The two numbers the AI visibility industry keeps confusing

There are a great many numbers circulating about AI and ecommerce just now, and most of them are quoted with far more confidence than they have earned. None of it needs to worry you. I have read the studies so that you do not have to, and the message underneath the noise is calm: AI has not remade ecommerce yet.

What the noise has produced is two figures for how well AI traffic converts. One is roughly 15 percent. One is roughly 2 percent. Both are real, both are measured, and the gap between them is where a great deal of money is about to be spent badly.

15.9%: Seer Interactive case study, late 2024 into 2025, against Google organic at 1.76%. 2.47%: analysis of 329 ecommerce brands. Both figures are genuine. They describe different businesses.

I kept hearing the 15 percent in rooms where people were making budget decisions. It did not match anything I had seen in a real retail P&L, so I went and found where it comes from, where the other number comes from, and which one an operator should actually plan against. The answer is more useful than either camp lets on, and it starts with giving the big number a fair hearing before taking it apart.

Where the 15 percent comes from, and why it is not wrong

The high figure is real. It traces to a Seer Interactive case study running from late 2024 into 2025, which found ChatGPT referrals converting at 15.9 percent against Google organic at 1.76 percent. That is a genuine measurement, honestly reported, and it has since been corroborated in peer-reviewed work showing AI-referred visitors converting at multiples of organic.

The mechanism underneath it is real too, and worth understanding rather than dismissing. Someone who clicks through from an AI answer has already done the research. The comparisons are handled, the shortlist is drawn, the brand has been pre-qualified by the model before the person ever lands on the site. They arrive further down the funnel than any cold organic visit. High intent converts at a premium. That is not marketing, it is how the channel actually behaves.

So anyone quoting 15 percent has a real number and a real reason behind it. The problem is not the measurement. It is what happens to the measurement next.

Why the same number collapses in retail

That 15.9 percent is one B2B software client. It rests on 1,370 AI conversions measured against nearly 14 million organic sessions, and the people who ran it are clear that it should be read as a case study, not an industry benchmark.

Point the same question at retail and the number falls out of the sky. One analysis of 329 ecommerce brands found AI traffic converting at 2.47 percent. A second, across 310 brands, put it at 2.68 percent. The largest study I could find, 973 sites and 20 billion dollars of revenue, found ChatGPT traffic converting worse than organic search, not better. A separate controlled analysis across 54 sites found no statistically significant difference in B2C conversion at all.

So the honest answer to how well AI traffic converts is that it depends entirely on whether you sell enterprise software or jackets. For most retail brands the figure sits in low single digits, in the same postcode as the numbers those businesses already see from organic search. Most people quoting the headline are quoting the wrong sector.

The borrowed authority problem

There is a second move that should make everyone uneasy, and it is the one worth naming most plainly.

The high conversion figure often arrives wearing a consultancy badge. McKinsey, Bain, Gartner. Those names carry it into boardrooms and settle the argument before it starts. Except none of them published a conversion rate. What they actually published is adjacent and far more carefully hedged. Gartner projects that 20 percent of digital commerce transactions will run through AI platforms by 2030. McKinsey sizes agentic commerce at up to a trillion dollars of US retail by the same year. Bain, in research from early 2025, found that zero-click search has already cut organic web traffic by an estimated 15 to 25 percent.

Every one of those is a long-range forecast about where the market is heading. Not one is a conversion rate. The logos get stapled onto a single case study after the fact, because a famous name makes a soft number sound settled. When you see a conversion figure attributed to a strategy consultancy, it is worth asking which report, because the honest answer is usually none of them.

The number is not even the point. The volume is.

This is the part that decides it, and it holds regardless of whose conversion rate you accept. Across every serious dataset, AI referral is still a rounding error of traffic. One thirteen-month analysis put it at under 2 percent of referral traffic. The 20-billion-dollar study put it nearer 0.2 percent of sessions. Ahrefs stated it most starkly: ChatGPT handles roughly 12 percent of Google’s search volume but sends 190 times less traffic to actual websites.

Fifteen percent of almost nothing is still almost nothing. Any revenue forecast that leans on the high conversion rate has quietly assumed a traffic volume that does not exist yet. Rate multiplied by volume is the only equation that pays a wage, and right now the volume is the binding constraint, not the rate. A brand that optimises hard for AI visibility today is optimising for a channel that, for most retailers, is not yet moving enough people to register on the revenue line.

None of which means it is a mirage

The channel is real, and the direction of travel is fast. On a stable US cohort, AI referral grew 6.5 times in twelve months while total traffic stayed flat. That is a genuine structural shift, not a spike.

A large part of it is also invisible. The shopper who takes a recommendation from ChatGPT, then types the brand into Google and buys, lands in branded search, where no one credits the AI that started the journey. The real influence of the channel is larger than the measured traffic, which cuts both ways: it means the volume is understated, and it means anyone quoting a clean conversion figure is measuring only the visible sliver. Adobe’s own data has swung from AI traffic converting 38 percent worse than other channels to 42 percent better inside a year, which tells you how unsettled the measurement still is.

This is worth preparing for now, seriously and with a clear head. The clear head is the whole point.

Three questions before a number touches a budget

This is an industry that does not yet agree on what AI discovery is, where it is heading, or the numbers that would make it a commercially viable channel rather than a line on a slide. In that fog, the temptation is to reach for the most flattering figure, borrow a famous name, and build a forecast on it. The discipline is to ask three questions of any AI conversion number before it informs a single decision.

Get those three right and AI discovery becomes a channel you can plan for with confidence, at the right pace, ahead of most of the market. Get them wrong and it becomes the most expensively cited rounding error in retail.

The businesses that win the AI discovery channel will not be the ones that believed the biggest number. They will be the ones that asked the smallest, hardest questions first.

Figures are cited to their sources in the text. The 15.9% figure is a single B2B software case study (Seer Interactive, late 2024 into 2025); the retail conversion figures are drawn from multi-brand analyses of 329, 310, 973 and 54 sites; traffic-share figures are from Ahrefs and Adobe. The Gartner, McKinsey and Bain figures are long-range market forecasts, not measured conversion rates. International figures should be read as directional signals of scale rather than UK performance targets.

Back to Thinking