How detection works
How Kaminari Ad turns a rendered page into tags, a severity, a classification, and explainable evidence.
A finding is not a black box — it is grounded in what the crawler actually observed.
What the crawler captures
Each scan renders the creative and follows every redirect on a real device from a residential IP, capturing screenshots, the full redirect chain, the page DOM and text, and any download or permission prompts.
How findings are produced
The captured evidence runs through three layers:
- System detectors — deterministic checks for known-bad patterns (forced downloads, cloaking, suspicious redirects) plus domain and URL reputation lookups.
- Custom rules — your own stopword, regex, blacklist, combo, and AI rules (see Detection rules).
- AI content analysis — a multimodal model reviews the screenshot and page text to classify the ad and flag deceptive or non-compliant content.
Each layer attaches tags to the scan. Every tag carries a
category and a severity (high, medium, or low). Alongside
the tags the scan gets a classification — detected brand, AI
content category, and IAB v3 mapping.
Why you can trust it
Every tag links back to its evidence — the screenshot, the redirect chain, the page text. When the AI flags something you can see exactly what it saw, which makes findings defensible to your partners and auditors.