What AI search sends you, and what it hides

Lead with the point. This article breaks down why AI search traffic looks misleadingly small on your analytics dashboards and explains how to track and estimate the real impact of AI recommendations on your brand.

What AI search sends you, and what it hides

AI looks small on every dashboard

Ask a marketing team how much business AI search brings in, and the answer is usually "very little". Their dashboards agree. In most Google Analytics accounts, ChatGPT, Gemini and Perplexity together barely register.

That number is real, but it only counts the people who clicked. Most people who see a brand in an AI answer never click. They read the answer, close the app, and come back later through Google or the app store. None of that is labelled as AI.

So the better question is: what can we measure today, and what are we missing?

What you can measure today

Five sources tell you something about AI today. Each one has a catch.

Source	What it tells you	The catch
Google Analytics	Visits from ChatGPT, Gemini, Perplexity and Claude, as referral traffic	ChatGPT does not always tag its links, so some of its visits land in "Direct". Gemini started tagging its links only around 30 September 2026.
Google Search Console, generative AI report	How often a link to your page was shown in AI Overviews and AI Mode (since June 2026)	No clicks and no search queries. Counts links only, not mentions of your brand.
Bing Webmaster Tools, AI Performance	How often your pages were cited in Copilot and Bing's AI summaries (since February 2026)	Citations only. No impressions, no clicks, no position in the answer.
Your server logs	Visits by ChatGPT-User, which fetches a page when a person asks ChatGPT something	Shows your page was read, not that you were recommended.
"How did you hear about us?"	Customers who say ChatGPT or another assistant	Depends on memory, and the numbers are small.

Together, these tell you who clicked and which of your pages were linked. That is useful. It is also the smaller part of the story.

The missing piece: being named

No platform tells you how often your brand was named.

Google and Bing count links to your pages. But AI answers often name a brand without linking to it. A Google AI Overview for "best credit card for UPI payments" can list five cards while linking only to review sites. Those five brands were shown to the searcher. None of them sees that impression in Search Console. The review sites do.

ChatGPT, Perplexity and the Gemini app share no impression data at all. So the biggest number in the chain, how many people saw your name, is the one nobody gives you.

Why this makes AI look small

Pew Research tracked real Google searches and found people clicked a link inside the AI summary on just 1% of visits. When a summary appeared, clicks on normal results fell to 8%, from 15% without one.

The rest of the effect comes later. Similarweb, which sells traffic data, found brands recommended by ChatGPT got 2.5 times more visits within a week, mostly from people searching the brand name. That is a vendor's number, not an independent one, but the pattern makes sense.

Those later visits show up as brand searches or direct traffic. So AI looks small, because most of its work is credited somewhere else.

Three ways to estimate what you cannot see

You cannot measure the missing piece directly yet. You can estimate it. Here are three ways, from simple to advanced.

A rough impression estimate. Take the monthly Google searches for the questions buyers ask in your category. Multiply by the share of searches that get an AI answer. Multiply again by how often AI names you, which you can check by asking those questions regularly. For example, 40,000 searches × about a third with an AI answer × named in 20% of them ≈ 2,700 times a month that people see your name. The share with an AI answer is itself an estimate: Pew Research found about one in five US searches in March 2025, and Similarweb, which sells traffic data, reported 43% by May 2026. Anyone can do this in a spreadsheet.
A lift test. Pick two products or service lines. Work on how AI describes one of them, and leave the other alone. Then compare searches for each product's name, direct visits and sign-ups, before and after. If the one you worked on moves and the other does not, you have strong evidence that the work caused it.
A marketing mix model. This estimates how much each channel adds to sales over time. Large brands can add AI visibility as one more input. Google's Meridian and Meta's Robyn are free, open-source tools for this. Both need a long history of weekly data, so they suit big advertisers more than small ones.

Start the record now

This record cannot be built later. You can look up last year's Google searches and last month's analytics any time. Nobody can look up what ChatGPT said about your brand last month. So start the record before you need it.