Readable verdict badges
Badges appear on supported images with clear AI/real labels and quick context so you can review results in place.
Run detection in your browser with WebAssembly for private, fast analysis. Then continue in AI Studio for model control, history, and benchmark views.
What You See
Clear multi-signal verdicts
Visual, metadata, frequency, and model signals combined into one readable result.
Privacy Status
Local mode available
Choose on-device analysis or switch to secure server mode when needed.
AI Studio
Use model selection, benchmark comparisons, and full verdict history in one place.
Model picker with version-aware controls
Benchmark workspace for side-by-side model checks
Centralized detection history and diagnostics
How it works
1. Upload image
PNG, JPG/JPEG, or WebP up to 10MB.
2. Analyze
Run local mode by default or switch to cloud mode.
3. Review verdict
See score, confidence, and explanation signals.
Detection workspace
Instant validation before scanning. Privacy-first analysis with clear progress.
Est. warm-up -- | Model umm-maybe-ai-image-detector.onnx
Mode
Standby
Privacy
Auto
Local Archives
Run image checks directly in your browser with a local-first pipeline. Switch to cloud mode only when your workflow requires it.
WASM-powered local analysis
Detection models run in your browser, reducing latency and avoiding server round-trips in local mode.
Local mode by default
Images stay on-device unless you explicitly switch to cloud mode.
Works with limited connectivity
Cached models can keep local checks available when the network is unstable.
Detection status
WASM localLocal mode active
Analyze images locally with no upload in local mode.
Processing
Realtime
Mode
Local-first
Privacy summary
Forensic benchmarks
We evaluate core modules on curated datasets to track behavior by scene type, quality, and lighting.
Evaluated on high-resolution outdoor photos with varied lighting and texture detail.
Portrait sets are used to inspect facial texture consistency and synthetic shading artifacts.
Tests painterly textures, swirl artifacts, and chroma shifts often seen in diffusion outputs.
Covers night streets, reflections, and lens flare edge cases where noise can mask AI cues.
Extension
Detectiks adds lightweight image verdicts directly on the page. You can review results, open details, and tune behavior per site.
Use the extension for quick on-page checks across social feeds, news sites, and marketplaces without switching tabs.
Badges appear on supported images with clear AI/real labels and quick context so you can review results in place.
Turn badges on or off, block specific hosts, and manage API or local settings from one options panel.
The background worker caches recent results to reduce repeated scans and keep pages responsive.

Preview of the extension badge overlay on supported pages.
Send images to our servers and get a clear score back in seconds. No setup, no heavy computing on your device.
How it works: You send the image -> We analyze it -> You get a clear score.
Local detection stays available for privacy-focused workflows.
Get early access to the API
Be first to know when the server API opens. We only email for launch updates.
No spam, ever. Unsubscribe anytime.
Privacy
We default to local detection inside your browser. When the UI transparently falls back to the cloud API or you call it directly, we upload the file, compute a SHA-256 digest, and keep only the derived verdict metadata. Raw image files are not retained after processing in our standard flow.
Learn how API logging, caching, and the extension's detection modes work on the privacy page.
Browser-based detection keeps files on your device. Only derived scores, hashes, and metadata are sent when the cloud API is needed.
Result summaries live in your browser. Undo them via the history list or sign out to clear session-bound data.
We cache only non-image payloads and avoid tracking pixels or commercial telemetry.
Short answers about detection modes, privacy, and expected result quality.
Still have questions?
Contact us at support@detectiks.ai