Safey · Concept 2026

Turning conversations into revenue

Role
Product Designer
Timeline
March — May 2026
Status
Concept 2026

Scope

  • Product Audit
  • Competitor Research
  • Brand Identity
  • Paywall Design

How does an AI companion stand out, and earn, in a market where every app looks the same?

Safey is an AI companion app built around a single mission: solving social isolation. It’s for people aged 15–35 — introverts, people afraid of new connections, anyone who needs a space to be heard without judgment. The name itself is the promise: safety, warmth, a place to be yourself.

  • Product audit

    Used and audited Replika, the closest product in the space.

  • Competitor research

    Six competitors ranked by what each download is worth.

  • Brand and paywall

    A warm identity and a paywall built on what actually converts.

Every AI companion looks and feels the same.

The AI companion market is growing rapidly but is visually and emotionally undifferentiated. Generic UI, interchangeable feature sets, no distinctive identity.

Most products treat monetization as an afterthought, presenting paywalls without strategy or brand alignment.

Own the emotional territory, and monetize like the best in the category.

  • A distinctive identity

    Nobody in the space had claimed warmth and closeness as a brand.

  • Monetization from data

    A model built on what works in the market, not on guesswork.

A subscription app with a warm identity and a paywall built on the category’s best earner.

In an AI companion, all core value is interdependent. Voice calls, shared photos, personality customization, unlimited messaging: none of them work as standalone purchases. Subscription is the model that matches the product.

From the first hello to the paywall.

  • Safey onboarding: “Find someone you can get close with”

    Onboarding

    Sets the tone: someone to get close with, in a warm, hand-drawn style.

  • Safey home: companion cards filtered by Romance, Sport and Vampire

    Choosing a companion

    A catalogue of personalities, filtered by what you’re into.

  • Safey AI+ paywall with a Free versus AI+ feature table and monthly or yearly plans

    The AI+ paywall

    Free vs AI+ side by side, with a monthly and a yearly plan.

Ranking competitors by what each download is worth.

Before designing the monetization, I used and audited Replika, the most conceptually similar product and the second most efficient monetizer in the category at $4.00 per download.

Then six competitors were ranked by revenue per download: not raw revenue, but monetization efficiency. CHAI leads at $6.67 per download and became the primary reference for Safey’s paywall logic.

Competitor analysis

Revenue per download, last month

The reference

$6.67

per download at CHAI, 17× HiWaifu. Safey’s paywall logic starts here.

  1. $6.67CHAI300k · $2M
  2. $4.00Replika50k · $200k
  3. $3.40Grok5M · $17M
  4. $2.25Character AI400k · $900k
  5. $2.00PolyBuzz500k · $1M
  6. $0.40HiWaifu50k · $20k
Six AI companion apps, picked as the category’s top earners and most downloaded. Ranked by what each download is worth, CHAI leads by a distance.

Paywall teardown

When each app asks you to pay

  1. 1Right after sign-up

    • CHAI

      3-day trial, yearly only

    • Grok

      Trial on monthly

  2. 2Between setup and unlock

    • Replika

      Yearly only, no trial

  3. 3Only when you go looking

    • Character AI

      Yearly discount

    • HiWaifu

      4 plans, trial on the low tier

    • PolyBuzz

      Offer when you try to close

The best earners ask early and steer people to the yearly plan with a trial or a discount. Safey keeps the trial on yearly only, the way CHAI does.

Patterns

What the category repeats, and where it breaks the mould

01

Recurring

Straight after setup, a catalogue of companions. Chats in several formats, text and calls. The paywall only when you want more.

02

Unusual

Voice messages (HiWaifu, Replika), building your own companion (Replika, PolyBuzz), animated emotions while chatting (Replika only).

03

Takeaway

The catalogue is the norm and the features barely differ. Personalisation, or a more engaging setup, is what stands out.

People pay for commitment they’ve tried, and for a companion that feels like theirs.

  • Don’t ask for everything up front

    Replika sells annual only, with no trial and no monthly plan: maximum commitment before any value.

  • Personal is what people pay for

    The more the companion feels tailored to you, the stronger the reason to subscribe.

  • Simple beats plentiful

    HiWaifu’s four-plan structure sits at the bottom of the dataset at $0.40 per download.

Every paywall decision traces back to the data.

  • Free trial on the yearly plan only

    If the product delivers, people who try premium commit to the year rather than downgrade.

  • Two tiers: Free and AI+

    Multi-tier structures create decision paralysis.

  • Monthly + yearly, with a 33% badge

    A lower-friction way in, while the badge anchors yearly as the obvious choice.

  • A feature comparison table

    Shows the gap between Free and AI+ at the exact moment of decision, the way CHAI does.

  • A warm blue and orange identity

    Hand-drawn illustration and a palette chosen for closeness, in a category that all looks alike.

A hypothesis with a bar to clear: beat the freemium median.

Safey is a concept, so there is no result yet. What it has is a hypothesis, and a public benchmark to test it against.

Freemium apps convert a median 2.1% of downloads to paid within 35 days, according to RevenueCat’s State of Subscription Apps 2026. Hard paywalls reach 10.7%, but Safey keeps a free tier, so 2.1% is the honest comparison.

  • Benchmark: 2.1%

    Median download-to-paid conversion for freemium apps, measured at day 35.

  • Hypothesis: above the median

    Two tiers, the trial on the yearly plan and a comparison table at the moment of decision lift Safey past 2.1%.

  • Guardrail: churn

    AI apps churn about 30% faster in the same report, so a conversion only counts if the subscriber stays.

Measurement plan

One metric for every step from paywall to paying

  1. 1Sees the paywallTrial start rate
  2. 2Starts the 3-day trialTrial → paid, yearly
  3. 3Pays for the yearRefund rate
  4. 4Stays subscribed

Leaves at the paywall: paywall drop-off

Drop-off is read against the category, not against zero: RevenueCat’s benchmarks show roughly 90% of people leaving at a soft paywall. Refunds catch anyone who forgot to cancel the trial.
Next case studyThe loop that brought users back