AI Companions Lab
The Lab publishes structured, reproducible tests of AI companion platforms. Every test is run against our published methodology with the same protocol for every platform. Status is always visible — we do not present a Planned test as if it has already produced results.
Testing — the test is running now. No numbers publish until it completes.
Planned — the test is on the roadmap. No numbers exist yet.
Current tests
Full 9-category hands-on review of Secret Desires, scored on the published methodology.
Comparative ranking of the AI girlfriend platforms we have completed hands-on testing on.
50 generations per character, four platforms, same prompt structure. Scoring face, hair, body and style consistency. Results publish only when the full protocol has run.
Day-1 / Day-7 / Day-30 recall of names, preferences and relationship context across the leading platforms. Full protocol on the methodology page.
Voice-reply quality, voice-call latency, turn-taking behaviour, and call reliability across platforms that support voice.
Short-form video generation quality, motion coherence and same-character consistency across platforms with video support.
Systematic evaluation of what each platform lets you tune when creating a custom character. Personality, appearance, communication style, roleplay parameters.
Independent read of each platform's privacy policy, data-handling claims, account-deletion behaviour and billing transparency.
What we don't do
We do not fabricate scores, invent screenshots, or claim to have tested a feature we haven't. When a test is Planned or in Testing, it says exactly that — with no numbers attached. When a test is Published, every score is traceable to real test runs against the methodology.
Chat, customize and create with AI companions on Secret Desires.
Try Secret Desires →