Citation Gap — Documentation

How to add Citation Gap to Claude and use it: on-page GEO analysis that compares your page against the pages AI answers actually cite. Free while in beta.

Citation Gap is a Model Context Protocol connector — available in Claude, ChatGPT, and other MCP-compatible assistants — that analyses any web page's Generative Engine Optimization (GEO) — also called answer engine optimization, or AEO — and tells you why AI answers cite your competitors instead of you.

Free while in beta

Citation Gap is free to use while it is in beta. No card, no trial clock, no limits beyond fair use.

It will not stay free forever — running a report renders live pages in a real browser, and that costs money. When paid plans arrive, single-page checks will stay free, and the competitive reports will be the paid part. Everyone who signed up during the beta keeps a monthly allowance of those reports at no cost.

If you are using it now, you are helping shape what gets built. Tell me what is missing: [email protected].

What it does

Point it at a page and the query you want to win, add a few competitor URLs, and it:

• fetches and renders the live page, the way an answer engine would,
• scores its on-page GEO signals (structure, schema, answer-ready content, trust signals, crawlability),
• measures it against the competitor pages you supply,
• returns a report with paste-ready fixes and a shareable link on its own domain.

The comparison is the point. A generic checklist tells you what a good page looks like; Citation Gap tells you what the pages already being cited for your query are doing that yours is not.

Add it to your assistant

Citation Gap isn't in the app/connector directories yet, so you add it manually — with the same URL everywhere:

https://citationgap.johnlawson.dev/mcp

In Claude: Settings → Connectors → Add custom connector → name it Citation Gap → paste the URL → Connect → sign in with Google.

In ChatGPT: Settings → Connectors (turn on Developer mode under the advanced settings if needed) → Add custom connector → name it Citation Gap → paste the URL → Connect → sign in with Google.

It also works anywhere else that supports remote MCP servers, using the same URL.

Example prompts

• Guided run: Run Citation Gap — it will ask for what it needs, one question at a time.
• Complete GEO report: Run a complete Citation Gap report for https://example.com targeting "your query", comparing against https://competitor-a.com and https://competitor-b.com — run it end to end, then show me the report.
• Technical & on-page audit: Run a Citation Gap technical analysis of https://example.com and list the on-page issues with paste-ready fixes.
• AI answer-engine readiness: Is https://example.com ready to be cited by AI answers?
• Structured data & EntityMap: What structured data should https://example.com have?
• Winning-page blueprint: What page would I need to build to get cited for "your query"? Here are the pages already cited: https://competitor-a.com, https://competitor-b.com.
• Your report history: Show my past Citation Gap reports, with links.

The reports

• Complete GEO report — the full report for a page, benchmarked against competitor pages you supply for a target query: information gain (concepts those competitors cover that your page doesn't, plus net-new ideas), on-page/technical and answer-engine issues, structured-data gaps, and a prioritised action plan with paste-ready fixes and an evidence appendix.
• Technical & on-page audit — the on-page and technical issues holding a single page back (indexability, titles/meta, headings, schema validity, links, mobile, security headers, crawlability), each with a paste-ready fix, plus an overall optimisation score.
• AI answer-engine readiness — whether AI answer engines can quote and cite a single page: direct-answer opening, question headings, FAQ depth, statistics and citations, author and freshness signals, and whether AI crawlers are allowed to read it.
• Structured data & EntityMap — a single page's structured data reviewed for AI engines: which schema.org/JSON-LD types are present, missing or invalid, Organization/Author entity signals, and EntityMap coverage — with ready-to-paste markup.
• Winning-page blueprint — a build-spec for the page you'd need to win the query, reverse-engineered from the competitor pages you supply (ideally the ones already cited for it).

Data handling

Citation Gap uses the page URL, query and competitor URLs you provide to run the analysis. It signs you in with Google (name and email) so your reports and free allowance stay tied to your account — but it does not store your name or email. Your account is identified by a pseudonymous, one-way key derived from your Google account, and the name and email from your sign-in are held only for the session and removed when you disconnect. You can erase everything tied to your account at any time by asking it to "delete my data". It does not read your conversation with the assistant, your files, or your Google data. See the Privacy Policy.