Half the web's traffic isn't human. Keep bots out of your numbers
Automated traffic passed human traffic in 2024. How bots inflate WordPress statistics, how to spot them, and what honest counting looks like.
If your visitor numbers jumped last month and sales did not, check who those visitors were. There is a good chance many of them were not people.
How much traffic is automated
Imperva, which sells bot protection and so has an interest in the topic, publishes a yearly Bad Bot Report from traffic across its network. Its 2025 edition found that automated traffic made up 51 percent of all web traffic in 2024, the first time in a decade that bots outnumbered humans. It counted 37 percent as malicious bots and 14 percent as good ones: search engines, uptime monitors, feed readers. A single vendor's network is not the whole internet, but the direction matches what we see in server logs.
Some of that traffic is AI. Crawlers from OpenAI, Anthropic, Perplexity and others read pages for training and to answer questions, and they are a growing share of requests on content sites.
How bots end up in your statistics
Server-side counters see everything that requests a page, including crawlers, scrapers and uptime checks. Unless they filter by user agent, the numbers are badly inflated.
JavaScript-based analytics only count visitors whose browser runs the tracking script, which removes most simple bots. Modern scrapers often use real headless browsers, though, and some pretend to be ordinary Chrome on Windows.
Referral spam sends fake visits with a referrer pointing at a site the spammer wants you to visit out of curiosity.
Signs a spike is not human
- A sudden jump from one country or one network, often a cloud provider
- Visits of zero seconds, one page each, at perfectly even intervals
- A single page receiving most of the extra traffic
- Screen sizes or browser versions that are oddly uniform
- Referrers you have never heard of, all new this week
What honest counting looks like
- Filter known bots by user agent. The well-behaved ones, including search engines and the major AI crawlers, say who they are.
- Count only visits that run the page's script, which drops most of the crude ones.
- Leave out your own team, so editing pages all day does not look like interest.
- Count bots separately, not never. Knowing that GPTBot or ClaudeBot read your pricing page is useful information; it just does not belong in visitor numbers.
- Block abusive bots at the firewall, so they cost your server nothing.
In Ironhelm
Visitor statistics count only visits from browsers that run the page's script and do not identify as bots, and can leave out logged-in people. AI visibility counts the crawlers separately: which AI assistants and search engines read which pages, and how many visitors AI assistants then send you. The firewall can block fake search engine bots and abusive addresses.
Sources
- Imperva, 2025 Imperva Bad Bot Report: How AI is supercharging the bot threat (April 2025)
- OpenAI, Overview of OpenAI crawlers
- Google Search Central, Verifying Googlebot and other Google crawlers
Ironhelm Suite is one WordPress plugin for security, speed, SEO and analytics. See plans or tour the features.