AI Traffic Is One Percent of Sessions. Why the Click Count Undersells It.

Your GA4 report says AI assistants sent about one percent of sessions last month. Someone on the team asks why AI visibility deserves budget at all. The honest answer starts with how AI search behaves. Semrush analyzed almost 69 million US desktop Google Search sessions from May to July 2025 and found that 92 to 94 percent of Google AI Mode searches ended without a click to any external site. Only 6 to 8 percent of AI Mode sessions led to an external visit.

That figure describes Google AI Mode, not AI shopping in general, and it is worth being precise about that. It still explains the problem. When the assistant answers the question in place, the shopper rarely needs to click, so the influence on what they buy happens before your analytics see anything. The click count measures what leaked out of the answer, not what the answer did.

The click was never the measure of AI influence

Ecommerce analytics grew up around the click because the click was the only trail a shopper left. Search engines could not finish the job, so the shopper visited the site, found the product again, added it to cart and checked out. Every step produced a session you could count.

An AI answer collapses most of that. A shopper who asks Google AI Mode or ChatGPT for the best waterproof hiking boot for wide feet, reads a comparison and later buys yours on Amazon or in a store leaves no referral session anywhere. The research happened, the decision happened, and your dashboard recorded nothing.

This is why two statements can both be true: AI traffic is one percent of sessions, and AI already shapes a meaningful share of purchase decisions. The one percent is the visible floor. For the visits that do arrive, including the ones that land in Direct because the app stripped the referrer, see how to measure AI referral traffic in GA4 and Shopify.

What the clicks that do arrive are worth

The visits that survive the zero-click filter behave differently from organic search visits. Shopify's Q1 2026 commerce data, published in May 2026, showed AI-referred orders growing nearly 13x year over year while AI-referred sessions grew more than 8x, with AI-referred sessions converting at nearly 50 percent higher rates than organic search. We looked at what the gap between orders and sessions means for a lean team.

Retailers running their own assistants report the same pattern. On its fourth quarter earnings call in February 2026, Walmart said customers who use its Sparky assistant have order values about 35 percent higher than customers who do not. Amazon has claimed Rufus generated $12 billion in incremental sales in its first season, a figure that AdExchanger noted Amazon has not substantiated. Treat the Walmart number as evidence and the Amazon number as a claim.

Being in the answer also lifts the organic clicks you still get. Position Digital's roundup of AI search research cites Seer Interactive's finding that organic click-through rate is 35 percent higher when a brand is cited inside an AI Overview. The few searches that do produce clicks go disproportionately to the brands that appeared in the answer.

What gets a product into the answer

The research on what AI engines cite looks different from the link-building playbook, and it points at product information rather than authority alone.

• Put the facts first. A Search Engine Land study of LLM citations found that 44.2 percent of citations come from the first 30 percent of a page. Product facts buried under tabs or below a wall of lifestyle copy are easy to miss.
• Third parties carry weight. AirOps found that in brand discovery queries, brands are 6.5 times more likely to be mentioned through third-party sources than through their own domains. Reviews, comparisons and press shape the answer alongside your own pages.
• Google reads your feed. Google's own guidance for appearing in AI features asks sites to keep structured data matching the visible page and Merchant Center information up to date. For shopping answers in AI Mode, the product data Google already holds decides whether a product can be shown at all. We covered why a brand shows up in ChatGPT but not Google AI Mode separately.

Why one query returns you and the next does not

A question we hear from ecommerce leaders is why the same prompt returns their brand on Monday and a competitor on Tuesday. That is how these systems behave. Position Digital's roundup cites Ahrefs research showing AI Overview content changes 70 percent of the time for the same query, and SparkToro research showing less than a one in a hundred chance that ChatGPT or Google's AI returns the same list of brands in any two of a hundred responses.

Two things follow. A single spot check proves nothing in either direction. And a one-time optimization project does not hold, because the answer is rebuilt every time from whatever product information and third-party evidence is current. Consistent presence comes from product data that stays complete and accurate, checked on a schedule.

How to tell if you are in the answer

Since the click will not tell you, you need a measurement that works upstream of it. The approach that holds up is simple. Write the questions your real customers ask, in their words and for their situations, for example a parent buying a first bike or a runner replacing worn shoes. Ask the same set every week in ChatGPT and Google AI Mode. Record whether you appear, whether you appear as a product card or only as a link, and which products are named.

That appearance rate is your baseline. When you fix product data, retest the same questions and compare before and after. Pair it with the AI referral view in GA4 and Shopify so sessions and orders sit beside appearances, reported as separate numbers. If you want a starting read, the free AI readiness report shows how your products currently appear for a set of shopper questions.

The decision

Fund AI visibility on the strength of influence and conversion quality, not on session share. The share will stay small for a while because the behavior is zero-click by design. The buyers it does send arrive decided, and the recommendation itself shapes purchases you will never attribute. The commercial risk is not a traffic drop you can see. It is a comparison you were never in.

This is the work Paz does for lean ecommerce teams. A Paz expert checks how your products appear for the questions your shoppers ask, finds the product information gaps that keep you out of the answer, and proposes fixes with the evidence behind each one. You approve, and the platform publishes the changes and tracks appearances, sessions and orders before and after. Learn more about AI visibility with Paz, or Talk to us.