Can ChatGPT cite you? There's a rubric for that

What it does It fetches public URLs and grades them 0–100 against a published, versioned rubric (v1.1) across five areas: Reachable, Understandable, Content Citability, Brand Credibility, and Answer Fit. The scorer is one Python file on the standard library — no key, no dependencies, roughly twenty seconds per site. Two further levels take your own OpenAI, Perplexity, Gemini, Anthropic, or OpenRouter keys: ask the engines one question and see whom they cite, or track a fixed question set week over week.

The interesting bit The rubric deliberately keeps citation results out of the 0–100. Readiness and outcome sit side by side and never get summed, because a site can score 90 and still lose every answer to a competitor with more third-party coverage — and the reverse happens too. Checks that can’t be verified mechanically, like independent mentions, are flagged “needs judgement” and dropped from the denominator instead of guessed.

Key highlights

• Zero dependencies: Python 3.8+ standard library, one file for the free score
• Open, versioned rubric — every check, its evidence, and its points are on display, so you can argue with it
• Citation checks run on your own keys across OpenAI, Perplexity, Gemini, Claude, and OpenRouter
• Doubles as an MCP server; only ask and run can spend money, and --read-only removes even that
• A public leaderboard of 317 well-known sites — a quarter of them unreadable to AI crawlers

Caveats

• Levels 2 and 3 need cli/geo_watch.py sitting next to cli/geo_score.py; the famous one-liner only covers level 1
• Four checks are judgement calls excluded from the denominator — the demo run reads “61 / 86 observable,” not a clean 100
• The MCP server reads provider keys from its own environment, not your shell; documented, but an easy way to lose an afternoon

Verdict Worth twenty seconds if you publish anything and wonder whether the answer engines can even parse it. If AI citations aren’t your problem yet, the free tier still works as a blunt crawlability and structured-data audit.