Your script tag is half an install

Every analytics install in the last fifteen years has been one step: paste a script tag. It works because the audience was people, and people run JavaScript.

The audience changed. The install did not.

The part that cannot be fixed

GPTBot will never execute your snippet. Neither will ClaudeBot, PerplexityBot, CCBot, Bytespider or any of the others. They fetch HTML, take what they came for and leave. There is no runtime, no document, no sendBeacon.

This is not a gap that better client-side engineering closes. It is a property of what those clients are. Any analytics product whose only intake is a browser script has a permanent blind spot exactly where the interesting question now lives, and no amount of work on the tracker changes it.

The part that was never missing

Your web server wrote every one of those fetches down. It has been doing it the whole time. The timestamp, the address, the method, the path, the status, the bytes, the response time and (the field everything depends on) the user agent.

The log has always held the other half of your audience. It just was not connected to anything that reported on it.

So the install is two halves

The script tag reports the people: engagement, scroll depth, client-side navigation, Core Web Vitals. A server cannot see any of that.

The log reports the machines, which agent, which operator, what for, from what address, with what status code. A browser cannot see any of that.

Neither half is optional if you want the whole picture, and neither half can be derived from the other. In Micaforge that is one script tag and one of:

npx --package=@micaforge/sdk-server micaforge-shipper /var/log/nginx/access.log

or a middleware in your Node app, which skips the file entirely.

What you find on day one

Usually two things.

The first is scale. On a documentation site the machine half is frequently comparable to the human half, and occasionally larger. It has been that way for a while; it was simply not being counted.

The second is distribution. The crawl is rarely spread the way your traffic is. A handful of pages get fetched constantly and some of your best writing has never been crawled at all, which is a fixable problem, and one you cannot see from a browser.

Install the second half the same day as the first. Otherwise you have installed a tool that measures the audience you already knew about.