2025 AI Mental Health Apps: The Path to $1B Revenue

Editor’s Take · Perspective of an AI Serial Founder（Content distilled by AI; opinions belong to the original author. No need to read the source after this.）

CTest CEO Ren Yongliang reflects on 14 years of entrepreneurship, pivoting from astrology tools to AI-driven psychological companionship and now developing hardware. Key evidence: 50 million users (B·official claim), 1 billion-level revenue (B·headline claim), LOVOT priced at 30,000 yuan (B·citing competitor). For monetization, this is a “high emotional stickiness + data closed loop” moat case, suitable for teams with psychology backgrounds or strong content operations capability. The biggest pitfall lies in long-term retention of AI companionship and the cost of hardware mass production. Action point: study their “tool-to-platform” monetization rhythm.

• Break down CTest’s monetization milestones from free tool to two-sided platform
• Reference LOVOT’s 30,000-yuan price point for emotional hardware cost estimation
• Avoid pure astrology prediction; focus on AI emotional companionship and resonance
• Beware of compliance risks and medical boundaries in AI psychological services
• Enter via “low-frequency necessity,” then gradually increase user retention frequency

1. What Kind of Opportunity Is This

CTest started as an astrology fortune-telling tool and has now moved into AI psychological companionship and embodied intelligence hardware. Users pay for emotional value, while the company generates cash flow through content subscriptions, service commissions, and future hardware sales. At its core, this is a long-term companionship business driven by “high emotional stickiness data.”

2. Independent Judgment

Worth referencing cognitively, but not advisable to fully replicate. Their 14-year accumulated content barrier and 50 million user base are extremely difficult to copy. My independent view is that without a professional psychology background or unique large-model fine-tuning capability, purely building an “AI astrology” product will easily fall into a homogeneous red ocean and face medical compliance boundary risks.

3. Cold-Start Path

First validation step: use an open-source large model to fine-tune vertical psychology/astrology corpora, develop a single-function WeChat mini-program (e.g., “daily horoscope + emotional soothing”), and avoid building a bloated all-in-one app. Cost magnitude is tens of thousands of yuan (compute + outsourced frontend), cycle 1–2 months, focusing on testing user willingness to pay and next-day retention.

4. Biggest Risks and Pitfalls to Avoid

First, the medical compliance red line. If AI companionship crosses into psychological diagnosis, it will face regulatory crackdowns; you must strictly limit it to “entertainment and counseling,” avoiding any claims of therapeutic efficacy. Second, the hardware mass-production trap. LOVOT’s 30,000-yuan price point remains close to cost; if CTest enters hardware without massive capital investment, supply chain loss of control can easily trigger a broken cash flow chain.

5. Case Review (How Others Did It)

• Stage transitions: In 2011, they only built tools to accumulate data; in 2016, they pivoted to a two-sided platform by bringing in counselors; in 2019, they launched AI conversations, and today they are developing robots. The rhythm is “first build traffic, then build transactions, finally build hardware,” with each step relying on the data and cash flow of the previous one.
• Aquisition and retention: They切入 at “emotional troughs” (e.g., relationship or workplace anxiety), use low-barrier content like astrology to attract users, and provide high-frequency interaction through AI conversations. 80% of users are women, with 80% located in Tier 1–2 cities, precisely targeting high-paying-power, high-emotional-need groups.
• Data closed loop: They treat user birth data and daily mood scores as explanatory variables for “cycles” and “chaos systems”; the more data, the better the model understands users, forming a “more you use it, the more accurate it feels” experience moat, rather than solely relying on generic large models.
• Hardware positioning: Referencing LOVOT’s 30,000-yuan price point, they position it as an “companionship” interactive terminal, not a function machine. Ren Yongliang emphasizes that “interaction innovation” is the biggest change brought by large models; the core value of hardware lies in triggering oxytocin-based emotional connections, not computing power stacking.
• Strategic discipline: They refused early acquisition offers, preserving cash flow to support innovation. When facing BP collisions with tech giants (e.g., personalized recommendations), they did not panic, insisting on finding vertical pain points in the “cracks between market and technology,” such as the niche field of AI泛psychology.
• Pitfall warnings: Three consecutive Spring Festivals saw R&D nodes delayed due to large-model technology iterations, reminding hardware and software teams to decouple and avoid letting AI waves disrupt existing business rhythms.

6. Dual-Track Executability

Cross-border: feasible. Acceptance of AI companionship hardware is very high in Europe and America; you can directly borrow the LOVOT model, but you need to solve the market education problem for the 30,000+ yuan price point. I recommend starting with software export (TikTok traffic + subscription model) before launching hardware. Domestic: proceed with caution. Pure AI astrology content suffers from severe homogenization; I recommend entering细分vertical scenarios within compliance red lines, such as “cognitive companionship for the elderly” or “emotional counseling for adolescents,” and binding B-side institutions (schools, corporate EAP) to reduce C-side customer acquisition costs.

Original · LateTalk: Read original →

Related tool recommendation (promotion): Startup Toolbox Xiaoe Tech Column