The honest answer here is simpler and more limited than you might expect. There's no fingerprinting system tracking your canvas rendering or audio signature, no AI watching for "pixel repetition" to catch fake camera feeds. What actually exists is more basic, and worth explaining plainly rather than dressing up.
A lot of the bot problem on other platforms comes from automated account creation — scripts that sign up thousands of fake profiles to spam links or scrape data. Chatzyo doesn't have accounts at all, which removes a big chunk of what makes platforms attractive to bot operators in the first place. There's no profile to create, no follower count to inflate, nothing to automate at scale in quite the same way.
That doesn't mean nobody ever tries to run a script against the platform — it means the economics are different, and the basic structure of a no-account, peer-to-peer platform happens to make some of the more common bot strategies less useful here than they'd be elsewhere.
The text chat filter that blocks phone numbers and email addresses also happens to catch a lot of the obvious spam patterns bots tend to use, since the two overlap — a script blasting the same link into every conversation runs into the same filter built for contact-sharing. It's worth being clear this is pattern-matching on text content, not a behavioral analysis engine tracking how you click or type.
Beyond that, the Report button is the main tool. If you connect with something that's clearly an automated feed or a bot account, report it the same way you would any other violation. A person reviews it, and the platform doesn't need a sophisticated detection system for an obvious case — most automated accounts are pretty easy for a human to recognize once flagged.
Someone playing a pre-recorded video loop instead of their live camera is a real thing that happens occasionally on any video platform, including this one. There's no automated "frame-consistency" system catching this in real time — it's not something Chatzyo's architecture is built to detect automatically, since the video stream is peer-to-peer and never reaches a server that could analyze it.
What that means practically: if something about a video feels off — too smooth, the same loop, no response to anything you say or do — trust that instinct and skip. There's no harm in being wrong about it, and reporting it afterward helps even if we can't independently verify a fake feed after the fact, since we don't have a recording either.
It would be easy to write a more impressive-sounding version of this page describing sophisticated detection systems — and that's exactly what existed here before this version. We'd rather you know the real, more limited picture: a small platform with basic rate-limiting on text patterns and a human-reviewed report system, not a security operations center running machine learning models on your video feed.
That's not a weakness we're trying to hide — it's the actual tradeoff of building a platform that doesn't ask for an account and doesn't route your video through a central server. If catching every bot automatically mattered more to you than keeping your video private and off our servers, a different kind of platform might suit you better. For what we think the right balance is, and how reports fit into the bigger picture, see how moderation actually works.