The research an agent does before you ever open a search box.
A growing share of product research now happens inside an ongoing conversation with an AI agent someone already trusts, rather than a fresh search each time. Nobody has a precise figure for how much yet, and any business that says otherwise is guessing. The direction, though, is real enough to plan around.
6 min read
The shift, stated carefully
Someone shopping for a new mattress a few years ago would have opened a search engine, typed something, scrolled a page of results, clicked into a few. Increasingly, that same person opens a conversation with an assistant they've used before, one that already knows they have a bad shoulder and a small bedroom, and asks it directly. Neither a search box nor a page of results gets opened. The research still happens, it just happens somewhere a retailer's analytics were never built to see.
This isn't a claim that search is disappearing. It's narrower and more defensible: for a growing category of purchases, the first research step is shifting from a search engine to a standing relationship with an assistant, one a person returns to rather than starting fresh with each time. How large that shift is, and how fast, isn't something anyone can state precisely yet. That it's happening at all is visible enough to take seriously.
Why this matters more than the exact number
A retailer waiting for a precise percentage before acting is asking the wrong question. What matters practically is that a meaningful, growing slice of research is happening somewhere a product page was never designed to be checked: inside a conversation, against facts the assistant already holds about the person asking, not against a page it's scrolling in real time. A retailer whose product data only works when someone's actively looking at a page is invisible to exactly the research that's growing.
What doesn't change underneath the shift
The mattress buyer with a bad shoulder and a small bedroom still needs the same specific facts either way: firmness rating, dimensions that fit the room, a trial period that's genuinely enforceable. Whether those facts get checked by someone scrolling a page or by an assistant answering from memory of an earlier conversation, the facts themselves don't change. What changes is where and how they get checked, and a retailer whose data only lives on a page built for scrolling loses exactly the research that's moving away from scrolling.
Where this shows up for a business
A retailer's facts become available to an agent directly through Selfe, not only through a page waiting to be found and read, so a purchase gets considered whether the research happens through a search result or through a standing conversation with an assistant someone already trusts.
Does this mean our website traffic will drop?
Possibly, for some categories, but a drop in page views isn't the same as a drop in genuine interest. It may just mean the research moved somewhere your analytics can't see.
How would we even measure this?
Honestly, not easily yet, which is exactly why the scale of the shift is treated as an open question here rather than a settled statistic.
The next twelve months in agentic commerce.
Most writing on where AI agents are heading falls into one of two camps: everything changes overnight, or nothing really changes at all. Neither is honest. Some things about agentic commerce are already settled enough to build for. Others genuinely aren't decided yet, and pretending otherwise doesn't help anyone plan.
What changes for discovery in the next twelve months.
Not a sweeping prediction. Over the next year, the practical change for discovery is narrower and more useful than "everything will be different": more of the requests reaching a business will already carry real constraints attached, rather than arriving as a vague, browsable query.
What happens to reviews once agents are the ones reading them.
A star rating tells a person something useful at a glance. It tells an AI agent almost nothing it can act on. As more of a property's reviews get read by agents rather than skimmed by people, what makes a review useful is shifting, and a 4.6 average isn't going to be the part that matters.