Answer engine scanner
What an answer enginecan quote from you
Something else reads your site now, and it is not a ranking index. It is a fetch that runs no JavaScript, a decision about whether to trust what came back, and an answer given to your customer without them ever seeing your domain. These eight checks ask whether you are legible to it.
- 8 AEO checks
- Engine 1.24.12
- Obeys robots.txt
- Abstains rather than guesses
What the scan reads
Six things
the bot checks
They answer three questions the page has to survive: can the bot reach it, is there a page once the scripts stop running, and can an engine tell who said it and when. All six are checked against the served document — never against a model.
Whether the crawler is let in
robots.txt is read for the AI user-agents it shuts out of the whole site — and the severity goes up when the same host publishes an llms.txt that invites those very bots in. One file saying come, the other saying no, is a mistake a site owner can actually fix in a minute.
Whether /llms.txt exists
The plain-text guide AI crawlers are increasingly given, written for a reader with no search result to fall back on. Its absence is reported as an opportunity rather than a violation, because most sites have not written one. HTML served at that path instead is a defect — the file is there and says something a crawler cannot use.
Whether anything arrives
A page that ships a nearly empty framework shell and paints itself in the browser is nothing at all to a fetch that runs no JavaScript. The scan compares what the server sent with what a render would need, and says plainly whether the words were in the document to begin with.
Whether it can be lifted
On pages over 400 words: section headings and paragraph boundaries are what an assistant uses to find a quotable passage. A wall of unbroken prose can rank on page one and still be impossible to excerpt — which is the difference between being found and being used.
Whether the questions are marked
This one fires only where three or more question headings or accordions already exist with no FAQPage schema around them. The questions are written, the answers are written, and the only thing missing is the markup that tells a machine this block is a question and its answer.
Whether it can be attributed
Organization or Person schema on the home page whose sameAs links tie the site to an entity assistants already recognise. A machine-readable author and publication date, checked on article pages only. And on articles over 600 words, whether the page cites anything outside itself — models weigh whether a claim can be traced before repeating it.
AI answer engines · 8 checks
All eight,
and what each one waits for
Most of this pillar stays quiet unless the page qualifies: an article check on a product listing is a false finding, and a schema check on a site with no questions written is noise. The conditions are part of the list below.
Getting in
2 checks
The two doors in front of the content: what robots.txt tells an automated reader, and whether the file written specifically for AI crawlers exists at all.
- AI crawler access
- llms.txt
Being readable
3 checks
Content that is in the HTML before a script runs, structure an assistant can excerpt a sentence from, and questions marked as questions.
- Server-rendered content
- Answer structure
- FAQ and HowTo schema
Being attributable
3 checks
Who is speaking, when they said it, and whether the claim can be traced to a source. Three signals a model weighs before it repeats your page to someone else.
- Entity schema
- Author and date
- Outbound citations
What each one looks at, and the cases where it deliberately says nothing, is on the full list of every check.
A citation you did not get leaves no trace in any dashboard you own.
Why this is its own pillar
The answer arrived
and the click did not
Security and SEO have always had a victim you can see: a breach, or a drop in traffic. This one does not. An assistant answers a customer’s question about your category, summarises three sites, names one of them, and sends the visitor there. The other two lose a customer they will never learn they had — no referral, no bounce, nothing in analytics that looks like a loss. That is why these checks sit apart from the SEO list rather than at the end of it.
The overlap with SEO is real and partial. Both read the same document, and structured data helps both. But they ask different questions. SEO asks whether a page can be indexed and ranked; this asks whether a passage can be extracted, trusted and attributed — which is why an unbroken 900-word page can pass every SEO check in the list and still be unusable here.
What this pillar will not pretend. No check can tell you whether you are inside a model’s training data, nobody can scan a competitor’s citation share, and this page will not sell you an “AI visibility score”. It does not query ChatGPT, Perplexity or Claude and it never claims to — a scan of your URL cannot see what an assistant decided. What it can see is the page their crawler fetches, and the eight properties of it that a site owner can actually change.
Check whether
you can be quoted
Paste a URL and all eight run against it, with the value behind every finding and a prompt that fixes it — plus the SEO and security pillars from the same fetch. Free account, no card.
Scan your site
ScanlyFix
Security, SEO and AI-answer-engine scanning for anything you put on the internet. Paste a URL, read the evidence, run the fix.
Home