GEO Scanner Experimental
SEO gets you ranked. AEO gets you read. GEO gets you cited. This scans a page for the signals that make AI answer engines quote it as a source.
Why there's no 0–100 score. A score would need a corpus of pages known to be cited by AI answers to calibrate against, and no such public benchmark exists yet. A number would imply a precision we can't back, so this reports what's observably present instead. Each check below is labelled with what it actually rests on. How to read the evidence labels
What the evidence labels mean
- Verifiable Directly observable fact about the page, or documented crawler behaviour.
- Studied Measured to raise citation rate in published research.
- Documented Follows a published spec or platform guidance; effect on citation rate not directly measured.
- Inferred Reasoned from how retrieval and answer synthesis work. Plausible, not measured.
GEO Checks
What GEO is, and why it's not AEO
Readable vs. quotable
AEO asks whether an answer engine can parse your page — schema, headings, crawler access. GEO asks the harder question: when the engine writes its answer, does it cite you? Being parseable gets you into the candidate set. Being quotable gets you into the answer.
What actually moves the needle
The KDD 2024 GEO study tested content edits against a live generative engine. The changes that raised a source's visibility most were adding statistics, adding quotations, and citing authoritative sources — not keyword tactics.
Statistics beat adjectives
A model writing an answer needs facts it can state and attribute. "Significantly faster" gives it nothing. "38% faster, measured across 1,200 pages" gives it a sentence it can lift, with your page as the citation.
Chunking decides what gets retrieved
Retrieval splits your page into chunks and pulls one at a time. A 1,500-word section under one heading chunks badly — the retrieved fragment has lost the context that made it meaningful. Keep sections to roughly 150–400 words.
JavaScript is the silent killer
GPTBot, ClaudeBot, PerplexityBot and CCBot fetch raw HTML and generally don't execute JavaScript. A client-rendered page can be excellent and still be an empty document to them. This is the one failure that voids every other signal.
Name yourself
An engine cites a source it can name. Declare an Organization or Person in JSON-LD with sameAs links to Wikipedia, Wikidata or LinkedIn, so your entity resolves to a known one instead of a guess.