The bots situation is crazy

In one of my previous articles, I said how I migrated from plausible to Liwan to Umami. But in the spirit of vibe-coding, I have decided to ditch Umami to a custom-made analytics. In Rust. It runs on DuckDB, uses 130MB of RAM, and ingests non-stop my caddy access log. So now, I get to see not only JS verified visitors, but also all the bots (in addition to RSS subscribers). Neat!

I turned on rotating Caddy logs around June 2026, when I was playing with GoAccess, and since then I have recorded 697,991 bots. These are obviously filtered out of the “organic” visitors, but it’s still an interesting number to see that was not available to me in Plausible or Umami. As with any new toy, you get excited in the beginning, so I was opening it every 15 minutes to just take a look. And when you look at something often, you start to notice some patterns.

For the past 30 days, I have noticed that in addition to the regular direct and search engine sources (hello people from kagi.com), one website occupied the top 5 most referred sources. In fact, it was higher than Reddit and HackerNews. It first appeared in September. Then a few hits in beginning to mid-October, when eventually having a big burst at the end of October, going quiet, and now, picking up pace again starting November 5th. I run Umami in parallel while checking how accurate my vibe-coded analytics, and the source appears there as well, so it’s not a fluke in my analytics.

Now, look. This blog is very diverse. I write on various topics ranging from Software engineering to Philosophy, to Travel — so links to my posts can circulate anywhere. But clearly not on the American Osteopathic College of Radiology?!

So I decided to dig deeper. Perhaps radiology people like my content? I am all in, everyone is welcome! Since these are all JS verified visits (i.e. appearing both in the access.log and sending an anonymous js-beacon), I started to dig. The top 5 visitor countries (at least if you trust free GeoIP database) are: the US (with 83% share), Singapore, Canada, India, and Mexico (together they amount to 92% of all visitors). The rest 8% are a bunch of countries that include various EU countries as well as countries like Saudi Arabia, Nigeria, Azerbaijan, and Honduras (among many others). But, hey I don’t blame anyone. No reason for a Nigerian NOT to read the aocr. Or perhaps my GeoIP database is outdated.

When analyzing traffic, especially from a weird website like aocr that seems to have no connection to my blog, you expect a pattern of a single article. Perhaps someone shared one of my posts in their internal forum or something. Maybe American radiologists want to go to Romania and climb a mountain, or they simply want to remember that we are all mortals. But no, the analytics shows that all of them land on at-least 30 different pages. So I looked at the browser: Chrome. No surprise, Chrome is by far the most popular browser, although I prefer Firefox. Then the OS: Windows. Again, expected for doctors. They rarely run Linux or Mac. Lastly, the viewport: Tablet (classified as 800x600 by Umami, and between 576 and 992 in width by my analytics).

Long story short, these are bots. Bots that run JavaScript, fake their user agent, and use a “real” referrer. Why? I don’t know. I mean, I could guess is that they don’t want to be classified as bots, so people like me would think that American radiologists enjoy my blog posts. But it kind of is sad, and makes me somewhat disgusted.

Take form it whatever you want.