Skip to answers
Knowledge BaseTraffic quality & conversion answersTechnical documentation ↗
Browse answers

Traffic Detective

Are bots visiting my website?

In short

Look at the visitor's browser and network evidence, the classification reasons and the recorded journey. A suspicious label warrants investigation, not automatic blocking.

Research context

OWASP distinguishes benign crawlers and monitoring agents from abusive automation. The first question is what the requests do and which evidence supports their classification.

Now that's a very nice definition. However, it tells you exactly nothing.

Your first approach might be to look at server logs.

But sometimes, you won't see anything.

Other times, you'll see an attacker trying every .env path variant they can think of.

Some studies have tried to identify bots from logs alone.

One Microsoft study found that traversal depth (how many pages a visitor goes through) is a strong indicator of automation, at least on search engine result pages. See the study's figure comparing pages visited by bots and humans (PageTrackedCount).

Based on this, you could use a tool like Microsoft Clarity to judge, visit by visit, whether the visitor was a bot or a human.

In practice, that's difficult. Even session replays don't tell you much, partly because interaction events can be faked. Few vendors get this consistently right. Many approaches to behavioral (biometric) analysis exist, but they often leave room for false positives, and human-like mouse movement generators such as WindMouse make the problem harder.

These are some of the behavioral quirks (signals) that can tell you whether bots are visiting your pages or skewing your data today.

It's an incredibly deep and difficult subject. The paper "Behavioral Analysis and Bot Detection in E-Commerce Navigation Data" discusses classifying visitors from their page-to-page transitions alone.

Button uses a different but parallel method to double-check its work. We find that this approach, while valuable, doesn't give the full picture. Our first rule is to do no harm to existing ad campaigns or business-critical functionality. You need to be very sure before you act.

Our approach is holistic. We examine four layers of each visit to identify and group inconsistencies. Those groups can then be used to reduce real business losses from ad fraud, form stuffing and other abuse.

Layer What Button checks Why it may matter
IP Network reputation, proxy / datacenter signals, location consistency Shows whether traffic comes from infrastructure that deserves more scrutiny
TLS Protocol and browser-family consistency Finds clients claiming to be one browser while behaving like another
HTTP Headers, browser claims, request consistency Surfaces automation and mismatched environments
JS Runtime behavior and multiple browser-side consistency checks Adds evidence without treating one browser quirk as proof

You can inspect all of this evidence, but you don't need to understand it to make a reasoned decision.

Next step: Why did Button flag this visitor? · Related guide: Protect the form and the site, not just the ad account

Sources and further reading

Button Solutions Knowledge Base Evidence before action.