How we measure what AI says, and what it sends you

1. 1.We ask each question many times. AI assistants give a different answer each time you ask. So we report how often you're named, for example "named in 60–75% of answers", instead of a single rank.
2. 2.We ask the questions your buyers ask. They're built from the roles, needs and deals in your CRM, and asked the way a real buyer would, in their country and language.
3. 3.We only count a visit as "from AI" when there's evidence. The browser's referrer, a link tag, or a landing page that AI quoted for that question. No evidence, no AI label.
4. 4.We identify companies only with consent. Company names worldwide. Named people in the US only. Never individuals in the EU.
5. 5.We tell you what we can't see. Some visits influenced by AI arrive with no trace. So our counts lean low: we'd rather miss a visit than claim one we can't back up.

Most visits from AI assistants show up in Google Analytics as "Direct", because many assistants don't tell the browser where the visitor came from. Our script, installed on your site, looks for evidence in this order:

1. 1.The referrer. When the browser says the visit came from chatgpt.com, perplexity.ai or another assistant. The most reliable signal.
2. 2.A link tag. When the link carries a tag such as utm_source=chatgpt.com.
3. 3.The landing page. When the page the visitor landed on is one AI quoted for a question we track. This is often the signal that finds AI visits Analytics misses, and it's possible because we already know which pages AI quotes.
4. 4.The visit pattern. Timing and device signals that support one of the above. Never used on its own.

When none of these is present, we don't guess. The visit stays unlabelled.

A visit from AI is a start, not a lead. The intent score shows how close a company looks to buying, based on what its visitors do on your site.

You choose the score at which companies are sent to your CRM.

Example: one company over four days

Your sales team sees the company, the question it asked ChatGPT, the pages it read, and its score, all in the CRM record.

Every method has blind spots. These are ours.

1. What people type into AI apps. No one outside OpenAI, Google or Anthropic can see real user questions. We see the answers to the questions we ask, which are built to match your buyers.
2. Visits that leave no trace. If someone reads an AI answer, then types your address or searches your name on Google, the visit looks like any other. Our AI visit counts are lower than the true number.
3. Visitors who decline cookies. They're not tracked or identified. In markets where many people decline, you'll see fewer companies.
4. Proof that AI caused a deal. We show that a deal started with a visit from AI. We can't prove the AI answer was the reason the buyer chose you.
5. Model updates. Assistants change without notice. A sudden shift in your numbers can come from an update, not from anything you or a competitor did.

• No tracking until a visitor accepts cookies, and none if they decline.
• Named individuals in the US only. Company-level data only in the EU and EEA.
• We don't sell visitor data.
• Any visitor record can be deleted on request.
• Your data is never used to train AI models.
• Collection from AI assistants uses their official access or follows their terms of use.

Read the full privacy policy and GDPR and visitor identification.