Your Next Customer Might Be an AI, Not a Person

My sister told me recently that she'd stopped comparing prices herself before buying anything over a certain amount. She just asks an AI assistant to do it, pull up a few options, summarize the reviews, tell her which one's actually worth the extra money. She still clicks the final buy button herself, for now, but everything leading up to that moment, the part that used to take her twenty browser tabs and an hour of her evening, now takes about ninety seconds and a conversation.

That shift has a name now, agentic commerce, and the numbers behind it are big enough that I think anyone selling anything online should understand what's actually happening, because it's not a small trend at the edges. It's already reshaping how a large share of people shop.

How far this has actually gone

The scale surprised me. Recent research puts the number of consumers already using some form of AI in their shopping process at around seventy three percent, not people who might try it someday, people already doing it now. Most of that is still what you'd expect, getting product ideas, having reviews summarized so they don't have to read forty of them, comparing prices across a few sites. Only about thirteen percent have gone all the way to completing an actual purchase after being pointed there by an AI assistant. But seventy percent say they're at least somewhat comfortable with the idea of an AI agent making a purchase on their behalf without them personally clicking through every step.

That gap, lots of people using AI for research, comparatively few letting it complete the purchase, tells you where things stand right now. The research phase has already shifted. The final click hasn't, not yet, but the comfort level suggests it's not far behind.

Why this is a genuinely new problem for sellers

Here's the part that I think matters more than the shopping convenience angle. For years, understanding your customer meant tracking their journey, what they searched, what page they landed on, what they compared, before they bought. That data lived in your analytics, mostly. If an AI assistant is doing the discovery, the comparing, and the reasoning inside a chat conversation somewhere else entirely, all of that behavioral data just doesn't reach you anymore. The person shows up at your checkout page having already made up their mind somewhere you can't see, and your usual tools have no record of how they got there.

There's also a real fraud concern that's come up alongside all this. A large share of financial institutions are reportedly expecting fraud attempts to increase specifically because of AI shopping agents, since an automated system moving through checkout flows quickly and repeatedly looks, on the surface, a lot like the exact pattern fraud detection systems were built to catch. Sorting out which of those automated visits are a legitimate assistant shopping for a real person, versus something malicious, is a genuinely unsolved problem industry-wide right now.

The technical piece worth knowing about

In January 2026, Google introduced something called the Universal Commerce Protocol, which is essentially an attempt to give AI agents a standard, agreed-upon way to read a merchant's product catalog and complete a purchase, rather than every AI company building its own custom way of scraping and interacting with every store's website. If that sounds familiar, it's a similar idea to how MCP standardized AI tools talking to apps like Gmail or Slack, just applied to shopping instead. Whether this becomes the actual standard everyone adopts or one of several competing approaches is still playing out, but the direction is clear: stores are going to need to be readable by an AI agent, not just a human browsing on a screen, and that's a fairly different design requirement than most small online stores have ever thought about.

There's real money already moving through this. Spending that starts with an AI platform's involvement is projected to be worth close to twenty one billion dollars in 2026, nearly four times what it was the year before, and some longer-range estimates put the total retail spend that could eventually shift toward this model in the trillions over the next several years.

What I'd actually pay attention to if I sold anything online

I don't think this means panicking or rebuilding your entire store overnight. But a few things seem worth doing now rather than waiting. Product information needs to be clear and structured, not buried in marketing language, because an AI agent summarizing your product for a customer is going to lean on whatever text is actually on the page, and vague, hype-heavy descriptions don't translate well into an honest comparison. Reviews matter even more than they used to, since AI assistants lean heavily on summarizing them, and a product with thin or clearly fake reviews is going to get flagged or skipped in a way a human browsing casually might not have noticed.

And it's worth just sitting with the idea that some meaningful share of your future traffic might not be a person browsing your site at all in the traditional sense. It might be something acting on a person's behalf, reading your page in a completely different way than a human would, deciding in seconds whether you're worth recommending. That's a strange thing to plan around, and I don't think anyone has it fully figured out yet, including the platforms building the tools. But given how fast the comfort level with this has grown, "figuring it out later" is starting to look like a more expensive choice than it did even a year ago.