🧠 CONDITIONAL

AI Psychologist Bot

Telegram bot with AI therapist: support, CBT techniques, and self-reflection

Category: health
Date: 2026-03-16
ID: b4abe88d…
AI Score ?Idea potential score (0-100). The final verdict (GO / CONDITIONAL / NO-GO) is the qualitative consensus of 5 AI models based on all risks and conditions - it can differ from the numeric score alone.
61/100
⚠ Conditional GO — risks exist, hypotheses need validation
Financial Dashboard — Key Numbers
Investment Required ?Total capital needed to reach break-even: servers, marketing, development.
$155
to get started
Break-even ?The month when monthly profit will cover all startup costs.
Month 3
from launch
MRR Target ?Monthly recurring revenue at which the project is considered successful and ready to scale.
$3,000
per month
Margin ?Share of each dollar remaining after infrastructure, APIs, and direct costs. 70%+ is healthy for SaaS.
80%
of revenue retained
Monthly Revenue Growth Forecast
Unit Economics — Numbers per Customer
Revenue per Customer ?How much one customer pays per month (ARPU). The higher, the fewer customers needed for strong profit.
$15/mo
per customer avg.
Profitability ?Share of each dollar remaining after servers, APIs, and other direct costs.
~92%
of revenue retained
Customer Lifetime Value ?Total revenue from one customer over the entire relationship (LTV). Ideally 3× above acquisition cost.
$72
lifetime total
Max Acquisition Cost ?Maximum ad spend per customer while keeping the business model profitable (CAC target).
$20
per new customer
LTV / CAC ?Ratio of customer lifetime value to acquisition cost. 3× and above is healthy.
3.6×
✓ Above benchmark
Development Scenarios
Scenario Revenue by Month 6 Revenue by Month 12 Key Assumption
Pessimistic $646 $928 CAC higher than forecast, conversion below 5%
Realistic $1,847 $2,652 On plan: CAC at or below target, churn at or below 5%/month
Optimistic $4,617 $6,630 Virality kicked in, CAC came in 2x below target
Why This Verdict

✓ Arguments FOR

  • A huge market with proven willingness to pay
  • 24/7 availability vs $100-200/session with a live psychologist
  • High retention with the right progress gamification

✗ Why not higher

  • Ethical risks when working with vulnerable users
  • Regulatory requirements in the mental health field
  • High competition: Woebot, Wysa, Replika, BetterHelp
🛑 When to Stop the Project
K1
No conversions after 100 clicks
leads == 0 AND clicks >= 100
→ pause traffic, diagnostics
K2
CAC exceeded LTV × 0.5
cac > ltv * 0.5 AND leads >= 5
→ stop traffic
K3
Two Gate failures
gate_failures >= 2
→ close project
What to Validate Before Scaling
1
Users in the situation of 'not ready for therapy but need support' are an underserved segment
2
Anonymity and 24/7 availability lower the entry barrier compared to a real psychologist
3
CBT techniques in a chatbot format have proven effectiveness (Woebot studies)
What the AI Models Said
~
Claude Opus (Critic)
An AI psychologist carries serious ethical and legal risks. For real mental disorders or suicidal thoughts, AI should not be the only source of help. Regulatory pressure is growing.
~
GPT-4.1 (Market Strategist)
The mental health tech market will reach $5B by 2027. With the right positioning ('AI coach for support, not a replacement for a psychologist'), this is a real product with high retention. Woebot and Wysa proved the model.
~
Grok-3 (Technical Analyst)
Mental health apps are one of the fastest-growing categories. 24/7 availability, anonymity, and a $10/month price versus $100/hour for a live psychologist make a strong value proposition. Calm and Headspace are valued at $1B+.
Milestones & Stages
M1
Development of the AI therapist core
Done
M2
Integration with the Telegram platform
Done
M3
Reflection and self-analysis feature
In Progress
M4
Testing with a focus group
Pending
M5
MVP launch in the App Store
Pending
Investment & Exit Scenarios

Total Investment Needed

$1,000
to reach profitability
Marketing $500
Development $250
Infrastructure $150
Operations $100
🚦 Strengths & Risks at a Glance

✓ Green Flags

  • Woebot shut down in June 2025, opening the niche for an LLM-based solution
    1.5M former users
  • TAM of $7.4B with a CAGR of +15.3% per year, the mental health app market is growing
    $7.4B TAM
  • 92% margin after LLM costs, infrastructure at $50/month
    92% margin
  • Telegram-native means bypassing the app store, full anonymity
    no App Store friction
  • Replika precedent: 2M MAU, $24M/year revenue on a freemium model
    2M MAU, $24M/year
  • Breakeven in month 6 with accumulated cash flow
    M6 breakeven
  • Acquisition cost of $17-333 (infrastructure plus outreach), LTV/CAC = 7.4x
    7.4x LTV/CAC

