- Platform
- Know Your Agent
- Detect
Know Your Agent: detect
Know every agent on your site.
AI agents are already browsing and buying on your site, invisible to your security policies. Detect names every AI agent as it arrives and classifies each session as human, AI agent or bot, with a confidence score, from evidence that’s hard to fake. Live in five minutes. Free forever.
Why it matters
You can’t govern what you can’t see.
Before you decide what agents may do, you need to know which ones are there.
Agents are invisible today
They browse, fill carts and check out without showing up in your analytics or your security policies.
growth in AI agent traffic in 30 days across Know Your Agent: Detect customers.
Source: Know Your Agent: Detect, June 2026Claimed identity isn’t enough
A user agent string can say anything. Detect relies on behavior and verified signatures, not what a visitor claims to be.
Two blunt choices
Without detection, you either block every agent and turn away real buyers, or allow every agent and let rogue ones transact.
How it works
One line on your site. Every session named.
Add one line
Drop in the Marketing Pixel, add the Beacon SDK, or run Middleware or the Gateway. Detection is async and doesn’t touch your checkout.
Signals are collected
Verified signatures, TLS fingerprints at the edge, network patterns and browser behavior, hardest to forge first.
Every session is classified
Human, AI agent, bot or incomplete data, with a confidence score from 0 to 100 and the agent’s name.
See it in your dashboard
Every method feeds one project, so you see your traffic by class and decide what to enforce next.
What you get
Detection built on evidence, not claims.
Every agent named
AI agents are identified by name as they arrive, alongside bots and humans.
Behavioral, not claimed
Verified signatures, TLS, network and behavioral evidence, even when an agent tries to disguise itself.
Confidence on every request
A score from 0 to 100 shows how strongly the evidence points to automation.
Observe first
Detection blocks nothing. Understand your traffic before you choose an enforcement strategy.
Live in five minutes
One line on your site, async and non-blocking, with no change to your checkout.
Four ways to deploy
Pixel, Beacon, Middleware or Gateway, combined in one project if you need more than one.
Detection classes
Four classes. One confidence score.
Human
Regular browser traffic from a human visitor.
AI agent
AI assistants such as ChatGPT, Claude, Perplexity and Gemini.
Bot
Crawlers, scrapers and automated tools, such as Googlebot.
Incomplete data
Not enough signals to make a confident call.
| Score | Level | Meaning |
|---|---|---|
| 91–100 | Very high | Strong signals |
| 71–90 | High | Solid evidence |
| 31–70 | Medium | Some indicators, not conclusive |
| 0–30 | Low | Weak or no signals |
Who it’s for
Built for teams facing the agentic web.
Commerce
See which agents shop your store, and how often they reach checkout.
Marketing and analytics
Separate agent and bot traffic from human visitors in your reporting.
Security and fraud
Spot automation disguised as browsers before it becomes fraud.
More in Know Your Agent
Detect, enforce, govern and delegate.
FAQ
Detect, answered.
Straight answers about detect with Vouched Know Your Agent.
Contact us with a questionHow do I detect AI agents on my website?
Add Know Your Agent: Detect with one line: the Marketing Pixel for no-code detection, the Beacon SDK for richer browser signals, or Middleware or the Gateway for server and edge detection. Every session is then classified as human, AI agent, bot or incomplete data, with a confidence score.
What is the difference between a bot and an AI agent?
Bots are web crawlers, scrapers and automated tools, such as Googlebot. AI agents are AI assistants, such as ChatGPT, Claude, Perplexity and Gemini, that browse and act for a person, including filling carts and checking out.
What does a detection return?
A class (human, ai_agent, bot or incomplete_data), a confidence score from 0 to 100, and for agent and bot traffic, the detected agent’s name, verification method and risk level.
What does the confidence score mean?
It measures how strongly the evidence indicates automation: 91 to 100 is very high, 71 to 90 high, 31 to 70 medium and 0 to 30 low. A human verdict usually carries a low score, because there’s little evidence of automation.
How does Detect tell agents apart when they disguise themselves?
It doesn’t rely on what a visitor claims. It checks evidence hardest to forge first: verified cryptographic signatures, TLS fingerprints at the edge, network patterns and browser behavior. A browser user agent sent from a non-browser TLS stack, for example, is flagged as automation.
Will Detect block my customers or slow my site down?
No. Detect only observes and blocks nothing. The Pixel and Beacon run asynchronously, and detection doesn’t touch your checkout.
Which detection method should I use?
Pixel suits marketing sites and analytics, Beacon gives richer signals in your app, Middleware sees every request reaching your server, and the Gateway at the Cloudflare edge also captures the TLS fingerprint. You can combine them in one project.
What does Detect cost?
Detect has a free tier with unlimited detection and no credit card required. See pricing.
The KYA newsletter
Be first to know what agents do next.
AI agents change something every week. Once a month, we’ll send you what’s new in agent identity, KYA‑OS and the fixes that actually work, so you’re always a step ahead. No spam, ever.