GA4 Bot Traffic: How One Visitor Made 88% of Our Sessions
From October 1, 2026, one bot in Singapore with a 1280x1200 screen made 88.1% of our GA4 sessions and 91.7% of Direct. How we spotted it, filtered it and stopped counting it.

In short: Before you celebrate a traffic jump, break it down by country, screen resolution and engagement. One country plus one odd resolution plus zero engagement is a bot. Build a segment that removes it from your reports, keep your tag out of automated browsers, and report growth from engaged sessions rather than raw sessions.
On this page12
Overview
In October 2026, our analytics looked like a breakthrough. Google Analytics 4 showed 15.8 times as many sessions on vanaxity.com as in September, almost all of them Direct.
It wasn't a breakthrough. One automated visitor, a bot in Singapore with a 1280x1200 screen, made 88.1% of our sessions from October 1 to November 15. This article shows how we found it, how we took it out of our reports and how we stopped counting it, with the numbers from our own GA4 property.
Disclosure: Vanaxity builds AI agents for marketing, and this is our own site's data, shared because the pattern is common. This article was drafted with help from Claude. The numbers come from the GA4 Data API, and the counts behind each chart are in our data file.
Key takeaways
- From October 1 to November 15, 2026, one bot made 1,312 sessions on vanaxity.com, 88.1% of all sessions and 91.7% of Direct sessions.
- Every session came from Singapore on a 1280x1200 screen, in Chrome on Windows, from a new user who viewed one page and was never engaged.
- It turned a 1.7-fold rise from September to October into a 15.8-fold one.
- GA4's built-in bot exclusion missed it, and GA4's data filters can't filter by country or screen resolution.
- An exploration segment removes it from reports. A one-line check on navigator.webdriver keeps the tag out of automated browsers.
Book a free fit checkMap your SEO, GEO and AEO workflow before you build.
Book a consultationThe growth that wasn't
GA4 put October's sessions at 15.8 times September's. Direct traffic led the rise, and Direct is the channel people read as brand demand: someone typed the address or used a bookmark.
When we took out the sessions with one country and one screen resolution, October was 1.7 times September. Still a rise, but a different story.
October 2026 sessions on vanaxity.com as a multiple of September's
Times September's sessions
Across the whole window, October 1 to November 15, the bot made 1,312 sessions: 88.1% of all sessions and 91.7% of the sessions GA4 filed under Direct. Its share barely moved, at 89.3% of October's sessions and 85.6% of those from November 1 to 15.
Share of vanaxity.com's GA4 sessions made by the bot
Share of sessions
The bot's fingerprint
The bot was easy to see once we looked past the totals. Every one of its 1,312 sessions had the same traits:
| Signal | The bot | Everyone else |
|---|---|---|
| Country | Singapore, every session | Many countries |
| Screen resolution | 1280x1200, every session | Many sizes |
| Browser and system | Chrome on Windows, desktop | Mixed |
| Channel | Direct, every session | Direct, social, referral and search |
| New or returning | New, every session | Both |
| Pages per session | 1 | 1.8 |
| Engaged sessions | 0% | 31.5% |
| Events | first_visit, page_view and session_start only | Scrolls, clicks and engagement too |
It didn't hammer one page. Its 1,312 sessions began on 811 different pages, and the most common first page was the Insights hub, in English (211 sessions) and Vietnamese (34). That is how a crawler moves: one page, the next link, a new visitor each time.
It also kept no hours. Every hour of the day had between 41 and 80 of its sessions, with none of the dip at night that real readers show.
The pair of traits is tight. In the same weeks, only 14 sessions from other countries used a 1280x1200 screen, and only 14 sessions from Singapore used another screen. Filtering on both together removes the bot and almost nobody else.
Other site owners have reported bot sessions from China and Singapore, filed under Direct and with no engagement, since October 2025. Delante described the pattern on December 30, 2025. Nine of the 14 other 1280x1200 sessions on our site came from China.
When it started
Our GA4 data starts on June 18, 2026. The bot's first sessions came on October 1, 2026, and it has visited every week since:
The bot's sessions on vanaxity.com per week
Sessions
| Week | Bot sessions |
|---|---|
| October 1 to 4 | 66 |
| October 5 to 11 | 69 |
| October 12 to 18 | 99 |
| October 19 to 25 | 504 |
| October 26 to November 1 | 155 |
| November 2 to 8 | 143 |
| November 9 to 15 | 276 |
The busiest stretch was October 19 to 21, with 490 sessions in three days. Then it dropped to a handful a day before climbing again. A start date this sharp is the first thing to look for: real audiences grow over weeks, while bots switch on.
