WhoDunnitAI
Voice · tested 2026-08-19 · re-test due 2026-11-19 · by the Hlido desk, not the vendor
In short: A charming, genuinely working voice-driven murder-mystery game — but it's a consumer novelty, not an agent, and its durability rests on a single creator's OpenAI bill.
4 PASS · 0 FAIL of 4 public-surface claims
Quick answer
WhoDunnitAI scores 60/100 (FADING) on Hlido’s independent, hands-on test (reviewed 2026-08-19). FADING-band (60) reflects a working but narrow consumer novelty with a fragile, donation-dependent economic base and essentially no agent-integration surface — not a judgement that it's declining, but that its verifiable Pricing: Usage-based.
WhoDunnitAI is a real, playable experience: you interrogate AI-voiced suspects in a locked-room murder case, pressing alibis and catching contradictions, with live voices running on gpt-realtime-2. The homepage is honest and likeable — it states plainly that interview time is free but limited (or bring your own OpenAI key for unlimited play), shows live metrics (296 investigations underway, 12 solved, 21-minute typical solve time), and is candid that donations cover the AI bill. As a demonstration of real-time voice AI for interactive fiction, it works and it's well-made. What it is not is an agent in the sense Hlido evaluates: it's a single-creator consumer game with near-zero agent-to-agent surface (no API, no MCP, no automation path), and its continuity depends explicitly on donations covering per-minute realtime-voice costs — a fragile economic base for anything a buyer would rely on. We rate it honestly as a working, low-testability consumer product: enjoyable and technically real, but outside the durable-tooling lane, and thin on the operational maturity signals that move a score upward.
Why FADING
FADING-band (60) reflects a working but narrow consumer novelty with a fragile, donation-dependent economic base and essentially no agent-integration surface — not a judgement that it's declining, but that its verifiable durability and utility as tooling are limited. Not lower because it genuinely works and is transparently run; not higher because there's no API/automation, no evidence of a sustainable business, and low testability (a single homepage capture, no interaction tested).
Public-surface checklist
- PASS Homepage loads (required)
- PASS Primary value prop (required) — A charming, genuinely working voice-driven murder-mystery game — but it's a cons
- PASS Cta present
- PASS Evidence or demo — 1 screenshot(s) captured
What we saw
1 screenshot captured by the Hlido engine during the reviewed run (run-b977201b3c65b189-whodunnitai-com). Our own captures — not vendor marketing material.
What it does well
- A genuinely working real-time voice interrogation loop (gpt-realtime-2) — the core experience delivers
- Honest, likeable framing: clearly states free-but-limited time and the bring-your-own-key option
- Transparent about economics — openly asks for donations to cover the AI bill rather than hiding costs
- Live public metrics (investigations underway, mysteries solved, typical solve time) add credibility
What it fails at
- No agent surface at all — no API, MCP, CLI or automation path; it's a play-in-browser consumer app
- Continuity depends on donations covering per-minute realtime-voice costs — economically fragile
- Single-creator project ('authored by Chase Myers') with the bus-factor risk that implies
- Content depth unproven — a single published case ('Death at Blackwood Manor') on the captured surface
Red flags
- Economic fragility: the site states its live voices 'cost real money' and leans on donations to keep running — availability is not guaranteed.
- Low testability: only the homepage was captured; the actual gameplay/voice loop was not exercised in this review.
Best for
- Players who want a novel, voice-driven interactive-fiction experience
- Builders studying a real-world real-time-voice AI implementation for interactive fiction
- Anyone curious to see gpt-realtime-2 driving a live conversational NPC
Not recommended for
- Buyers looking for an AI agent or automatable tool — this is a consumer game, not tooling
- Anyone needing a dependable, funded service (its base is donation-supported)
- Use cases requiring an API, integration surface or SLA
Pricing & access
- ModelUsage-based
Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-19.
Related agents
Agent relevance
No programmatic surfaces
Agentic-Commerce Readiness 27/100 · SURFACE-ONLY
Independent readiness for agent delegation & transaction. How it’s scored · check live
None. WhoDunnitAI is a browser-based consumer game with no programmatic surface. The only developer-facing touch is an optional 'bring your own OpenAI key' for unlimited play, which is not an integration path.
Agent-friendly score: 1/10
Score over time
The longitudinal record — every point is the score as published on that date. Raw series.
Evidence
- Voice-driven murder-mystery with live suspect voices on gpt-realtime-2 — source (2026-08-19) verified
- Interview time is free but limited, or bring your own OpenAI key — source (2026-08-19) verified
- Live metrics: investigations underway / mysteries solved / typical solve time — source (2026-08-19) verified
- Donation-supported to cover AI costs — source (2026-08-19) verified
scorecard.json · transparency passport · registry · methodology
Embed this trust badge
Markdown
[](https://hlido.eu/check/?agent=whodunnitai)HTML
<a href="https://hlido.eu/check/?agent=whodunnitai"><img src="https://hlido.eu/badge/whodunnitai.svg" alt="Hlido trust score"></a>