AI traffic

The visits analytics never records.

AI engines read your website every day, and your analytics has no idea. They do not run JavaScript, so they are never counted. Reflexa counts them where they actually arrive — separating a page opened to answer somebody right then from one merely crawled to learn, page by page and engine by engine.

By Reflexa Technologies — the team building Reflexa, the agent that works on your site · October 8, 2026

Analytics was built for people

Google Analytics counts a visit when a browser runs its JavaScript. An AI engine fetching your page to answer somebody's question does not run it. So the single most interesting thing happening on your website right now — engines reading your pages to answer real buyers — is invisible in the tool you already have.

The agent counts those requests at the source: it sees them arrive, reads who is asking, and files each one.

Three different things, never added together

They are kept apart because they mean opposite things. A page crawled a hundred times and never opened to answer anyone is not a success — it is a page AI knows and does not find useful. Blending them into one "AI traffic" number would hide exactly the thing worth knowing.

Who gets counted

ChatGPT, Claude, Perplexity, Gemini and Copilot each get a counter of their own. Amazonbot, Applebot, Bytespider, CCBot, cohere-ai, DuckAssistBot, YouBot and Diffbot are counted together as one group. Everything is matched on the user agent the request arrives with and the referrer a person arrives from — observed, in your own logs, not modelled.

Page by page is the point

A site-wide total tells you almost nothing. The dashboard groups your pages by what AI actually does with them: the ones it answers with, the ones it has crawled but never used, and the ones it has never touched at all. That last group is usually the biggest, and it is the one that turns into work — because a page AI has never seen is a different problem from a page it has seen and passed over.

What it is not

An engine reading your page is not an engine recommending your company. They are different events and we never present one as the other. This counts reads, crawls and click-throughs — facts from your own server — and says nothing about what any answer said.

If no counter is connected, the dashboard says it cannot see rather than showing you a zero. An absent counter is not a measurement of nothing.

Read my site →

Questions people ask.

How is this different from server logs?

It is server logs, read properly: classified per engine, split by why the request was made, attributed to the page, and kept day by day. The agent does it for you, or a few lines in your CDN do it if you have no agent.

Can bots fake being ChatGPT?

A user agent can be spoofed by anyone. That is why the three kinds are kept apart and why we never turn the count into a claim about recommendations — it is a count of requests to your server, described as exactly that.

Why does a page get crawled but never used?

Usually because the engine can reach it but cannot lift anything cleanly from it — the text only appears after JavaScript runs, or the page never opens with the answer. Those are two of the 23 checks.

Is AI traffic on the free plan?

No. The free plan is the structural check of your pages. Counting AI visits needs the agent on the site, which is part of Starter and Growth.