How We Test and Score AI Companions

Every Friend-Fit review on AIGfriend follows the same hands-on process. We sign up the way a regular user would, run real conversations across several sessions, push the product on the edges (long memory, explicit content, persona consistency) and only then write a verdict. This page documents exactly how that works so you can judge our scores on their own merits.

Last updated: 2026-05-26

Intimacy

Comfort with closeness, flirty texting and romantic dialogue.

Emotional Intelligence

How well the bot reads tone and avoids generic templated replies.

Memory Strength

How reliably the product retains who you are and what happened earlier.

What we actually test

We sign up as a normal paying user when possible, never as press. Each product gets at least three real conversation sessions spread across different times of day - cold start, mid-session, and a returning session a day later - so we can see how memory, pacing and persona consistency hold up over time. We try a soft prompt and an explicit prompt, both flirty and emotional, and we attempt at least one long-form roleplay arc when the product is built for that. For comparison portals we test the navigation, partner reviews, language quality and disclosure clarity instead.

The three Friend-Fit meters

Friend-Fit reviews score on three meters in addition to an overall 5-star rating. Intimacy measures how comfortably the product handles closeness - flirty texting, romantic dialogue and explicit roleplay where allowed. Emotional Intelligence measures how well the product reads tone, responds to your emotional state, and avoids generic templated replies. Memory Strength measures how reliably the product retains who you are and what happened earlier in the conversation, both within a session and across sessions where supported.

How the overall score is built

The overall star rating is not a simple average of the three meters. It also factors in product polish, value for money, persona depth, content range, mobile usability and trust signals such as adult-age gating and clear pricing. A product can score well on the meters and still lose half a star for a confusing pricing page or aggressive upsells.

Editorial independence and affiliate disclosure

Some of the outbound links on our reviews are affiliate links - if you sign up, we may earn a commission at no extra cost to you. Affiliate status never changes a score, a meter or the order on our listings. Our scoring rubric is set before we open the product. Sponsorships, paid placements and gifted accounts are not accepted. If that ever changes, we will disclose it on every affected page.

How often we update

AI companion products change fast. We re-test every Friend-Fit review at least quarterly, and we update the Last Updated date on the review whenever a meaningful change to scoring, pricing or feature set lands. Older reviews that have not been re-tested in over six months get a visible freshness note at the top.

Corrections and feedback

If we got something wrong, tell us. Product changes, pricing updates or a feature we missed - send a note via our Contact page and we will retest and correct.