Is AI Search Actually Making You Money? How to Tell

Why AI search revenue is harder to measure than it looks

Every brand has been told to get ready for AI search. Very few can say what it's worth to them today.

The usual instinct is to look at traffic. That's the wrong place to start. AI assistants answer the question inside the chat, so a buyer can compare three brands, pick one and only click once, or not click at all. Measured by sessions, AI search looks tiny. Measured by what those sessions are worth, it looks a lot bigger.

Adobe's data on US retail sites found that in March 2026, visits from AI sources converted 42% more often than visits from everywhere else, and earned 37% more revenue per visit. A year earlier, AI visitors converted at around half the rate. A separate 12-month study of 94 ecommerce brands by Visibility Labs found ChatGPT referrals converted at 1.81%, against 1.39% for non-branded organic search. The same study found ChatGPT revenue was still only about 1.5% of what those brands earned from non-branded organic.

For most brands, AI search is a small channel that's growing fast and sends visitors worth more than average. Measuring it by traffic alone undersells it.

Are you in the answer?

Before anything else, find out whether AI assistants recommend you at all. If they don't, there's no revenue to measure and the job is visibility first.

Most brands aren't on the list yet. Our State of AI Visibility study tracked 2,541 ecommerce brands across ChatGPT, Perplexity and Google AI Overviews, and 72.6% of them appeared in fewer than 5% of AI shopping answers in October 2026.

Use a fixed set of real buyer questions for your category. Think of what a customer actually asks, like the best brand for a specific use, a gift idea or a comparison with the market leader. Ask the same questions on ChatGPT, Perplexity and Google's AI Overviews, every week, and record which brands get named.

Two things matter in what you record.

Share of answers

Share of answers is the percentage of those answers that name your brand. It's the closest thing AI search has to a ranking. An AI answer usually names a short list of five or six brands and has no page two, so the question is whether you're on the list.

Mentions versus recommendations

Being named and being recommended are different results. An answer that says "some people also consider your brand" counts as a mention. An answer that says "your brand is the best choice for this" is the one that sells. Track them separately.

You can do this by hand with a spreadsheet, or check where your brand stands for free on Found, our AI visibility ranking, which runs a fixed question set for each category every week. Whatever you use, make sure the method is written down. Our Found methodology is a good template for what a credible one includes.

The visits and orders you can trace

The second layer is revenue with a clear trail back to an AI tool.

When a buyer clicks a link in ChatGPT, Perplexity, Gemini or Copilot, the visit usually arrives with that tool's name as the referrer. Most analytics setups lump these in with other referral traffic, so ask whoever runs yours to give AI assistants their own channel. Then you can see visits, conversion rate, orders and revenue per visit from AI on their own, next to organic search and paid.

Shopify adds a second source. Shopify's agentic storefronts let buyers find and buy your products inside AI chats, including ChatGPT, Google AI Mode and Gemini, and Microsoft Copilot. Orders placed through those channels show up in your Shopify admin with channel or referrer attribution, so you can see which AI channel each order came from.

Google is the gap in this layer. Clicks from AI Overviews and AI Mode land in your normal Google organic numbers, so they can't be pulled out as AI traffic. Google started rolling out generative AI reports in Search Console in June 2026, but only to some sites. They show which of your pages appeared in AI features and how often. They don't show the clicks or revenue behind them.

The revenue you can't trace

This is the layer most brands skip.

A common buying journey now looks like this. Someone asks ChatGPT for the best option in your category, sees your brand on the list, then opens Google, searches your brand name and buys. Your analytics credit that sale to branded search. The AI assistant that actually introduced you gets nothing. The Visibility Labs study reached the same conclusion. Standard analytics probably understates how much ChatGPT influences sales.

You'll never trace all of it, but you can get a fair read with two cheap signals.

• Ask your customers. Add a post-purchase question like "Where did you first hear about us?" and include "ChatGPT or another AI assistant" as an option. Within a few weeks you'll have a self-reported share of AI-influenced orders.
• Watch branded search demand. If your share of AI answers rises and more people start searching your brand by name in Search Console, the two are probably connected.

Put AI search revenue on one scorecard

The three layers only make sense side by side. A rise in share of answers should be followed by more AI-referred visits, more AI-referred orders and a higher share of customers saying AI is where they found you. When those move together, you can be confident the work is paying off.

Run the weekly measures every Monday and the monthly ones at month end. A single number never tells the story. The trend across all six does.

What to be sceptical of

AI search is new enough that a lot of the reporting around it is loose. A few things deserve a hard look.

• Screenshots of a single AI answer. Answers change between runs, accounts, locations and days. One screenshot proves the brand appeared once. It says nothing about how often.
• Dashboards that claim to know your exact AI revenue. Nobody can see the buyer who read a ChatGPT answer and then searched your brand on Google. I'd treat any tool that reports AI revenue to the dollar with suspicion. It's reporting the traceable part and leaving the rest out.
• Visibility scores with no published method. If you can't see which questions were asked, on which platforms and how often, you can't compare the score week to week.
• Reading too much into the first month. Expect four to six weeks of baseline before any change in the numbers means something. Most brands want answers sooner, and that's how they end up chasing noise.

The real cost of doing this properly is ownership. Someone has to run the questions, update the scorecard and read it every week. Without that, the numbers go stale within a month and the decisions go back to guesswork. Our complete guide to AEO for Shopify Plus covers how to set up a baseline you can repeat.

Set a revenue goal

Once you can measure all three layers, you can set a goal that means something to the business.

"Appear in more AI answers" is a fine activity, but it's a weak goal. Nobody in finance can tell you what it's worth. A goal like "lift revenue from AI-referred and AI-influenced orders by a set amount over the next 100 days" can be tracked and judged. It also forces the right work. Visibility only counts when it turns into buyers.

That's how we set goals in Lift, our 100-day AEO partnership. We agree the revenue goal and the baseline in week one, then report against all three layers as the work goes in.

The bottom line

AI search is already sending some of your most valuable visitors, and most brands can't see it clearly yet. Measure whether you're in the answer, what you can trace and what you can't, and read them together. If you can't say what AI search is worth to you today, start by finding out whether you're on the list.