How it works, explained

How On Guard works, one page per question.

The walkthrough shows it. These four pages say it plainly: where detection runs, what a detection does, why there are no screenshots, and exactly what your ally receives.

Explicit content shows up, your ally knows. They learn that it happened, never what.

4 guides · Checked August 18, 2026 against the shipping app

What stays on your device, and what leaves it

Your device

What is on your screen

Frames are sampled as you scroll and browse.

On Guard engine

Classifies each frame locally. Nothing is uploaded.

A verdict

Clean, suggestive, or explicit. The frame is discarded.

Crosses the boundary

  • A liveness heartbeat
  • A detection event, with no content
  • Your account, settings, and billing

Never crosses it

  • Screenshots or frames
  • URLs or browsing history
  • What you were looking at
The frame stays. The verdict travels without it.

Questions answered

Is this the same as the how-it-works page?

No. That page is a walkthrough you scroll, from inside the app. These four pages are the plain-text answers behind it, one question each, so you can read or share exactly the part you need.

What is the shortest honest summary?

Detection runs on your device. Explicit content gets a gate, a way out, and one check-in text to your ally. Suggestive content gets a nudge and nothing else. No screenshots are kept or sent, and your ally learns that it happened, never what.

What about iPhone?

On iPhone and iPad, On Guard runs no detection engine of its own; it works through Apple's Screen Time controls. The ally check-in, account, and settings are the same everywhere.