Webclat / Ecommerce
ecomm.webclat.com
Webclat / Ecommerce  /  agentic commerce

AI agents are shopping on your site. Here's what you're not seeing.

Some of your customers are no longer people clicking through a browser. Plain-English rundown of what that means for your reported sales and what a dashboard misses.

Quick answer

Some of the traffic buying from your site right now is not a person clicking through a browser. It is an AI shopping agent, acting on a customer's behalf, and it often completes the purchase without ever loading your website - which means the tag that reports your sales never runs. The money still lands in your bank. Your dashboard just does not know where it came from.

Landscape as of September 2026: this shifts monthly. OpenAI pulled its agent checkout model back from one-click purchase to recommend-and-redirect in March 2026; Google launched a competing approach the same quarter. Treat the mechanics below as accurate to right now, and expect the details to move again before the year is out.

The scale of it, in numbers you can check

Adobe Analytics measured a 4,700% year-over-year jump in AI-driven visits to US retail in 2025. By March 2026, that traffic was converting 42% better than everything else on the site - a reversal from a year earlier, when the same comparison ran 38% worse. Shopify, separately, reports AI-attributed orders grew 11 times over between January 2025 and March 2026. These are named, dated figures from the vendors' own published research, not our estimate.

Put plainly: the customers arriving this way are not a curiosity segment. On the numbers above they are converting better than your average visitor, and the volume is growing fast.

Why your reporting misses most of it

Standard web analytics (the tag that counts every visit and sale on your site today) works by running a small script in a visitor's browser. A shopping agent that reads your product listing and completes checkout through a direct request never opens that browser, so the script never runs, so nothing is counted. Industry estimates put the invisible share at roughly 70.6% of AI-referred traffic, with the visible remainder still undercounted by a factor of 3 to 4. [hypothesis, per named 2026 research - the exact share on your own site could read higher or lower; it is a direction to take seriously, not a figure to quote as your own without checking it].

Counted, but miscounted

An agent driving a real browser on a customer's behalf is tracked normally, but it behaves nothing like a person - instant decisions, no hesitation - and gets blended into your human numbers uncorrected.

Not counted at all

A purchase completed through a direct agent-to-merchant request never touches your website, so there is no page view, no session, and no sale in your dashboard - only the order in your fulfilment system.

Filtered out on purpose

Some bot-filtering tools quietly discard anything that looks automated, which can throw out a genuine high-intent agent along with the traffic it was actually built to stop.

We built the invisible order and watched it disappear

Rather than describe this, the Webclat demo store runs a real version of it. A working agent-checkout order was sent through end to end, and the standard analytics platform it reached (Google Analytics) came back with its own verdict: the order had no cookie to attach to and no session to join, so it could not be reported the normal way. The order itself landed correctly in a database under our control - the money is real, the record exists, it just never reached the report a person on the marketing team would open. See the full run, including a version driven from a command line with no browser whatsoever, at store.webclat.com/agentic-commerce.

Common questions

Are AI shopping agents actually buying things, or is this hype?

It is measured, not hype. Adobe Analytics recorded a 4,700% year-over-year increase in AI-driven visits to US retail sites in 2025, and by March 2026 that AI-driven traffic was converting 42% better than everything else - a full reversal from a year earlier, when the same comparison ran 38% worse. Shopify separately reports AI-attributed orders grew 11x between January 2025 and March 2026.

If it's already happening, why does my dashboard show almost none of it?

Because a shopping agent often completes a purchase through a direct request to your systems rather than by loading your website in a browser. Your analytics tag needs a browser to run in, so a purchase with no browser leaves no signal for the tag to fire. Research on this puts the invisible share at roughly 70.6% of AI-referred traffic - a directional estimate from named 2026 industry research, not a number to take as exact for your own site without checking it.

Is this worth doing anything about, or can it wait?

It depends on how much of your traffic is already agent-driven and whether that share is growing, which is exactly the thing a business cannot know without measuring it deliberately. Given the growth figures above, waiting means the invisible share keeps compounding while your reported numbers look unchanged - the risk is not that today's number is wrong, it's that you cannot yet tell how wrong it is.

Price the unmeasured slice on your own numbers.

Every order on the Webclat demo store is synthetic verification traffic against a fictional catalogue, run to show the mechanics honestly rather than to sell you a case study. To see what an unmeasured slice like this could be worth against your own traffic and spend, request an audit below.