The owner of a small online homeware shop notices something odd in her sales. A handful of new customers arrive on her website, go straight to one product page, and buy within minutes. No browsing, no comparing, no abandoned basket. When she asks one of them how they found her, the answer surprises her: “I asked ChatGPT for a handmade ceramic serving bowl under R800 that ships to Cape Town, and it suggested you.”
That customer had done all her comparison shopping before she ever reached the website. The AI assistant had read the reviews, checked the prices, weighed the options and arrived at a shortlist. By the time she clicked, she was ready.
For years, small businesses have competed to be found in search results. A new kind of discovery is growing quickly alongside it: being recommended by an AI assistant that a customer has asked for advice. And the early evidence suggests the people who arrive this way are unusually likely to buy.
One Like insight: When an AI does the browsing, your website’s job shifts from persuading the visitor to confirming the recommendation.
The numbers behind the shift
Adobe tracks online shopping across more than a trillion visits to US retail websites, which gives it an unusually clear view of where shoppers come from. In its analysis of the 2025 holiday season, traffic to retail sites from generative AI tools, meaning shoppers clicking a link suggested by a chat assistant, rose 693% compared with the year before.
That percentage needs context: it’s growth from a small starting point, and AI assistants still send far fewer visitors than search engines or social media. The more interesting finding is the quality of those visitors. During the 2025 holidays, AI-referred shoppers converted 31% more often than visitors from other sources, and were noticeably less likely to leave immediately. That’s a reversal from early 2025, when AI visitors converted less often than average. Shoppers, it seems, are learning to trust the recommendations and arriving ready to act.
The data is American, from large retailers, and covers a specific season. South African adoption may differ, and the pattern could change as AI tools evolve. But the direction matches what many business owners are starting to notice anecdotally: customers who arrive having already been told, by an assistant they trust, that you’re a good fit.
What AI assistants look for
Nobody outside the AI companies knows exactly how their assistants choose which businesses to recommend, and the methods keep changing. But the assistants can only recommend what they can find and understand. That points to a few sensible habits.
Be specific about what you sell. Customers ask AI assistants detailed questions: a price range, a location, a material, an occasion. Product pages that state those details plainly, such as size, materials, price, delivery areas and lead times, give an assistant something concrete to match.
Make your claims verifiable. Recent reviews, clear policies on returns and delivery, and consistent business details across your website and directories all help build a trustworthy picture. As we explored in Your Best Reviews Might Be Too Old to Count, recency matters to people, and it’s reasonable to assume it matters to the tools that summarise them.
Make the landing page do one job. AI-referred visitors tend to arrive with a specific need. If the page they land on confirms it quickly, with the price, availability and a clear way to buy or book, the recommendation turns into a sale. If they have to hunt, they may go back and ask the assistant for an alternative.
[Infographic placement] The new shopping path. Two journeys side by side. Traditional: search → browse several sites → compare → shortlist → buy. AI-assisted: ask an assistant a detailed question → receive a shortlist → click one link → confirm details → buy. Caption: Adobe (2025 holiday season, US): AI-referred retail traffic up 693% year on year, converting 31% more than other sources.
Testing your own visibility
The simplest test costs nothing. Ask two or three AI assistants the questions your customers would ask, with the detail they’d use: “Where can I buy a handmade ceramic serving bowl under R800 with delivery to Cape Town?” or “Who does same-week bathroom repairs in Pinetown?” Note whether you appear, what the assistant says about you, and whether it’s accurate. If you don’t appear, look at the businesses that do. Their websites will usually tell you what information the assistant found useful.
It’s also worth adding “an AI assistant” as an option when you ask new customers how they found you, so you can see whether this channel is starting to matter for your business.
A few questions can help:
- Do our product or service pages state prices, specifics and delivery areas plainly?
- Are our reviews recent and our business details consistent everywhere?
- When an AI-referred customer lands on our site, can they confirm and buy in under a minute?
- Are we tracking how customers found us, including AI assistants?
Which brings us back to the homeware shop and its unusually decisive new customers. The owner didn’t do anything special to earn those recommendations. Her product pages simply happened to say clearly what the bowls were made of, what they cost and where she delivered. As more customers start their shopping with a question to an assistant, that kind of plain, specific information may turn out to be one of the most valuable things a small business can publish.
Sources
- Adobe, “AI-driven traffic surges across industries”, Adobe Quarterly AI Traffic Report covering 1 November to 31 December 2025, based on more than one trillion visits to US retail sites.
