Survelium Docs
Docs menu

AI traffic

People from AI

How many visits came from people who clicked in an AI assistant.

People from AI is the number of visits to your site that came from AI assistants — people who clicked a link in a ChatGPT, Perplexity, Claude, Gemini or Copilot answer.

Formula

people from AI = visits whose referrer or utm_source matches an assistant's domain

A visit counts on a page view that comes from an assistant. We recognise the assistant by:

  • referrer — the address of the page the visitor came from (e.g. https://www.perplexity.ai/...);
  • utm_source — the parameter some assistants add to the link themselves (ChatGPT adds utm_source=chatgpt.com).

Recognised domains: chatgpt.com, chat.openai.com, claude.ai, perplexity.ai, gemini.google.com, copilot.microsoft.com, you.com, phind.com.

Clicks on a ChatGPT ad are listed separately, as ChatGPT Ads, so paid visits do not pass for ChatGPT recommending the site. They are recognised by the ad's link: utm_source=chatgpt with a paid utm_medium (paid_social, search_paid) or the oppref / olref parameters.

Example

In the last 7 days, 30 visits came with utm_source=chatgpt.com, 12 with the referrer perplexity.ai and 5 with the referrer gemini.google.com:

people from AI = 30 + 12 + 5 = 47

If the 7 days before had 40, the number shows +18%.

What counts and what doesn't

Counts Doesn't count
Page views with a referrer or utm_source from an assistant Visitors the assistant does not mark (mobile apps often send no referrer)
Visits reported by the server integration or by the beacon Clicks from AI Overviews or AI Mode: Google sends them as traffic from google.com
Crawlers — they are in crawler hits

On a site with both (server integration and beacon), the server wins: while the server has reported a visit from AI in the last 7 days, visits from AI reported by the beacon are not saved, so nothing counts twice.

We keep nothing that identifies the visitor: no IP, no cookies, no full referrer — only the assistant, the page, the time, the device type (read from the user agent) and, for visits reported by the beacon, the country.

How to read it

  • It is a minimum: some people sent by assistants arrive without a referrer and show up as direct traffic in any analytics tool, this one included.
  • Put it next to live fetches: many live fetches and few people mean assistants read your pages, but people get the answer without clicking.
  • Referrals → Landing pages shows which pages bring people from AI — usually the ones assistants cite.

Where it appears

  • AI traffic → Overview — the number, the change and the "People from AI" series in the chart.
  • AI traffic → Pages — the "People from AI" column, on the landing page.
  • AI traffic → Referrals — the daily chart, the assistants (with each one's change) and the landing pages.

CSV export

AI traffic → Referrals → Export CSV downloads every AI visit in the period chosen on the page, one row each, newest first: Date (UTC), Assistant, Landing page, Country and Device. Country is empty for visits reported by the server integration.

View as Markdown · llms.txt