✗ Red Flags

  • H1, willingness to pay $15/month, has not been market-tested, a landing page test is needed
    high
  • Freemium conversion of 1.5% (adjusted down from 3%), industry average is 2.18%, this may not be enough
    high
  • High churn in mental health apps, Day-7 retention below 20% is a stop criterion
    high
  • ChatGPT competes for free, within 12 months a GPT-6 free tier could appear
    high
  • Character.AI lawsuits in 2025-2026 over product liability set a precedent for regulating AI mental health
    high
  • Churn of 12% in M1-M3 (adjusted up from 8%), which means LTV is only $125 over 8.3 months
    medium
  • Maximum loss of $105 in the first 2 months, requires starting capital
    medium
⚠️ Risk Matrix
Risk Probability Impact Mitigation
Legal liability during user crises
🛑 Kill trigger
 35% High 3-level crisis detection, a static response with hotline numbers, positioning as a 'coach' rather than medical care
Character.AI precedent for regulating AI mental health
🛑 Kill trigger
 60% High Ban on minors (18+), terms of service, insurance
ChatGPT as a free competitor
 85% High Specialization plus personalization plus memory history plus Telegram convenience, high switching cost
Freemium conversion below 2% after 500 users
🛑 Kill trigger
 45% High A/B test hard paywall (trial) vs freemium, pivot to a 7-day trial
High churn, Day-7 retention below 20%
🛑 Kill trigger
 50% High Improve onboarding, personal history memory, mood patterns
LLM budget overrun (over 30% of revenue)
 25% Medium Claude Haiku by default, hard session limits, switch to a cheaper model
Acquisition cost above $50 through organic channels
🛑 Kill trigger
 40% Medium Diversify channels (Reddit, Product Hunt, SEO, Yandex), focus on virality and word-of-mouth
GDPR non-compliance when serving EU users
 30% Medium Data minimization, a /delete_my_data command, privacy policy
💸 Monthly Cash Flow (Realistic Scenario)
Period Revenue Expenses Net Cumulative
M0 / Launch $0 $50 -$50 -$50
M1 $45 $100 -$55 -$105
M3 $180 $150 +$30 -$75
M6 ✅ BREAKEVEN $450 $250 +$200 +$125
M9 $900 $400 +$500 +$625
M12 $1,575 $600 +$975 +$1,600
M18 $2,925 $1,200 +$1,725 +$3,325
🏁 Competitive Landscape
📡 Market catalyst: Woebot vacated the niche in June 2025, Character.AI set regulatory precedents in 2025-2026, CAGR +15.3%/year
Competitor Size Take Rate Weakness
BetterHelp
 scale unknown N/A
Live therapist at $65/week, expensive for the mass market, no AI automation
Wysa
 scale unknown N/A
B2B model, FDA approval slows things down, $99.99/month is expensive
Replika
 2M MAU N/A
A companion, not a coach; positioned around romance, not mental health; $14.99/month
Headspace
 scale unknown N/A
Meditation, not CBT/ACT/DBT; $12.99/month; not specialized
ChatGPT
 N/A 0%
Free, but generic, no structured methodology, no user history memory
Woebot (SHUT DOWN 06.2025)
 1.5M former users N/A
Shut down, but left a vacuum; was scripted CBT, not LLM-powered
🛠 MVP — Week-by-Week Plan
Week 1
  • Python + aiogram 3.x Setup
  • Claude Haiku integration
  • One CBT dialogue module
Bot responds to first messagesCrisis detection works
Week 2
  • 3-level crisis detection (keywords, semantics, hotlines)
  • Supabase JSONB user memory
  • Freemium: 20 messages/day
  • Paddle integration for Premium
Crisis detection covers 95% of casesPaddle checkout worksHistory memory is accumulating
Week 3
  • First traffic: Telegram outreach to channels
  • Reddit, useful posts
  • Landing page: rimen.pro/landings/ai-coach/
200+ first users5%+ fill out the landing page form = GO to full MVP
Week 4
  • Product Hunt launch
  • Validate freemium to premium conversion
200-500 visits from Product HuntConversion of 1.5%+ to paying users
Week
  • A/B test: freemium vs hard paywall (7-day trial)
  • Add ACT methodology
  • Analyze Day-7 retention
Day-7 retention >20%Trial conversion above 12% (if pivoting)Net result is positive user feedback
Week
  • Scaling: Telegram Ads after $500 MRR
  • SEO content: 10 articles on CBT techniques
MAU 800, 12+ paying$500 MRR reachedOrganic traffic is growing
🏰 Competitive Moat

✗ Easy to Copy

  • Telegram bot architecture
  • Basic freemium model
  • Crisis detection prompts
  • Basic CBT questions

✓ Hard to Copy

  • Accumulated history plus user memory (switching cost)
  • Brand trust in mental health (built over a long reputation)
  • Mood patterns plus personal understanding (hard to model)
  • Network effect on Telegram (invites from existing users)
⏱ Moat forms by: M6+ (trust plus personal history); M12+ (strong moat through accumulated data)
📊 Acquisition Cost by Channel
Channel CAC Notes Profitable?
Telegram outreach (organic) $17–50 An automatable channel, infrastructure cost only. $50/month hosting divided by 3 paying users = $17, or outreach at $5/user divided by 1.5% conversion = $333 at most ✓ Yes
Reddit plus forums $0 One-time setup, useful posts, high intentionality, 50-150 users per month ✓ Yes
Product Hunt $0 Launch-day effect, 200-500 visits in a day, free visibility ✓ Yes
Telegram Ads $10–25 Scalable after $500 MRR, a paid channel, requires budget ✗ No
SEO (long-term, Yandex plus Google) $0–3 6+ months until first traffic, 100-1000 users per month by month 6+, low CAC ✓ Yes
🔬 Anti-Optimism Audit
1
Freemium conversion assumed at 3%
→ Reality: industry average is 2.18%, conservative estimate is 1.5% after the Anti-Optimism Audit
-15 points (from 61 to 52)
2
Churn in months 1-3 was assumed at 8%
→ Reality: mental health apps typically see 12% monthly churn (higher than casual apps)
-10 points
3
Breakeven assumed at month 3 with full scaling
→ Reality: breakeven at month 6 with accumulated cash flow, peak loss of -$105 instead of -$60
-20 points
4
Cumulative profit by M18 assumed at $35K (optimistic forecast)
→ Reality: $3.3K (Floor-realist), a 10.5x overestimate
-20 points

Have your own idea?

5 AI models attack it from all angles and deliver an honest GO / NO-GO verdict with real numbers in 24 hours.

Validate for $39 →