How likely is your traffic to be from real humans?

How likely is your traffic to be from real humans?

How to handle the rise of bots and autonomous agents

51Degrees Device Detection IP Intelligence AI Device Detection

Not everyone reading your pages is a person. Agents read them for people, and crawlers copy them for training. Your analytics counts the agents as users, and the crawlers it does not recognize. We’ve combined our superpowers into one property, Human Confidence, so you can tell the humans from the bots.

Human Confidence

HumanConfidence is the property returning one word; High, Medium, or Low.

Value Meaning Action
Green human face for High High A person is reading the page. Let the request through.
Amber face, half human and half robot, for Medium Medium Mixed signals. Could be a person, could be a script. Challenge with a CAPTCHA, a login, or, on a business-to-business site, a notice like ours at 51degrees.com/suspicious.
Red robot face for Low Low Almost certainly no person is reading it. Send it to a page telling the AI how to get a license and engage its human, like ours at 51degrees.com/ai-notice. Log the visit and issue an invoice to the AI agent.
HumanConfidence values and suggested actions. Business-to-business sites might choose different treatments.

Analytics

Add HumanConfidence to your analytics solution as a custom dimension and then segment data by Low, Medium, and High. The volumes will change over time. Learn when AI clients performing agentic browsing becomes unacceptable to you or AI companies seeking training and grounding data need to be charged.

Monthly visits by HumanConfidence from May to September, with known crawlers removed
An example of monthly visits by HumanConfidence from May to September, without known crawlers, showing the rise of Low and Medium confidence traffic. With known crawlers included, the Low bars would be significantly higher.

Charge the AI scrapers 💲💲💲

The Movement for an Open Web (MOW) publishes a free Search Only Contract that any website can adopt. It allows search indexing and human, non-commercial use. All other access, including AI training, building datasets, and bulk scraping, needs a deal or a fee, set by default at £500 for each article, image, or video used. Add these lines to robots.txt and cite the contract in your terms.

User-Agent: *
# Terms https://m4ow.uk/socw/2.txt
TDL: https://m4ow.uk/socw/2.txt
Allow: /

An AI scraper or bot that accesses your site has accepted the contract, and use beyond search indexing and non-commercial use attracts the fee. Use HumanConfidence to log those visits as evidence, and invoice the operators. Non-payment can lead to enforcement through the County Court. UK publisher Trusted Reviews is already logging the bots that ignore its contract.

Read the MOW guide to the contract, the MOW launch announcement, and Press Gazette’s interview with Chris Dicker of Trusted Reviews. Our free robots.txt generator writes the rest of the file for the uses you allow each crawler.

What next?

Start with a subscription to the 51Degrees cloud service and add Human Confidence to your website in three steps.

  1. Get a resource key, which identifies your subscription to the cloud.
  2. Add the 51Degrees script to your pages.
  3. Send HumanConfidence to Google Analytics, or whatever analytics you use, as a custom dimension.

Then you learn how much of your traffic is real humans. A step-by-step guide for Google Analytics follows next week.

A publisher or a CDN can decide which bots get in and stop paying to serve the rest. Adtech can drop Low before the bidstream, so nobody pays to show an ad to a script.

51Degrees has separated robots from people since 2009, on-premise and in the cloud, and added IP intelligence in 2025.

Make the AI scrapers pay. Or feed them, if you sell to businesses and want AI answers to cite you. Either way 51Degrees gives you choices.

Read more about our approach to AI at 51degrees.com/ai.