Signup-to-first-message on desktop: 1 minute 47 seconds — the longest in our test cohort to date, because of the age verification gate. GirlfriendGPT requires 18+ confirmation before unlocking adult content; this is the right call legally and ethically but it does add measurable friction for evaluation.
Character library exploration was the focus of week 1. The platform ships 100+ pre-built character tags and categories — fantasy, NSFW-specific, anime, realistic, scenario-based — substantially deeper out-of-the-box than Candy AI or SweetDream. We sampled 12 characters across 6 tag categories and found that the variety isn't just cosmetic: personality differences across characters were real, with distinct conversational tones, vocabularies, and roleplay defaults.
Free tier evaluation: limited daily messages and reduced NSFW depth. Enough to evaluate the chat feel but not enough to commit to an extended scenario. We upgraded to Premium ($14.99) on day 4 to remove caps for the remainder of the test cycle.
First NSFW-specific observation: the moderation posture is genuinely permissive. Where Candy AI flagged a couple of scene-types as warning gates and Kindroid surfaced a consent-confirmation interrupt, GirlfriendGPT allowed escalation through the same scenes without intervention. This will be a strong differentiator for the audience the platform targets; users wanting clearer guardrails should evaluate elsewhere.