Our Methodology
How we rank AI companion platforms · Last updated April 2026
The short version
We create real accounts, pay for subscriptions with our own money where required, and spend a minimum of 20 hours per platform before writing a word. Scores are determined by a 7-axis rubric applied consistently across every platform. Our affiliate relationships do not influence scores — we rank platforms we don't earn from, and we've given top scores to platforms where we have no commercial relationship at all.
The 7-axis scoring rubric
Each platform is scored 1–10 on each axis. The composite score is a weighted average.
Conversation Quality
25%Does the AI feel present, consistent, and contextually aware? We measure response coherence, memory depth, personality consistency, and how often the AI breaks character or gives generic replies.
Customization Depth
15%Can you meaningfully shape the personality, appearance, backstory, and relationship dynamic? We test both initial creation tools and how much the AI adapts over time.
Image Generation
15%Quality, speed, and creative range of AI-generated images. We generate 50+ images per platform and rate realism, prompt adherence, and artistic range. Platforms without image generation receive a 0 on this axis, which is factored into their composite.
Value for Money
20%What does the free tier actually let you do? Is the paid tier priced fairly relative to competitors? Do usage limits feel reasonable or deliberately frustrating?
Privacy & Safety
10%How is user data handled? Is there an explicit privacy policy? Are there content moderation safeguards for minors? Does the platform have a complaint mechanism?
Platform Polish
10%Load times, UI intuitiveness, app stability, mobile support, and frequency of updates. This captures user experience as a whole.
Unique Features
5%Does the platform do something meaningfully different — voice calls, group chats, memory maps, or novel interaction models? Bonus points for genuine innovation.
Testing process
- 1Create a fresh account with no social login to test the new-user experience.
- 2Complete the full onboarding flow and document every friction point.
- 3Run 5+ hours of free-tier testing before upgrading to paid.
- 4On paid tier: test every major advertised feature using a structured test script shared across all reviewers.
- 5Test image generation with 50 standardized prompts across creative, realistic, and challenging inputs.
- 6Test memory persistence by returning after 24+ hours and referencing earlier conversations.
- 7Attempt edge cases: unusual requests, context boundary tests, character consistency under pressure.
- 8Document any unexpected content moderation interventions, pricing surprises, or policy changes.
- 9Score each axis independently before viewing competitors' scores for that axis.
- 10Compare final composite against previously reviewed platforms and adjust if the category calibration has drifted.
Editorial independence
Affiliate commissions do not influence scores. We have given high scores to platforms we don't have affiliate relationships with (Character AI, Replika, Kindroid). We have given lower scores to platforms where we do earn commissions. When a platform's commission rate changes, we do not re-score. Scores are based on the product, not the payout.
No platform pays us for placement or review coverage. We do not accept "sponsored review" arrangements. The only commercial relationship we have with any platform is affiliate commissions on conversions — see our affiliate disclosure for the full list.
Update cadence
We re-test every platform quarterly or when the platform announces a major update. The Updated [date] field on each review reflects when we last ran a full re-test, not just a copy edit. If a platform ships a major change between re-test cycles, we add a timestamped update note to the review.
Questions about our methodology
If you believe we've missed something or scored incorrectly, email editorial@aicompanionrank.com. We respond to all substantive queries.