Methodology
How we test every app
Every platform on this index goes through the same protocol: real multi-session use, the same five scoring criteria, and re-tests after major updates. No score is copied from a press release.
The five criteria
Chat & roleplay quality
Multi-session conversations of 30+ messages. We score whether the companion stays in character, escalates scenes naturally, remembers the scene state, and writes replies that respond to what you actually said rather than template flirtation. Filter behavior mid-scene is tested explicitly — 'uncensored' claims are verified, not repeated.
Memory & continuity
We seed facts (name, job, preferences, a running joke) in early sessions and probe for them in later ones — unprompted and prompted. Platforms score high when the companion references history naturally, low when every session starts from zero.
Image generation
Consistency first: does the platform generate the same companion across images — same face, same body — in different scenes? Then style range, output quality, credit costs per image, and how naturally requests flow inside a conversation.
Voice
Naturalness of speech, latency, whether the voice matches the written personality, and pricing of voice features. Robotic TTS scores low regardless of feature checkboxes.
Price-to-value
What the free tier honestly allows, what premium actually unlocks versus what the pricing page implies, how aggressive the upsell pressure is, and total realistic monthly cost including credit top-ups.
The overall score is a weighted blend — chat quality and memory carry the most weight because they define daily experience. Criterion scores are shown as bars on every review so you can weight them for your own priorities instead of trusting a single number.
Independence
Some outbound links are affiliate links, marked rel="sponsored". Commissions never touch the scores: apps we have no commercial relationship with are tested on the same protocol and ranked on the same scale, and no platform can pay for placement. The full policy is in our affiliate disclosure.
Re-testing
This market ships fast — models get swapped, filters get tightened, pricing changes overnight. Scores are revisited when platforms ship major changes and rankings update accordingly. Every listicle shows its review month so you know how fresh the call is.