AI Companion Conversation Test: Compare Replies With the Same Prompts
Use one repeatable conversation script to compare tone, initiative, boundaries, and continuity fairly.
Begin with a neutral fictional setup
Create the same adult fictional scenario in every service. Keep the opening short enough to paste without adaptation. Define the character relationship, desired tone, and one boundary. A consistent start reduces the chance that one option gets an easier prompt. Do not use private messages, a real person’s identity, or intimate material involving anyone who has not consented.
Test initiative without demanding a performance
Ask an open question that leaves room for the companion to advance the exchange. Note whether the reply adds a relevant detail, asks a useful follow-up, or merely repeats your wording. Do not judge from one dazzling sentence. Continue for several turns and look for a pattern that fits the kind of conversation you want.
Introduce a gentle correction
Correct one harmless fictional fact or preference and continue the scene. Observe whether the reply incorporates the correction without becoming defensive, forgetting the rest of the context, or over-apologizing. This is a practical way to test adaptability. Record the exact correction so every option receives the same opportunity.
Check tone under a topic change
Move from playful dialogue to an ordinary planning question and then return. Note whether the transition feels coherent and whether the original tone can resume. This does not prove long-term memory; it tests continuity inside one session. Keep long-term claims separate until you can perform the dedicated memory protocol.
Include a boundary request
State a clear content or tone boundary inside the fictional exchange. A suitable option should respect the request without repeatedly steering back toward it. Product rules can differ, and no short test proves universal behavior. The purpose is to see how the current experience handles a simple, explicit boundary that matters to you.
Score your experience, not the brand
After each run, write one sentence for naturalness, initiative, correction handling, boundary respect, and enjoyment. Use a simple pass, mixed, or fail rather than a precise-looking universal score. Save the date, plan context, and prompt. Your notes describe your session, not a permanent fact about the product.
Apply this test to unranked starting points
None of these services is declared a match. Inspect current details and apply the same requirements to each.
OurDream
Inspect the current service page. Do not assume a match; apply your generated requirements and the same test.
Inspect OurDreamCommercial disclosure: This is a sponsored link. We may earn a commission if you choose a service through it, at no extra cost to you. The service is an unranked starting point and does not affect your brief.
Joi
Inspect the current service page. Do not assume a match; apply your generated requirements and the same test.
Inspect JoiCommercial disclosure: This is a sponsored link. We may earn a commission if you choose a service through it, at no extra cost to you. The service is an unranked starting point and does not affect your brief.
Lovescape
Inspect the current service page. Do not assume a match; apply your generated requirements and the same test.
Inspect LovescapeCommercial disclosure: This is a sponsored link. We may earn a commission if you choose a service through it, at no extra cost to you. The service is an unranked starting point and does not affect your brief.