How CitationWorks works

What we ask

We read your website and work out what you sell, who you sell it to, and where. Then we write the questions a buyer would ask an AI assistant before choosing someone like you.

Those questions never name you. "Who is the best accountant for startups in Austin?" tells you something. "Is Acme any good?" doesn't.

You can add your own questions and pause any of ours. Your plan sets how many run, from 3 to 100.

Where we ask

Every question goes to ChatGPT, Claude and Gemini with live web search on, the way your buyers use them. An assistant without search only knows what it memorised, and it would ignore almost every small business.

Most plans scan every day. So you see a trend, not a single snapshot.

One honest note: we reach ChatGPT through OpenAI's own models with web search. That follows what chatgpt.com says closely, but it is not a copy of it.

How we score an answer

A separate AI model reads each answer and records what happened. Were you named? How high? Which competitors were named or linked? Was your own site cited? Did the answer contradict the facts on your own site?

An answer we couldn't score never counts against you. It isn't a miss. A day with no scan is a gap in the chart, never a zero.

Each week you get a short brief: what changed, and the few things worth doing next.

What we don't claim

AI answers change from one ask to the next. That is why we measure over time, so one odd answer doesn't move your numbers much.

We can't control what an assistant says, and we can't see inside its training data. We show you what it says, what seems to drive it, and what to try next.