Your analytics is hiding your AI search revenue

Often it does not. In data OtterlyAI shared in October 2026, Google Analytics credited Claude with 0.1% of conversions. When customers were asked how they found the company, 10.6% said Claude. People often read an AI answer and then come back later by typing your name. Analytics credits that visit to search or direct traffic. Asking customers is the simplest fix.

OtterlyAI sells AI search monitoring software. It compared two views of its own sign-ups. The first was Google Analytics 4 with last-touch attribution. The second was what customers said when asked how they heard about the company.

Solid: GA4 last touch Hatched: what customers said

GA4 gave Google Search 35% of conversions. Only 15.7% of customers named it. For Claude the gap ran the other way. GA4 gave it 0.1%. When asked, 10.6% of customers named Claude. ChatGPT moved from 7% in GA4 to 11% in the customer answers.

OtterlyAI also reported these figures from customer answers. Claude and ChatGPT together accounted for about a quarter of its revenue. The average customer who came from Google spent around 15% less than one who came from Claude. From June to July, new paying customers who named Claude rose 113%. ChatGPT rose 25% and Google 32%.

An AI assistant often answers the question without sending a click. The buyer reads the answer, remembers your name and comes back days later. They might search for your brand or type your address. Last-touch analytics gives that sale to Google or to direct traffic. The AI answer that started it gets no credit.

OtterlyAI describes this as a funnel. The AI answer creates brand coverage at the top. Below it sit three kinds of visit. Some are AI bots and agents. Some are people clicking a link in the answer. Others are people arriving later through branded search or direct. Only the middle one shows up clearly in analytics.

This is one company. It sells AI search software, so its customers are more likely than most to use AI assistants. The figures come from what customers remembered, which is never exact. So I would not use 10.6% as a benchmark for your business.

What the data does show is the size of the gap between what analytics sees and what customers report. If your business is anything like this one, you may be undervaluing AI search by a lot. You will only know by asking. My note on how to read an AI benchmark claim sets out the questions I put to any vendor's numbers.

Ask every new customer how they found you. Report the answers next to your analytics every month. Check more than one AI assistant, because they often disagree. My note on how ChatGPT, Claude, Gemini and Perplexity answer the same question found they split on naming a brand in more than one case in four.

Free template: the question to add

Add one required question to your sign-up, checkout or enquiry form. Ask "How did you first hear about us?" and offer these options.

• Google search
• ChatGPT
• Claude
• Gemini
• Perplexity
• Microsoft Copilot
• Another AI assistant
• LinkedIn
• A friend or colleague
• Other (please tell us)

Keep the AI assistants as separate options. Each one behaves differently, so grouping them as "AI" hides the difference.

Free template: the monthly report

Put the answers next to your analytics in a table like this one. The example below is filled in for an imaginary business so you can see how it reads.

Scroll sideways to see every column

In this example, analytics gives the four AI assistants 8% of new customers. The customers themselves say 29%, which is 58 of the 200. Copy the layout and put in your own numbers. Fill it in each month. After three months you will see whether AI search is bringing in customers that analytics does not show.

Getting help

To see how often AI assistants name your brand, an AI visibility tool such as LLM Scout tracks prompts across engines. If you want an agency to improve your visibility in AI answers, Orchid Box is an AI search agency that offers AI search optimisation. To set up the measurement and the team routine around it, see work with me.