Ahead of Singles’ Day, here’s a set of numbers from the US. Shoppers who click through from AI to US retail sites now generate 83% more revenue per visit than people who arrive through other channels, and yet according to Adobe, a year ago they were nowhere near as valuable as everyone else. This is how we read it: AI still sends very few customers, but those customers are valuable, and within a year they’ve gone from trailing to leading, so don’t write off the thin AI line in your analytics as noise just because the numbers are small.

What happened

This comes from a retail data analysis that Adobe published on its official blog on 29 October, based on online transaction data from US retail sites covering more than a trillion visits. What it calls “AI traffic” means visits where a shopper clicked a link in an AI answer and landed on a retailer’s site; the comparison is with all visits from non-AI sources, such as email, paid search and social.

The 83% is the figure for September, and the widest gap that Adobe has measured this year. In Adobe’s own words, “In just 12 months, AI traffic has gone from being worth less than half of a typical visit to nearly twice as much.” The same goes for buying: just over a year ago, AI visitors were a little less likely to buy than everyone else, and this September they were far more likely to.

The gap hasn’t widened every single month, though. Adobe’s holiday season forecast, published in late September with data only through August, notes that the lead slipped a little in August compared with July; it was the 29 October post that showed it opening up again in September. Adobe also forecasts that over this year’s US holiday shopping season, that lead will be twice what it was in the same season last year, but that’s only a forecast, not something that it has actually measured yet.

Why it matters to you

First, the data that points the other way: very few people come from AI so far. Ahrefs, which makes SEO tools, tracks more than 110,000 websites worldwide across all industries, not just retail, and in October visitors arriving from AI made up less than 1% of the total. Adobe’s post doesn’t say what share of retail traffic comes from AI either, so the 83% is about how much each visit is worth, not a sign that AI is already bringing businesses a lot of sales.

There may be few of them, but there are more all the time: Adobe found that visits from AI in September were up around 70% on the same month last year. The way we see it, this is the part that deserves your attention: the volume is only just getting started, but the revenue from each visit has already pulled ahead.

These are all US figures, and we haven’t found anyone who has publicly measured what AI-referred visits are worth in Taiwan, so we can’t say Taiwan is the same. What business owners in Taiwan can take from it is a way of reading their analytics: for the AI line, look at the number of visitors and their value separately, because few doesn’t mean worthless.

Say you run a second-hand camera shop in Shilin, Taipei, with your own online store as well. Over Singles’ Day you open your analytics report and see only a handful of visits that list an AI address like chatgpt.com as the source, so naturally you’re inclined to skip past them. But suppose that one of those visits is a customer who first asked ChatGPT, “I want to take photos of my kids running around. Which entry-level second-hand mirrorless camera should I pick, and where in Taipei is a safe place to buy one?” The answer listed a few shops, you were one of them, and that’s why the customer clicked through. If you hadn’t been in the answer, the customer wouldn’t have found you that way, and that visit wouldn’t be in your report at all.

What to do about it

It’s better to act early, but don’t rely on that line in your report alone. Suppose that a small shop gets only a few visitors a month from AI: one sale more or one sale fewer and the figures swing wildly, so it’s hard to tell whether things are getting better or worse. Adobe’s figures, on the other hand, are drawn from over a trillion visits, and a small shop can’t read its own numbers the same way. What you should be looking at is the answer that customers read before they click through: whether you’re in it, and how it describes you.

We can pull up that answer in a demo so that you can see how it describes you today. One of our consultants goes through it with you. Since these customers arrive by clicking a link in an answer, the consultant follows the same route: asking ChatGPT, there and then, the kind of question that a customer who’s ready to buy would ask, checking whether you’re in the answer and, if you are, whether the link in it actually leads to your website. After the answer comes your website’s GEO audit score: that’s geoweb.tw’s free website audit, which checks how well AI can understand your site across 12 areas.

If you want to take it further, that’s where managed GEO comes in. You carry on running the shop and shipping orders, and our team works on three things behind the scenes: making the information that customers want on your website clear and correct; bringing what other websites and platforms say about you into line with your site; and regularly putting the same set of customer questions to ChatGPT, Gemini and other leading AI engines to see whether the answers mention you and get you right.

Further reading