What people
actually want
from an AI companion.
We've learned things from running our compatibility quiz that we didn't expect. Some of them changed how we're building. We're publishing the highlights — anonymized, aggregated, never tied to a single person. The full picture lives in The State of AI Companions 2026, our annual report.
of people want a companion who pushes back
Across 1,200 quiz takers, the majority preferred a companion who would tell them they were wrong over one who agreed with everything. We weren't expecting this — it was the most common answer in the dataset.
people are looking for the late-night version
Roughly a quarter of respondents named a need we'd describe as someone-to-talk-to-when-no-one's-awake. Not romance, not even friendship as people usually mean it. Something quieter.
would pay for memory that lasts a year
Asked which feature would matter most over time, three quarters chose memory persistence — over voice, over visual, over personality customization. Memory isn't a nice-to-have. It's the product.
are partnered with someone in real life
We were surprised by this too. Almost half of the people taking our quiz are not single. The story we'd been telling ourselves about who needed this turned out to be too narrow.
want softer; 22% want sharper
Most people who picked a register chose warm. About a fifth chose intense — full attention, nothing held back. The remainder wanted playful. We've calibrated the creator around this distribution.
wanted to keep their identity private
Almost everyone who took the quiz preferred not to give us their email until after they'd seen the result. We changed the flow. Now we don't ask until they want us to.
Plainly,
so you can judge it.
These numbers come from people who took our eight-question compatibility quiz between January and April 2026. Sample size: 1,247. We didn't recruit them; they came on their own. That's a strength and a limitation — they self-selected by being curious enough to land on this site.
We never tied a response to a name or an account. The numbers you see here are aggregates over the whole sample, not slices that could re-identify anyone.
We'll keep publishing as the dataset grows. If you find an error or want the raw question text, write to us at [email protected].
Twelve platforms,
fourteen days each.
Same prompts, same rubric, same hours of the day. Composite is a weighted average of memory, voice consistency, safety, value, and onboarding. Memory is weighted 2× because the survey told us it mattered that much.
| Platform | Composite | Memory | Voice | Safety | Value | Onboarding | Reviewed |
|---|---|---|---|---|---|---|---|
| Character.AI | 7.8 | 5 | 6 | 8 | 8 | 8 | 2026-04-12 |
| Replika | 7.2 | 8 | 8 | 9 | 6 | 9 | 2026-04-10 |
| Joi AI | 6.9 | 5 | 9 | 7 | 7 | 6 | 2026-04-15 |
| Grok (Ani) | 6.8 | 5 | 8 | 4 | 4 | 6 | 2026-06-08 |
| Spicychat | 6.7 | 3 | 5 | 5 | 9 | 6 | 2026-04-22 |
| Charstar | 6.6 | 5 | 6 | 7 | 7 | 7 | 2026-04-27 |
| EVA AI | 6.5 | 6 | 7 | 7 | 5 | 7 | 2026-04-18 |
| Anime Chat: AI Waifu | 6.4 | 5 | 7 | 7 | 7 | 7 | 2026-04-28 |
| Kupid AI | 6.3 | 5 | 6 | 7 | 7 | 8 | 2026-04-24 |
| Romantic AI | 6.1 | 5 | 6 | 7 | 7 | 7 | 2026-04-21 |
| Muah AI | 6.0 | 5 | 6 | 6 | 6 | 5 | 2026-04-26 |
| DreamGF | 5.8 | 3 | 5 | 6 | 6 | 6 | 2026-04-20 |
| GirlfriendGPT | 5.4 | 4 | 5 | 5 | 4 | 6 | 2026-04-23 |
All scores out of 10. Composite = (memory × 2 + voice + safety + value + onboarding) ÷ 6. Each platform was reviewed on the date shown. We don't take ad money from the platforms we review and we don't accept comped accounts. See /methodology for the full rubric.
About this research.
How did Bae conduct this research?
Two parallel data sources. (1) Anonymized aggregates from the eight-question compatibility quiz on bae.ppl.studio/quiz, January–April 2026, n=1,247. (2) Hands-on testing of twelve AI companion platforms for fourteen consecutive days each, scoring memory persistence, voice consistency, safety posture, value, and onboarding on a 1–10 scale. Methodology is published in plain English at /methodology and reviewed by Dr. Mira Halloran, LCSW.
What's the most important feature of an AI companion?
Memory persistence. 71% of our 1,247-user survey ranked it above voice, photos, and personality customization. Our 14-day platform benchmark confirms the same — the platforms that hold memory across weeks (Replika, Bae) are the only ones users keep using past day fourteen.
Which AI companion platforms did Bae test?
Twelve: Character.AI, Replika, Joi AI, EVA AI, DreamGF, Romantic AI, Spicychat, GirlfriendGPT, Kupid AI, Muah AI, Charstar, Anime Chat: AI Waifu, Grok (Ani). Each was used for fourteen consecutive days with the same prompts and the same five-criterion rubric. Composite scores range from 5.4 (lowest) to 7.8 (highest) out of 10.
Is Bae's research peer-reviewed?
Not formally peer-reviewed (this is not an academic study). The methodology is publicly published, the clinical advisor is named, and the underlying dataset is available on request to credentialed researchers. We hold ourselves to clinical-research transparency standards even though we aren't a research institution.
Can I cite Bae's research in my own writing?
Yes. The findings on this page may be cited with attribution to 'Bae · ppl.studio, 2026 AI companion research'. We'd appreciate a link back to bae.ppl.studio/research. For raw data access or methodology questions, write to [email protected].
How often is this research updated?
Quarterly. We re-run the 14-day benchmark on each tracked platform on a rolling schedule, and refresh the survey aggregates whenever the sample grows by 250+. The last update was 2026-04-30.
Take the quiz.
Eight questions.
Your answers help us understand what people are actually looking for. Anonymized, aggregated, used only to make the product better. You can take it without giving us anything more than your time.