Why GA4 didn't filter it
GA4 already removes known bots. Google says it excludes traffic from known bots and spiders using its own research and the IAB's International Spiders and Bots List, and that you can't turn this off or see how much it removed. This bot wasn't on the list, or didn't look like anything on it.
Nor can you add your own rule at collection time. GA4's data filters cover three cases: developer traffic, internal traffic by IP address, and web hostnames. None uses country or screen resolution, and the bot's hits came from vanaxity.com itself, so a hostname filter doesn't help either.
That leaves two places to act: in your reports, and in the tag itself.
How to find and filter it in your reports
- In GA4, open Explore and start a free-form exploration.
- Add Country and Screen resolution as rows, and Sessions and Engaged sessions as values.
- Look for one country and one resolution with many sessions and an engagement rate near zero. Ours was Singapore and 1280x1200.
- Add Date as a row for that pair. A sharp start date confirms it.
- Build a session segment that excludes sessions matching both values, and apply it when you report traffic.
- Compare Direct sessions with and without the segment, and report the second number.
Segments only apply in explorations, so the standard reports still include the bot. In those, add a comparison that excludes the country, or report engaged sessions, which the bot never produced.
How we stopped counting it
Our tag doesn't load GA4 on page load. It waits for the visitor's first tap, click, key press or scroll, or for 15 seconds, whichever comes first. The bot's sessions show no engagement at all, which fits a browser that opens a page and waits for the timer.
Browsers driven by automation tools, such as Playwright, Puppeteer or Selenium through WebDriver, report it by default: navigator.webdriver is true. So in November 2026 we added one check, and the tag now stops before it loads anything in such a browser:
(function () {
if (navigator.webdriver) return; // automated browser: load no analytics
var loaded = false;
function load() {
if (loaded) return;
loaded = true;
var s = document.createElement("script");
s.async = true;
s.src = "https://www.googletagmanager.com/gtag/js?id=G-XXXXXXXXXX";
document.head.appendChild(s);
}
["pointerdown", "keydown", "scroll"].forEach(function (type) {
addEventListener(type, load, { once: true, passive: true });
});
setTimeout(load, 15000);
})();We tested it in Chrome under Playwright, with every request to Google blocked so no test reached our property. As is, with navigator.webdriver true, the page queued no GA4 configuration and made no request to Google, after a scroll or after 17 seconds of waiting. With the flag hidden, as in a reader's browser, GA4 loaded both ways.
A bot can hide the flag, so this check is a filter, not a wall. If this one does, the segment above still takes it out of our reports. The bot can still read the site, and that's fine: the pages are public. We just don't want it in our numbers.
What to report instead
- Engaged sessions, not sessions. The bot made 1,312 sessions and zero engaged ones.
- Direct traffic with the segment applied. Raw Direct was 91.7% bot for us.
- New users only with a bot check. Every one of the bot's sessions was a new user.
- Growth from a date range that starts before any spike, with the spike explained.
Bots read server logs differently from analytics. For a look at which crawlers really read a site and how many fake their names, see our week of crawler logs.
What this data can't tell you
- Who runs the bot. GA4 shows no addresses, and nothing in the sessions names an operator.
- Whether every session from Singapore at 1280x1200 is the bot. A real reader with that screen in Singapore would be filtered out too, though the numbers suggest very few exist.
- Whether the webdriver check catches it. If the bot hides the flag, only the segment will.
- How it compares with your site. Our site is small, so one bot moves our totals far more than it would move a busy site's.
GA4 bot traffic FAQ
Why is my GA4 Direct traffic suddenly so high?
Often a bot. Break Direct sessions down by country, screen resolution and engagement rate. On our site, one bot from Singapore with a 1280x1200 screen made 91.7% of Direct sessions from October 1 to November 15, 2026.
Is traffic from Singapore with a 1280x1200 screen a bot?
On our site, yes: 1,312 sessions with zero engagement, one page each and only automatic events. Check your own data for the same pattern before you filter, since real readers can share either trait alone.
Can I block bot traffic in GA4?
Not by country or screen resolution. GA4's data filters cover developer traffic, internal traffic by IP address and web hostnames. Use an exploration segment for reports, and keep your tag out of automated browsers.
Does GA4 filter bots automatically?
It excludes known bots and spiders using Google's research and the IAB's International Spiders and Bots List. You can't turn this off or see how much it removed, and it missed the bot on our site.
What is navigator.webdriver?
A browser property that is true when automation software, such as Playwright, Puppeteer or Selenium, controls the browser. A tag can check it and skip loading analytics, though a bot can hide it.
Sources and data
The counts behind this article are in our data file, read from vanaxity.com's GA4 property through the GA4 Data API.

