Bots read my blog more than people do

A week ago I wired a small home-grown analytics module into this blog. I wanted to answer something basic: is anyone visiting at all? The first numbers turned out more interesting than I expected — and not in the way I thought.

Over seven days the site got 214 requests:

• 92 from real people (66 unique visitors);
• 122 from bots.

So 57% of my blog's traffic is machines. They show up more often than humans, and the gap wasn't a fluke — it held almost every day.

Who actually came

The full list for the week:

Three things stood out.

1. AI crawlers found the site almost immediately

ClaudeBot — 13 visits, GPTBot — 8. Together that's 21 requests, roughly one in six bot visits.

The site is a week old. It has no backlinks, I announced it nowhere and submitted it to no directories. Anthropic's and OpenAI's crawlers came anyway and read everything that was published.

It used to take months of waiting to be indexed. Now there's a second distribution channel for content, and it appears to move faster. Note that none of these visits show up as "visitors" in normal analytics: someone who asks ChatGPT and gets an answer sourced from my article may never land on my site at all.

2. Almost half the bots give nothing back

Let's count: MJ12bot (20) + SemrushBot (17) + PetalBot (2) + KBCnetContactBot (6) + LohiSoftBot (12) = 57 visits. That's 47% of all bot traffic.

What do I get from it? Nothing. These are SEO aggregators collecting data to sell inside their own products, plus contact scrapers of dubious value. They won't bring me a reader and won't put me in search results — they just take.

For a small blog it doesn't matter: a few dozen requests a week won't hurt anyone. But on a larger site this becomes a bandwidth line item, and you start to see why people block such bots in robots.txt.

3. People arrive directly, not from search

Sources for human visits:

• direct — 74
• internal navigation — 15
• search — 3

Three visits from search in a week. Which is completely normal: a new site, search engines have only just started indexing it, there are no rankings yet.

What it does show is where an audience really comes from at the start: people you handed the link to yourself. There's no magic in "write an article and traffic will come" — not in the first months, anyway.

What this means in practice

Don't measure only humans. Looking at "visitors" alone, the first week is dull: 66 people. But in parallel the site has already been read by search engines and by two major language models. Those are different channels, and the second one is invisible to ordinary counters.

Your content must be in the HTML. I wrote about this separately: AI crawlers don't execute JavaScript. If my blog rendered on the client, ClaudeBot and GPTBot would have carried away an empty page, and those 21 visits would have meant nothing.

Decide about bots deliberately. I left everything open: the blog exists to be read, and if a model cites my article, good. But now I at least know who's visiting, and I can block the pure takers whenever I want.

Caveats

Honestly about the limits of these numbers: this is one week of one small personal blog. The sample is tiny and you shouldn't extrapolate to large sites. Bots are identified by User-Agent, which can be spoofed, so some of the "humans" could in theory be disguised crawlers. I excluded my own visits from the stats, but a few of my curl checks did land in the data — they're the "other" row.

Still, the overall picture seems stable: on a new site there are more machines than people, and among the machines a growing share reads not for search, but for training and answers.

I'll come back to these numbers in a couple of months — I'm curious whether the ratio shifts once search traffic appears.