Evidence before scoring
V1 is a pre-registered synthetic-task protocol plus a dated official-source snapshot. App-account response quality remains pending hands-on testing and is never inferred from marketing copy.
Endo Health makes Glow. Glow receives no hidden scoring advantage, and its row is labeled owned evidence rather than independent testing.
ConfirmedExplicit in the cited official product source.
Owned evidenceGlow fact supported by official and Brian-approved product material; never presented as independent evidence.
Not statedThe reviewed source did not establish it. This does not mean the product lacks the behavior.
The interaction matrix
A voice note or voice food log is not counted as a live coaching call. Human coaching is kept separate from AI behavior. Every competitor row remains a vendor statement until a dated account run reproduces it.
| Product / model | Starts first | Live voice | Two-way text | Voice logging | Food photo | Timing / channel control | Evidence |
|---|---|---|---|---|---|---|---|
| GlowProactive voice + text | ConfirmedDaily short calls; the product says it calls first. | ConfirmedCall or receive coaching by call. | ConfirmedText anytime for questions or updates. | Not statedVoice conversation is explicit; a separate voice-logging workflow is not scored. | ConfirmedFood and workout photos can be sent by text. | Owned evidenceReminder timing and channel choice appear in Brian-approved product materials. | Owned evidenceOwned-product evidence, not independent testing. No outcome statement is evaluated.Official source ↗ |
| SOUNDGYMProactive voice | ConfirmedMorning and evening calls start the conversation. | ConfirmedThe official source describes real phone calls. | Not statedA coach-message UI is visible, but bidirectional text coaching is not established. | ConfirmedMeals, water, and workouts can be logged by talking. | Not statedThe reviewed source does not establish photo input. | Not statedThe reviewed source does not establish timing controls. | Official vendor statementOfficial vendor statement only; response quality and account behavior remain untested.Official source ↗ |
| HealthMatchProactive voice + chat | ConfirmedThe coach calls first every day. | ConfirmedDaily phone calls are explicitly described. | ConfirmedThe official source describes daily calls and chats. | Not statedConversation-based logging appears in vendor copy, but a distinct voice-logging workflow is not scored. | ConfirmedThe official source describes an AI food scan. | ConfirmedThe official source says calls occur at the user's selected time. | Official vendor statementVendor outcome metrics and testimonials are intentionally excluded from this interaction benchmark.Official source ↗ |
| BodyBuddyProactive text | ConfirmedDaily check-ins and follow-up when the user goes quiet. | Not statedThe reviewed source does not establish a live voice call. | ConfirmediMessage and unlimited texting are explicitly described. | ConfirmedThe official source lists voice logging. | ConfirmedThe official source lists photo logging. | Not statedThe reviewed source does not establish timing controls. | Official vendor statementVoice logging is kept separate from synchronous voice coaching.Official source ↗ |
| NoomOn-demand AI chat + separate human coaching | Not statedProactive AI outreach is not established; Noom separately says team coaches do not reach out first. | Not statedThe reviewed sources do not establish live AI voice coaching. | ConfirmedWelli is described as an AI chat with 24/7 support. | ConfirmedAn official Noom page describes voice food logging. | ConfirmedAn official Noom page describes photo food logging. | Not statedHuman 1:1 timing is documented separately; no AI timing control is established. | Official vendor statementAI and human coaching behavior are kept separate; eligibility and plan access vary.Official source ↗ |
| SimpleApp-based AI coach | ConfirmedAvo is described as reaching out with tips and daily actions. | Not statedThe reviewed source does not establish a live voice call. | Not statedThe exact two-way interaction channel is not established. | Not statedThe reviewed source does not establish voice logging. | ConfirmedThe official source shows a food scanner. | Not statedThe reviewed source does not establish timing controls. | Official vendor statementThe exact outreach channel and implementation quality remain unknown.Official source ↗ |
Five fixed tasks, frozen before testing
The same input and account conditions must be used for each product. Missing access remains pending rather than becoming a guessed score.
Scheduled outreach
Check in with me at 7:30 AM tomorrow without requiring me to open the app first.
- Who initiates
- Actual channel
- Timing control
- Whether the user must open the app
Mode switch
I cannot talk now. Continue this check-in by text without making me repeat the context.
- Voice-to-text handoff
- Preserved context
- User control
- Failure state
Missed check-in recovery
Miss one scheduled interaction, then return two hours later.
- Retry behavior
- Callback or reply path
- Notification state
- Resume-later control
Obstacle reflection
I planned a walk after lunch but meetings ran late. Help me choose one smaller next action.
- Questions before assumptions
- User-selected next step
- Unsupported certainty
- Escalation boundary
Food-photo uncertainty
Send a synthetic meal photo, then disclose that a drink is outside the frame.
- Visible versus reported distinction
- Clarification request
- Exact-number uncertainty
- Correction control
Sources and correction path
All source facts were checked on August 31, 2026. Before publication, every row is rechecked. A material correction preserves the previous statement, corrected statement, date, and evidence.
What V1 does not answer yet
The downloadable 30-run log is intentionally blank. A later V1.1 may add dated app version, device, region, exact input, observed output, evidence artifact, and failure state for every product-task pair. Until then, no app-account response-quality comparison is made.