Bot and Spam Traffic, and What Filtering Actually Removes

"Are these real people" is the right question, and the honest answer is that some of them are not, that automatic filtering removes a specific subset, and that the subset it removes is usually not the one distorting your report.

Here is what each category actually is, because the fix is different for each and lumping them together is why this problem persists.

Four different things, commonly called the same thing

1. Known crawlers and bots

Search engine crawlers and other identified automated agents. These are numerous and they are the category platforms handle for you: measurement tools maintain lists of known agents and exclude them from reporting.

This is what automatic bot filtering covers, and it works well for what it covers. It is also the category least likely to be the thing bothering you, because it is already being removed.

2. Referral spam

Sessions injected into your reporting by parties who never visited your site, to get a domain name in front of somebody looking at their own report. It shows as referral traffic from domains you do not recognize, frequently with an unusually high or unusually low engagement pattern.

Automatic bot filtering does not necessarily catch this, because it is not a crawler visiting your site; it is data being sent to your property.

3. Automated traffic that is not on any list

Scrapers, monitoring services, uptime checkers, security scanners, competitors' tools, and anything custom. These behave like a browser, are not in any published list, and are recorded as ordinary visits.

This is the category filtering handles worst, and it is frequently the one producing an inexplicable segment.

4. Your own team

Not bots at all and easily the most common distortion in a small account.

You, your colleagues, your agency, your developer, everybody who loads the site during a build. On a low-traffic site this can be a substantial proportion of everything recorded, and it is concentrated on exactly the pages you care about most.

What automatic filtering does and does not remove

It removes: traffic from agents on a maintained list of known bots and spiders.

It does not remove: internal traffic, automated traffic that identifies itself as a browser, referral spam sent directly to your property, or invalid clicks on paid campaigns.

And what any given platform's automatic exclusion covers has changed over time, so the correct move is to read your own platform's current documentation rather than an article's description of it, including this one. That is why nothing here names a setting.

The practical consequence is the one worth carrying away: if you switched on automatic filtering and the strange segment is still there, that is expected. It is not in the category that switch addresses.

What to do about each

Internal traffic

The highest-return fix and the easiest.

Exclude your own visits. Every serious platform provides a way to do this, usually by network address or by a marker set in the browser. Which mechanism is available, and what it is called, is a question for your platform's current documentation.

Two things to know about it, whichever mechanism you use. It applies going forward, so historical data keeps the internal visits. And it will not catch people working from home or on mobile networks, so it is a large improvement rather than a complete one.

Do this first, before anything else in this article.

Referral spam

Exclude the referrer where your platform allows it, or filter it in reporting.

Do not visit the domains. That is what the exercise is for.

And accept that it recurs. New domains appear, so this is maintenance rather than a fix.

Unidentified automated traffic

The hard category, and honesty is the right approach.

Look for the signature rather than trying to identify the agent: a single page, no engagement, one session per visit, an unusual regularity in timing, a concentration in one location or one network, or a spike arriving as a flat line rather than as a curve.

Where it is identifiable and consistent, filter it. Where it is not, the better answer is to segment your reporting toward behavior that only a person produces rather than to chase exclusions. Conversions, form submissions, purchases and multi-page sessions are all evidence of a person, and a report built on those is far less sensitive to whatever is in the session count.

Invalid clicks on paid campaigns

A separate system entirely. Ad platforms operate their own invalid-traffic detection and adjust billing for what they identify, on their own terms and their own timescale, and none of that has anything to do with your analytics filtering.

Do not attempt to reconcile the two. They are counting different things by different rules and expecting them to agree is a category error, which is a general property of this territory rather than a fault in either.

Two rules that keep you out of trouble

Filters apply going forward and cannot be undone in the data. A filter that removes real traffic removes it permanently from what you collect. So before applying anything, check what it would have removed by looking at the segment first, and prefer excluding at the reporting layer where you are unsure, because that is reversible and a filter is not.

And keep an unfiltered view if your platform allows it. Having somewhere to look when a number changes unexpectedly is worth the small overhead, and it is what lets you tell a filter from a real event. This site covers that distinction separately, and a filter added last Tuesday is one of the standard explanations for a cliff nobody caused.

The proportionality point

Bot and spam traffic distorts session counts far more than it distorts conversions, because automated traffic rarely completes a form or buys anything.

So the practical answer for most accounts is not an elaborate filtering setup. It is to report on conversions, revenue and engaged sessions rather than on raw sessions, and to exclude your own team.

That reframing removes most of the problem without touching a filter, and it is a better reporting habit anyway. Our article on choosing what a conversion should actually measure covers the setup side of that, and our article on what wasted spend looks like in an account covers the paid version of the same idea: a number that counts activity rather than outcomes is a number that flatters whatever produces activity.

The short version

Exclude your own team first. Filter referral spam as it appears. Report on outcomes rather than sessions.

And check what your own platform's automatic exclusion currently covers, in its own documentation, because it covers less than most people assume and what it covers has changed before.

Similar Posts