AI Detection is available for our PRO and Enterprise plans
AI-Powered Detection adds a layer of AI agents on top of Fraud Blocker's standard protection, to catch sophisticated invalid traffic (SIVT) that looks like a real buyer. Fraud bots now use AI to mimic human behavior, rotate through clean IP addresses and slip past rule-based checks. AI-Powered Detection is built to catch them, and it tells you in plain language why each visitor was blocked.
What is AI-based detection?
With AI-Powered Detection on, every ad click is checked by two systems at once:
Our signal-based detection engine: This is our standard system that scores each click on proven fraud signals: VPNs and data centers, blacklisted IPs, excessive clicks, risky geography, accidental clicks mouse dynamics and more.
Our AI agents: They review each visit as a whole and look for the subtle patterns that signal-based checks miss.
Together, they separate real customers from bots, click farms, competitors and automated AI agents.
The AI doesn't just label traffic "good" or "bad". It works out who, or what, is actually clicking your ads.
How does Fraud Blocker's AI-based detection work?
Every click is analyzed. Our AI agents review each IP that clicks your ads. They weigh many signals together, from click timing and repeat visits to device, network and on-page behavior.
Flagged clicks raise the fraud score. When the AI flags an IP as bot activity, that adds to the IP's fraud score alongside the standard signals.
High-scoring IPs are blocked automatically. An IP whose score crosses your auto-blocking threshold is blocked in your connected Google Ads and Meta Ads accounts, just like any other fraud type. Where the AI's flag was the strongest signal, the Primary Fraud Type reads "AI Detected Bot".
You choose the AI model
Fraudsters upgrade their tools constantly, so we don't rely on a single AI provider. We test the latest frontier models and offer the best ones for analyzing your traffic. You can pick your model per website:
As of October 2026, here's the models you can choose from:
Model | Provider |
GPT-6 Luna (default) | OpenAI |
Claude Haiku 5.5 | Anthropic |
DeepSeek V4.1 Flash | DeepSeek |
GLM 5.3 Flash | Z.ai |
There is no additional cost regardless of which model you choose. And we'll keep adding models as better ones become available.
It's trained on Fraud Blocker's large network of data
Our AI agents learn from a real-time stream of click data across the 25,000+ websites that Fraud Blocker protects and activity from 60,000 million monthly IP addresses and devices. That means they can recognize fraud patterns seen elsewhere on our network, often before those patterns reach your campaigns.
How to turn on AI-Powered Detection
AI-Powered Detection is set per website. Repeat these steps for each website you want to protect with AI.
Log in to your Fraud Blocker dashboard.
Select the website from the website menu in the left sidebar.
Go to Customizations.
Find the AI-Powered Detection card and switch the toggle to On.
Optional: pick a different model from the Your Model menu on the same card. You can change the model at any time, even while AI-Powered Detection is off.
That's it. Once it's on, the top bar shows AI Enabled, and AI-flagged IPs start appearing in your reports. To turn it off, switch the same toggle back to Off.
Where to see AI detection results in your dashboard
While AI-Powered Detection is on, blocked fraud shows in purple instead of red, so you can tell AI-driven blocking apart at a glance. The top bar also shows an AI Enabled badge.
"Why?" on every auto-blocked IP: In the IP tables on Fraud Score and Reports, each IP that was auto-blocked on its fraud score has a "Why?" link next to its Primary Fraud Type. Click it to see a plain-language explanation of the block
AI Detected Summary: On the Fraud Score page, the Blocked Fraud Type card adds an AI Detected Bots row. It shows the share of your blocked clicks that came from AI-flagged traffic.
AI reasons in your exports. When you export the IP table as CSV or Excel, or print it, an AI Detection Reason column is included right after Primary Fraud Type.
Frequently asked questions
How is it different from Fraud Blocker's standard protection?
Our standard protection uses a behavior-based engine to spot invalid clicks from signals like IP reputation, device fingerprinting and click patterns. AI-Powered Detection adds a second layer: AI agents that review each visit as a whole and catch sophisticated bots that blend in with human traffic. It works alongside your standard protection, so you keep everything you have today and gain an extra layer for the toughest fraud.
Which plans include AI detection?
All Pro plans. It isn't available on Starter. See Which plans include AI-Powered Detection? above.
How do I turn on AI detection?
Go to Customizations → Detection Rules and switch on the AI-Powered Detection card for each website. See the step-by-step instructions above.
Can I see why a specific click or IP was blocked?
Yes. Every IP that was auto-blocked on its fraud score has a Why? link in the IP tables on Fraud Score and Reports. For AI Detected Bot rows, it shows the AI's plain-language reason and the model that flagged the IP. The same reason is included in CSV, Excel and print exports.
Which ad platforms does AI detection work with?
AI-flagged IPs are blocked automatically in Google Ads and Meta Ads, just like other fraud types. For Microsoft Ads, IPs can't be blocked automatically, but AI-flagged traffic still appears in your reports.
I'm an agency. Can I use AI detection across my client accounts?
Yes. AI-Powered Detection is set per website, so you can turn it on for each client website under your Pro plan, and choose a different model for each one if you like.
What's the difference between "AI Detected Bot" and labels like "OpenAI-bot" or "Gemini AI-bot"?
Labels like OpenAI-bot and Gemini AI-bot mark known AI crawlers, which are not blocked or billed. AI Detected Bot means our AI agents flagged the visit as fraudulent bot activity.
What if a real customer gets blocked?
AI-flagged IPs follow the same auto-blocking threshold as every other fraud type. You can unblock any IP at any time with the Status switch in the IP table. Use the Why? link to see the reason before you decide.
Do I need to change my tracking code?
No. AI-Powered Detection uses the data your existing Fraud Blocker tracking code already collects. The analysis runs on our servers, so it adds nothing to your website's load time.
Still need more information?
Here's our launch video that may help or contact us!
Related articles: