I sent an AI agent and a scripted browser to the same online store. They looked nothing alike.

by•

Today I ran a small test on a demo store. First I browsed it myself. Then I asked Claude (through its Chrome extension) to look around and pick a product. Then I ran a basic Playwright script on the same pages.

The scripted browser was easy to spot. It announces itself as automated through the navigator. webdriver, and plenty of sites catch that already.

The agent was the interesting one. It ran inside a normal Chrome with a real profile, so on the surface it looked like a person. The difference only showed up in how it used the page: how it moved, typed and clicked.

That got me thinking about analytics. If agents run in real browsers, most tools will count them as human visitors. Your traffic, bounce rate and conversion rate quietly start including software that shops on someone's behalf.

So I'm curious how others see it:

- Would you want agent visits counted as users (a real buyer sent them) or separated out?

- Has anyone noticed odd sessions in their analytics that turned out to be an AI agent?

8 views

Add a comment

Replies

Best

I'd count them separately, not drop them. A buyer who sends an agent is still demand, but the funnel is different: an agent doesn't need your hero section, it needs prices and stock it can read. A flag, not a filter.