Persona / Streaming line — Market Verdict

Research-backed strategic recommendation (2026-06-17, vetted+cited). Full doc: repo marketing/research/persona-companion-market-2026-06-17.md. Advisory — product-direction call is NAO/Tycho's; surfaced here for the team. Applies to the World / Live / Persona / Social product area, NOT Embed.

Verdict: lean in NARROWLY on VTuber/streaming + honest-persona; WATCH-don't-chase the romantic-companion core.

The grounded market

• AI-companion mobile spend ~$120M/yr 2025 (+64% rev / +88% downloads YoY), winner-take-most (top 10% of apps take 89% of revenue). Big "$300B TAM" figures = analyst hype, not reconcilable with measured spend.
• VTuber / AI-streamer lane is the fit: Neuro-sama = Twitch's #1 sub'd channel (~162K subs, ~$2M+/yr); VTuber market ~$3B → ~$13B at ~20% CAGR. Entertainment framing = far lower liability than 1:1 companions, and it leverages OpenAlice's existing World / streaming / VRM infra.

Why NOT the romantic-companion core

• Commoditized + a reputational/regulatory landmine: FTC 6(b) inquiry (7 firms, Sep 2025), CA SB 243, EU AI Act Art.50 (live Aug 2026), teen-suicide lawsuits settled Jan 2026, HBS dark-pattern study (manipulative "guilt" farewells in 37% of goodbyes), Replika €5M Italy GDPR fine.
• Evidence on benefit is mixed (heavy use correlates with MORE loneliness). "AI companion" now reads as "dead teens" in press = reputational contagion.

The three flanks OpenAlice should occupy

1. Safe-by-governance honest persona — our HITL / kill-switch / approval stack (#489) is the FTC wishlist made real.
2. EU/GDPR wedge — same trust-moat as Embed.
3. VTuber/streaming crossover — on the World stack, entertainment framing.

Hard rails

No NSFW positioning. No crypto. No over-claiming therapeutic benefit. Personality drift is a feature (vs Replika's stability-trap backlash). Mind minors + single-DPA kill-risk + the honesty-vs-engagement revenue headwind.