The short answer: look at your visitors by city. Places like Council Bluffs (Iowa), Ashburn (Virginia), Boardman (Oregon) and Lanzhou (China) turning up day after day, with visits that last a second and an engagement rate near zero, are almost always automated. GA4 can't filter by city, so you either compare with them excluded in your reports, or use a tool that leaves them out for you.
Doesn't Google already filter bots?
Partly. GA4 automatically excludes known bots and spiders, using Google's own research and the IAB's International Spiders and Bots List. You can't turn that off, and you can't see how much was removed.
What gets through is everything not on the list: headless browsers, link-preview fetchers, uptime checkers, scrapers and AI crawlers that run JavaScript. Many run in cloud data centres, so GA4 places them in the towns where those data centres are. Council Bluffs and The Dalles host large Google data centres; Ashburn and Boardman host Amazon's.
The signs
In Reports, open Demographic details (under User attributes), switch the table to City, and look at each city's engagement rate. Bot traffic usually shows:
- Data-centre towns near the top for a site that has no reason to be popular there.
- Engagement rate near 0%. An engaged session, in Google's definition, lasts longer than 10 seconds, has a key event, or has two or more page views. Bots rarely do any of those.
- Very short visits, around a second.
- Persistence: the same places every day, often across several of your sites.
- Direct traffic or "(not set)" sources more than usual.
On a small site this can be a large share of all users, enough to make a quiet day look busy.
What you can do in GA4
- Compare without them. In a report, add a comparison (or in Explorations, a segment) that excludes the suspect cities. This changes the view, not the data.
- Data filters won't help. GA4's data filters cover internal traffic (by IP address), developer traffic and web hostnames, not cities or countries, and their effect is permanent.
- Unwanted referrals (up to 50 per data stream) stop a referring domain from starting new sessions. Useful for payment pages; not a bot filter.
- Block at your host. If you use Cloudflare or similar, its bot controls can stop some automated traffic before the Google tag ever runs.
Leaving them out automatically
GA Bridge does the comparison for you, across every site. Every hour it looks at the last 28 days by city, today included, and marks a place as a likely bot when its visits look hollow (almost never engaged and gone in a couple of seconds, or one page view and nothing else) and it keeps happening: on several days, on several of your sites, or in a burst. Those places are then left out of every number, live counts included, using GA's own figures minus the bots, and a line under your sites says how many sessions were left out.
Nothing is hidden silently: the Likely bots sheet lists each place with its evidence ("0% engaged · 12 days · 3 sites · data centre"), and you can mark any of them Not a bot. If a real person in Council Bluffs reads your page, the city isn't hidden.
Questions
Why do I get visitors from Council Bluffs in Google Analytics?
Council Bluffs, Iowa, hosts a large Google data centre, so automated traffic running there (crawlers, link-preview fetchers, headless browsers) is located there. Visits from it that last a second with no engagement are almost always bots.
Does GA4 filter bot traffic?
It automatically excludes known bots and spiders on Google's and the IAB's lists, and that can't be turned off. Bots not on those lists still get through.
Can I exclude a city from Google Analytics 4?
Not with a data filter: GA4's data filters cover internal traffic, developer traffic and hostnames. You can exclude cities from a report's view with a comparison, or in Explorations with a segment.
How does GA Bridge decide what's a bot?
Every hour it looks at 28 days of activity by city across all your sites, today included. A place counts as a likely bot when its visits are almost never engaged and last under a couple of seconds, or are one page view and nothing else, and it recurs on several days or sites or comes in a burst. Known data-centre towns such as Ashburn need no repeat. You can see the evidence and correct it.
