This is example research content. Replace it with your own write-up.
Every analytics tool counts page views. Almost none of them tell you how many belong to a real person. Over six weeks I logged user agent, screen size, language and referrer for every hit on this site, then sorted them by hand.
Method
- One site, static HTML, no client-side router.
- Analytics events kept raw for six weeks, then exported as CSV.
- A session counted as automated when it had a headless browser fingerprint: fixed viewport, one page per session, no referrer, no pointer movement.
What came out
Roughly two thirds of the raw page views were not people. The bulk came from preview crawlers and screenshot services, all reporting a normal desktop Chrome user agent, all with square viewports like 1280×1200.
Once those were removed, the remaining traffic was boring and believable: search referrals, a handful of newsletter clicks, and direct visits that mostly landed on the one page people bookmark.
What I changed
Nothing on the site. The fix belongs in reporting, not in the browser: filter at query time, keep the raw data, and never build a dashboard on top of an unfiltered count.