Measuring GEO: A Weekly Probe to Check If AI Actually Cites You
A cheap weekly probe using Perplexity Sonar checks whether AI assistants actually cite your content, turning GEO into a measurable trend.
GEO advice — FAQ schema, Q&A phrasing, clean HTML — is everywhere, but almost nobody verifies whether any of it results in an actual AI citation. The author built a lightweight weekly probe that fires a fixed set of user-phrased questions at a grounded model (Perplexity Sonar via OpenRouter) and checks whether the returned citation URLs contain owned domain patterns.
The measurement is deterministic — a plain substring match, not an LLM grading itself — and scoped to owned paths like owner/repo rather than bare project names, since generic names can collide with unrelated projects and falsely inflate the citation rate.
Key engineering details include treating API failures as distinct 'error' records instead of false 'not cited' results, so an empty API credit balance never gets mistaken for a citation collapse, and waiting until content is actually crawled and indexed before measuring, since publish-to-citation follows a crawl-rank-cite pipeline that takes weeks.
The probe costs under a dollar a month and is meant to complement passive signals like Bing Webmaster Tools' AI Performance report and Google Search Console, since different AI assistants draw from different underlying search indexes.