Half Your Website Traffic Isn't Human
I went through the analytics on the sites I manage a few weeks ago expecting a tidying job. On one of them, 41% of the recorded traffic turned out to be coming from my own development machine. The tracking script was firing every time I opened a page to work on it, and every one of those was being counted as a visitor.
That was my own bug and I have fixed it. But cleaning up the rest of the data turned up something more general: across the sites I look after, roughly one session in five came from datacentres and scraping operations rather than from anybody who might ever buy anything.
If you have looked at your visitor numbers and wondered why they never turn into enquiries, this is part of the answer.
The Traffic That Is Not a Customer
Four kinds, in rough order of how much they distort things.
Crawlers and scrapers. Search engine bots, SEO tools, AI training crawlers, price scrapers, and a long tail of software that reads websites for reasons of its own. The well-behaved ones identify themselves. Plenty do not.
Datacentre traffic. Sessions that look like an ordinary browser but originate from cloud hosting rather than from a home or a phone. If your analytics shows visitors from Council Bluffs, Ashburn, Boardman or Falkenstein, those are not towns your customers live in. They are where the large cloud regions are.
You and your own team. Every time you check the site, an employee looks something up, or your web person opens a page to edit it. On a small site this is easily the biggest single source of fake traffic, precisely because the real numbers are small enough for it to matter.
Referral spam. Fake referrals designed to get you to visit the referring site out of curiosity. Less common than it was, still around.
How to Spot It
You do not need special tools for this. You need to look at the shape of the data.
- Geography that makes no sense. A Shoreham hairdresser with a steady stream of visitors from Virginia, Iowa and Guangdong is not going international.
- Zero-second sessions with a single pageview. A real person who lands and leaves still takes a few seconds. A run of sessions with no duration at all is usually automated.
- Traffic that is too even. Humans have a daily rhythm. If your hourly chart is flat through the night, something is visiting on a timer.
- Spikes with no cause. No campaign, no post, no mention, and a hundred extra sessions on a Tuesday.
- Odd browser and device combinations. Very old browser versions, or screen sizes that do not match anything on sale.
- Pages nobody should be reading. Repeated hits on files and URLs that have never been linked from anywhere.
Why Your Analytics Does Not Filter It For You
It filters some. Google Analytics excludes traffic from a published list of known bots and datacentres, and that catches the honest ones. What no list can catch is software that identifies itself as an ordinary browser, and a growing share does exactly that, sometimes routed through residential connections so the geography looks plausible too.
Privacy-focused and self-hosted alternatives generally filter less, not more, because they are deliberately doing less profiling. That is a fair trade, but it means the raw number is dirtier than it looks.
You Are Probably in Your Own Numbers
This is the one people miss, and it is the one that caught me.
The obvious version is you and your staff visiting the site. The less obvious versions are worse: a developer with the site open all day while working on it, a preview or staging copy carrying the same tracking code, or a monitoring service loading a page every minute.
Mine was the first of those. The tracking tag had no restriction on which domain it was allowed to run on, so it fired happily from a development server on my own laptop and reported it as real traffic to the live site.
What to do about it:
- Restrict your tracking to your real domain, if your tool supports it. This is the fix that holds, because it works no matter whose machine the page is opened on.
- Exclude your own network by IP address as a second line.
- Keep tracking out of staging and preview sites entirely.
- If you have had this problem, consider deleting the historic bad data rather than living with it. A clean short history is more useful than a long misleading one.
What It Costs You to Believe the Wrong Number
This is not only a vanity problem.
You conclude a page is popular and write three more like it, when the visits were a crawler. You conclude the website is working because traffic is up, and stop asking why the phone is not ringing. You compare this month to last month and read a change in bot activity as a change in customer interest. Then you set your marketing budget on the basis of it.
The direction of the error is always the same. Fake traffic inflates the total, so the site always looks like it is doing better than it is.
What an Honest Number Looks Like
Small businesses are routinely told they should be getting thousands of visitors. For most local trades, that is not what success looks like.
A plumber covering three towns might get a few dozen genuine visits a week and be doing perfectly well, because a meaningful share of them ring up. A quiet site where one visitor in twenty gets in touch beats a busy one where nobody does, and the busy one is often busy with robots.
The number that matters is not sessions. It is how many people took the action you built the site for. Everything above that is context.
Measure the Thing You Actually Want
Most small business analytics setups count arrivals and stop there, which leaves the important part invisible. A contact page with thirty views and three enquiries is a completely different situation from thirty views and none, and by default you cannot tell them apart.
Worth tracking as events:
- Contact form submissions, recorded on success rather than on the click.
- Clicks on your phone number.
- Clicks on your email address.
- Directions and map clicks.
- Booking link clicks.
- WhatsApp or messaging clicks.
Six events, an hour of setup on most platforms, and it turns a traffic report into something you can make decisions from. It also gives you a sanity check on the traffic itself, because bots overwhelmingly do not fill in forms.
An Hour of Cleaning Up
In order:
- Look at your traffic by country and city, and ask which of those places contain plausible customers.
- Filter out the obvious datacentre and overseas traffic if your tool allows it, and note what you removed rather than removing it silently.
- Restrict your tracking script to your live domain.
- Exclude your own IP address, and keep tracking off any staging site.
- Set up the handful of conversion events above.
- Re-read last quarter with all of that applied, and be ready for the real figure to be a lot smaller.
The smaller number is the useful one. You can work with it.
The Bottom Line
Nearly every website report I have looked at overstates its audience, usually by a lot, and my own was no exception. That is not a reason to stop measuring. It is a reason to measure the thing you care about and treat the headline number with some suspicion.
If your site is quiet, the traffic count is rarely where the problem is, and neither is your PageSpeed score. The place to look first is usually your Google Business Profile and whether the site answers the question people arrived with.
