What you’ll learn in this article
- Points to check before rejoicing that “people from overseas suddenly increased” in your analytics
- How to tell bots and crawlers apart — something you can’t see from “headcount” alone
- The process of verifying a small sense of something being off, together with Claude, using data
Every time I open the Google Analytics (GA4) home screen, it feels a bit like checking answers. “Who in the world came to see my blog today?” It’s a small, secret moment of enjoyment.
But as I was gazing at it as usual, a number jumped out that made me do a double take. Wait — these are all from overseas!?
Today’s story is about that.
Previous: Reviewing my tags after the Redbubble Premium certification, is here
Active users by country (last 28 days)
*Excerpt of the top 5 countries, from the “by country” data on the GA4 home screen
Number one is of course Japan (64). So far, as expected.
But in second place was a number I never imagined. 45 people from Singapore. Compared with China (10) and the United States (6) in third place and below, it stood out by a head.
“Maybe Taisho Roman and Japanese-style illustrations resonate with people in Southeast Asia too…”
For a moment, my heart swelled with that expectation. This blog was launched in 2022, rested for about three years, and restarted in 2026. The trigger for the restart was my challenge of selling illustrations overseas. I felt as if that hard work had finally been rewarded.
Hey, Wait a Moment — Greenland’s “0”
But my elation was short-lived.
The GA4 home screen has a feature that displays country data on a world map. Happily, I moved my mouse around over the map.
As a test, I hovered over Greenland. The display that came up was “0.” Well, of course (lol). Nobody would come to see my blog from that far away.
But there, I suddenly came to my senses. “So, who exactly are these 45 people from Singapore?”
The second most, after Japan, is a small country in Southeast Asia. Thinking about it calmly, it’s a little strange.
“Are these really (humans) coming to see it?”
Anxiety crossed my mind. Still, since I was at it, I decided to dig into the numbers a little more carefully, together with Claude.
Digging Down to “City” with Claude
By the way, this isn’t the first time I’ve verified access numbers together with Claude. Before, there was the case of “My blog disappeared from Google search!? (in Japanese)” which I investigated together with AI. This time too, when in trouble, my first move is to consult.
GA4 has a report that lets you see access sources in units finer than country — “City.” I don’t usually look that closely, but while consulting Claude, I decided to check the data from this angle.
By the way, how to open this “City” report is summarized for beginners in a separate article: “How to see access by country and city in GA4.”
Then, looking at the place names lined up in the list, I couldn’t help exclaiming “Huh?” Mixed in with familiar, real city names like Matsuyama (Matsuyama) and Osaka (Osaka), for some reason “Singapore” was displayed as-is, with the country name.
If location information were captured at “city” level precision, the name of some district within Singapore (a unit like “XX Ward,” for example) should be displayed. Instead, it stopped at country-level granularity, “Singapore.” That alone was slightly suspicious material.
The Decisive Blow Was the Place Name “Prineville”
And comparing the numbers, my unease turned into conviction. What I focused on was the “average engagement time” — in other words, the average time spent on a page from opening it to leaving.
Average engagement time by city
*Prineville is a place name known for hosting large data centers of companies like Meta and Apple
Average engagement time by city
Access from Matsuyama was 1 minute 28 seconds, from Osaka 2 minutes 2 seconds. Both are natural numbers, like people properly reading the articles.
But Singapore’s average for all 45 people was “0 seconds” (hey now!). That means not a single second was spent before leaving the page. It’s hard to imagine 45 humans all doing that.
What clinched it further was another unfamiliar place name lined up right nearby in the list: “Prineville.”
Looking it up, this is a place name in Oregon, USA, known for hosting large data centers of companies like Meta (formerly Facebook) and Apple.
Access from a data center location, with a session duration of 0 seconds. With this much lining up, honestly, the answer was as good as decided.
Conclusion: These Were Bots
The true identity of the “45 people from Singapore” is most likely not human readers, but mechanical access by crawlers and bots — that was the conclusion I reached together with Claude.
A brief explainer | What is a bot?
■ What is a bot?
Not a human, but a program automatically coming to view a website. It’s a shortened name for “robot,” and it’s also called a “crawler.” There are “useful bots,” like Google going around collecting articles for search, and “nuisance bots” that probe for a site’s weaknesses.
■ Why do they mix into overseas access?
Bots operate from data centers (facilities where servers are lined up in rows) all over the world. Because of that, GA4 sometimes records them as if they were “people who came from that country.” This time’s Singapore and Prineville may also have been such data center locations.
■ Signs to suspect a bot
*GA4 has a mechanism to exclude well-known bots, but it can’t prevent everything. This happens routinely on many sites.
The basics of bots are also written in my previous article, “If your blog’s PV dropped and you’re anxious | Check bot contamination and Search Console” (in Japanese). This time is its overseas edition.
💬 If a word you don’t know comes up, feel free to ask via the chat at the bottom right of the screen.
Honestly, I was disappointed… My faint hope — that my work might have reached someone across the sea — missed completely. Still, the cause became clear and my unease disappeared, and I learned one “way of reading numbers” (I have to think of it that way).
What I Learned — “Engagement Time” over “Headcount”
What this incident impressed on me is that I mustn’t let my emotions swing just by looking at the number of accesses from overseas. The “headcount” number, precisely because it’s the easiest to see, may also be the number that’s easiest to be misled by.
This time’s learning points
- Don’t rejoice at overseas access numbers by “headcount” alone
- Check “engagement time” and “bounce rate” together
- If session duration is extremely short or close to 0 seconds, suspect bots
- Displays with coarse city-level granularity (still the country name, etc.) can also be a sign of bots
From now on, when I look at headcounts by country, I’ll make sure to check them together with “engagement time” and “bounce rate” (the percentage who leave immediately after opening a page). Even if the headcount is large, if the session duration is extremely short, that’s a sign to suspect bots.
Note that this time, the only one I could verify properly was Singapore. For access from other countries — China, the United States, Germany, and so on — I haven’t done the same check yet. Perhaps the same kind of bots are mixed in among them too, or conversely, there may be people who genuinely came to read (though the possibility is small…). I’ll check that again another time.
A morning that should have ended in elation unexpectedly became a day of studying GA4. If you’re someone whose heart dances at overseas access numbers like me, please take a look at the session durations together with them sometime.
The five overseas platform comparison roundup is here
Continued: The story of my FAA account suddenly being “closed,” is here
Comments