Get cited, not just ranked.
When ChatGPT or Google’s AI answers, nobody clicks. They read who got quoted. We make you quotable.
Four questions, then the fix.
Can the engine reach you? Can it read you? Does it know who you are? Is anything worth quoting? Then we fix whatever said no.
The method, step by step
Can the engine reach you
Twelve named crawlers, each checked against your robots.txt individually rather than assumed from a wildcard rule. Blocking them can be the right call. Blocking them by accident is the common one.
Can it read what it fetched
A page whose body arrives empty and fills in with JavaScript is invisible to any engine that does not run scripts. Measured per page, reported as a finding with the word count the server actually sent.
Can it resolve who you are
Organization or LocalBusiness markup, sameAs links to the profiles that describe the same entity, consistent naming, and a presence in the records models use to disambiguate companies.
Is any passage worth quoting
A direct answer in the first paragraph, a specific number in the first sentence, question headings with short answers under them, tables and lists, a named author, and outbound citations.
Then fix it
The answer block is drafted. The FAQ markup is generated from the questions already on the page. The llms.txt is built from the pages worth pointing at. The entity markup is written with your real profiles in it.
The 12 AI crawlers we check on every run
Each one is tested against your robots.txt with the same longest-match rule a real crawler uses.
| User agent | What it feeds |
|---|---|
| GPTBot | Trains and grounds ChatGPT answers |
| OAI-SearchBot | Powers ChatGPT search results and citations |
| ChatGPT-User | Fetches a page when a user asks ChatGPT about it |
| ClaudeBot | Claude's crawler |
| Claude-User | Fetches a page when a user asks Claude about it |
| PerplexityBot | Perplexity's index and its citations |
| Google-Extended | Gemini grounding and AI Overviews training |
| Applebot-Extended | Apple Intelligence and Siri answers |
| Bytespider | TikTok search |
| meta-externalagent | Meta AI answers |
| CCBot | Common Crawl, which many models train on |
| Amazonbot | Alexa and Rufus answers |
Not advice. The actual files.
llms.txt, written from your site
Grouped by section, every entry with a one-line summary taken from the page itself. Curated down to the pages you would actually want quoted, not a dump of every URL.
robots.txt with the engines you choose
Your existing rules kept, an explicit allow block added for the answer engines you want citations from, and the sitemap line added if it was missing. Marked critical risk, so a person reads it.
Entity and FAQ schema
Organization or LocalBusiness with the social profiles found on your own site as sameAs. FAQPage built from the questions already in your headings, answered from the paragraphs already under them.
Answer blocks
Two to three sentences that make sense quoted on their own, with the specific number first. Placed under the H1, above everything else, because that is the passage engines lift.
Headings that name the thing
"Overview" tells a model nothing. Generic headings are found and rewritten to say what the section answers, which is what decides whether a passage gets retrieved at all.
Citable facts
Pages with no figures and no sources are flagged. Models cite pages that contain specific numbers, dates and named sources, and a page of adjectives is not one of them.
Can ChatGPT even see you?
Find out in about ten seconds. Free, no account.