by different machinery. An assistant has to fetch your page, parse it without running your JavaScript, work out what your business is and what it sells, and then decide you are worth naming. A site can pass every SEO audit and fail at the first of those steps.
Most sites block them without knowing
The common failure is not subtle wording, it is a robots.txt rule someone added years ago, a bot filter at the CDN, or a storefront that renders entirely in the browser and serves an empty shell to anything that does not execute JavaScript. Each one removes you from that engine's answers outright.
It is checkable, so we check it
This is not a prediction market. Whether GPTBot is allowed, whether your product copy survives with JavaScript off, whether your entity and Q&A markup parses — all of it has a definite answer today. We fetch, parse and ask, then tell you which of those steps you are failing.
Named checks
What the visibility scan actually does
Six probes, in the order an assistant would hit them. Each one is a pass, a fail or a not-applicable in your report, with the evidence attached.
- robots.txt
- Fetched and parsed, not assumed. A single stale Disallow is the most common reason a site is absent from answers.
- 12 named AI crawlers
- GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot and the rest, each checked by name rather than by a blanket rule.
- Raw HTML, JavaScript off
- We read the document the way a retrieval bot does. If your copy only exists after hydration, it does not exist to them.
- Entity & Q&A schema
- Organization, Product and FAQPage markup — what tells a model what you are, rather than leaving it to infer.
- Title, canonical, sitemap
- The plumbing that decides which of your URLs an engine treats as the real one, and whether it finds the rest of them.
- Live model questions
- We put real questions to real models about your category and record whether you are named, and alongside whom.
Questions
Before you run it
Is this the same as SEO?
No, though they overlap. SEO optimises for a ranked list of links. This checks whether an assistant can retrieve and parse your page at all, and whether it has enough structured information to name you inside a written answer. Sites that rank well and are never cited are common enough that we built the check.
Do I need to let AI crawlers in?
That is your decision and there are honest reasons to say no. What we do is make the consequence explicit: blocking a retrieval bot removes you from that engine's answers, and most sites blocking them today did not choose to. Once you can see which are blocked, the choice becomes a real one.
Does it work on a JavaScript-heavy store?
Yes, and that is exactly where it earns its keep. We fetch the raw document with no JavaScript executed, which is what most retrieval bots see. A React or Vue storefront that looks fine in your browser and returns an empty shell to a crawler is the single most expensive failure this scan finds.
How long does it take?
About two minutes. You give us a URL, no account and no card, and the full report arrives as a link you can share with whoever maintains the site.
Run it on your own site.
Your score, and every finding ranked by what it costs you. If it comes back clean, that is a useful thing to know too.