In May 2026, we asked a small group of existing Claire users a direct question: does a scheduled voice conversation actually make reflection easier to begin and repeat? The answers were encouraging. They were also limited. We are publishing both parts.
Download the research packet (PDF) for a dated copy of the methodology, aggregate results, scoring notes, and limitations.
What we asked
The anonymous survey used 12 statements and a four-point scale from strongly disagree to strongly agree, with no neutral option. It covered trust, usefulness, ease of use, feasibility, technical friction, accessibility, and recommendation intent.
We invited 31 non-test users. Twenty-three had at least 10 active weeks with Claire when invited. The invitation tracker recorded 17 completions overall, a 54.8% completion rate, and 16 completions among the 23 users with 10 or more active weeks, a 69.6% completion rate.
| Measure | Count | Rate or note |
|---|---|---|
| Non-test users invited | 31 | 100% |
| Invitees with at least 10 active weeks | 23 | 74.2% of invitees |
| Completed invitees in the tracker | 17 | 54.8% of invitees |
| Completed invitees with at least 10 active weeks | 16 | 69.6% of experienced invitees |
| Known, non-test response rows | 18 | Not used as a completion rate |
| Response rows tagged to invitees with at least 10 active weeks | 17 | 94.4% of response rows |
The anonymous response table contains 18 known, non-test response rows, while the invitation tracker contains 17 completed invitees. Because the responses were deliberately separated from identity, we cannot reconcile that one-row discrepancy without weakening the anonymization boundary.
The clearest signal: reflection felt easier to repeat
All 18 response rows agreed that Claire helped them reflect more consistently. All 18 also agreed that it was easy to start or answer a Claire call, and that Claire's responses felt appropriate for what they shared.
Sixteen of 18 agreed that Claire made journaling easier than writing on their own. Seventeen of 18 agreed that they would recommend Claire to someone who wants to record reflections for therapeutic reasons. These are product-perception findings, not measures of treatment or clinical benefit.
| Survey statement | Strongly disagree | Disagree | Agree | Strongly agree | Favorable | Adjusted mean |
|---|---|---|---|---|---|---|
| I felt comfortable sharing personal reflections with Claire. | 1 | 0 | 5 | 12 | 17 of 18 | 3.56 / 4 |
| I trusted Claire to respond in a supportive and nonjudgmental way. | 2 | 0 | 1 | 15 | 16 of 18 | 3.61 / 4 |
| I understood that Claire is an AI tool, not a human therapist. | 0 | 0 | 1 | 17 | 18 of 18 | 3.94 / 4 |
| Claire helped me reflect more consistently. | 0 | 0 | 6 | 12 | 18 of 18 | 3.67 / 4 |
| Claire made journaling easier than writing on my own. | 0 | 2 | 3 | 13 | 16 of 18 | 3.61 / 4 |
| Claire's responses felt appropriate for what I shared. | 0 | 0 | 6 | 12 | 18 of 18 | 3.67 / 4 |
| It was easy to start or answer a Claire call. | 0 | 0 | 7 | 11 | 18 of 18 | 3.61 / 4 |
| The time required to use Claire felt manageable. | 0 | 0 | 10 | 8 | 18 of 18 | 3.44 / 4 |
| Technical issues made it harder for me to use Claire. | 5 | 11 | 1 | 1 | 16 of 18 | 3.11 / 4 |
| It was hard to find a private time or place to use Claire. | 7 | 8 | 3 | 0 | 15 of 18 | 3.22 / 4 |
| Claire could be useful for veterans who may benefit from private, voice-based reflection. | 1 | 0 | 6 | 11 | 17 of 18 | 3.50 / 4 |
| I would recommend Claire to others who want to record their reflections for therapeutic reasons. | 1 | 0 | 3 | 14 | 17 of 18 | 3.67 / 4 |
The table reports aggregate counts only. It does not reproduce any participant's open-text feedback or row-level response pattern. Optional feedback is summarized by theme below. Favorable means agree or strongly agree. For the two negatively worded statements about technical issues and finding a private time or place, favorable means disagree or strongly disagree, and the mean is reverse scored so that a higher number is favorable for every row.
The result we trust most is not a sweeping claim. It is a clear signal from a small group of experienced users: removing the blank page made reflection easier to repeat.
Trust started with a clear boundary
Every response agreed with the statement that Claire is an AI tool, not a human therapist. That item received the survey's highest average score: 3.94 out of 4. Sixteen of 18 also agreed that they trusted Claire to respond in a supportive and nonjudgmental way.
We do not read those scores as permission to blur the line. The boundary is part of the trust: Claire can support a reflection habit, but it does not diagnose, provide treatment, or replace a therapist, crisis service, or human relationship.
Where responses were less uniform
The statement “Technical issues made it harder for me to use Claire” had an adjusted mean of 3.11 out of 4. Sixteen of 18 respondents disagreed, while two agreed. The statement “It was hard to find a private time or place to use Claire” had an adjusted mean of 3.22. Fifteen respondents disagreed and three agreed. Both were negatively worded, so disagreement was the favorable response.
A scheduled call only helps when it arrives reliably and fits into a real day. The technical-issues responses give us a specific area to investigate. The private-time-or-place responses describe the circumstances in which people used Claire; they do not, by themselves, establish a product privacy problem.
What users suggested in optional feedback
Nine of the 18 known, non-test response rows included nonempty optional feedback. We grouped those responses by theme and paraphrased them so no participant's response is reproduced. Five entries contained an identifiable product suggestion:
- Clarify data access. One entry asked for clearer information about who, beyond the user, could access saved or processed recordings, transcripts, or derived insights. This is separate from the scored statement about finding a private time or place.
- Pause an interrupted call. One entry suggested a pause-and-resume option when a conversation is interrupted.
- Make schedule changes faster. One entry suggested a quicker way to change the call time after missing consecutive calls.
- Remember what was already discussed. One entry asked Claire not to repeat a scheduled check-in question when that topic had already come up earlier in the conversation.
- Offer an optional backstory. One entry suggested a personal backstory for Claire to create more relational continuity.
Of the remaining four entries, two offered praise without requesting a change, one reported no applicable feedback, and one positively commented on intimate questioning. Each theme above came from one optional-feedback entry, so these are ideas to evaluate, not majority preferences.
What this survey cannot tell us
This was not a randomized or controlled study. It did not measure symptoms, diagnoses, or clinical outcomes. The sample was small, self-selected, and made up mostly of experienced users; 17 of the 18 response rows were tagged to invitees with at least 10 active weeks. The demographic mix was also not representative enough to support broad population claims.
Nonresponse matters, too. Seventeen of 31 invitees were marked complete in the tracker. People who chose to answer may have had a different experience from those who did not. That is why we are using these results as directional product evidence, not as proof that Claire works for everyone.
The honest takeaway
The survey does not justify a sweeping wellness claim. It does support a narrower idea worth continuing to test: for this small group of existing users, replacing the blank page with a scheduled conversation made reflection feel easier to start, easier to repeat, and worth recommending. We will keep publishing what we learn, including the friction and the limits.