We built the report our own analytics refused to show us.

Citealytics started with a spreadsheet. A growing share of sign-ups said they had “found us through ChatGPT”, and no analytics tool we ran could show a single one of those sessions as anything other than Direct. The traffic was real, the revenue was real, and the reporting was blank.

What we measure

Two things, both from your own server-side request stream. First, answer-engine crawlers: which bots fetch which URLs, matched against a public registry of tokens, regexes and IP ranges that we keep current. Second, human referrals from assistant surfaces: the visits that follow a citation, split per engine, attributed to the page that earned them.

What we refuse to collect

No cookies. No localStorage. No fingerprinting. No retained IP addresses. No cross-site identifiers of any kind. This is a product constraint rather than a setting, which is why there is no consent banner to configure and no consent-mode gap to explain to a legal team.

How we operate

Revenue, customer count and events processed are published on the open startup page, updated as the numbers move rather than when they flatter us. The crawler registry is public on the crawlers page so you can audit our classification instead of trusting it, and every registry change is recorded in the changelog.

We are small and intend to stay focused: one channel, measured properly, rather than a general-purpose analytics suite competing on feature count. If a tool already reports something well, we would rather point you at it than reimplement it badly.