Viral Generator Builder

Scope Boundary

Generator Type Decision

First match based on business goal.

The Viral Loop Architecture

Every generator follows the same 5-stage loop. Optimize each stage independently.

Stage 1: LAND (visitor arrives)
  |
Stage 2: INPUT (visitor provides minimal data)
  |
Stage 3: PROCESS (the "magic" -- your algorithm)
  |
Stage 4: RESULT (personalized, shareable output)
  |
Stage 5: SHARE (result posted to social/messaging)
  |
  +---> New visitor lands (Stage 1) -- loop repeats

Result Design Principles

The result page IS the product. Everything else is scaffolding to reach it.

Input Optimization

Every input field is a dropout point. The conversion cost of each additional field is ~10-15% of remaining users.

The first-question rule: The first input/question must be answerable in <3 seconds and feel relevant to the promised result. "What's your name?" or "Pick the image that speaks to you" -- not "Describe your work habits in detail."

Algorithm Selection

Algorithm gotcha: Weighted scoring quizzes need result distribution testing. If 60% of test-takers get the same result, the quiz feels broken. Test with 100+ random answer combinations. Adjust weights until no result exceeds 35% frequency and no result is below 8%.

Monetization Decision

Anti-Patterns

1. The Generic Result

Results that could apply to anyone. "You're creative and sometimes prefer being alone." Zero identity signal, zero share motivation. Symptom: share rate below 3%. Fix: Each result should make 70% of non-recipients say "that doesn't describe me at all."

2. The Friction Wall

Too many inputs before any payoff. 15-question quiz with no preview of what the result looks like. Users have no reason to invest time without knowing the reward. Symptom: >50% drop-off before completion. Fix: Show an example result on the landing page. "See what type you are in 5 questions."

3. The Invisible Share

Share buttons buried below the fold, no pre-written share text, no OG image configured. The result page works great -- but nobody shares because sharing is hard. Symptom: high completion rate but share rate <2%. Fix: Share button at eye level on result page. Pre-fill share text. Generate unique OG image per result.

4. The Ugly Card

OG image is auto-generated with no design thought. Default gray box with tiny text. In a social feed, it looks like spam. Nobody clicks. Symptom: share-to-click ratio below 1%. Fix: Design OG images at 1200x630px. Bold text, strong colors, readable at thumbnail size. Result visible without clicking.

5. The One-and-Done

No reason to return or re-engage. User takes quiz once, shares, never comes back. All viral value captured in one cycle. Symptom: 0% return visitor rate. Fix: Daily/weekly changing generators, "compare with friends" feature, new quiz variants, result collections.

6. The Random Disappointment

Non-deterministic results that change on reload. User shares "I got Midnight Architect," friend tries with same inputs, gets different result. Trust destroyed. Symptom: comments saying "I got something different when I tried again." Fix: Hash-based determinism for input-dependent generators. Seed randomness with user input.

Rationalizations

• "More questions means more accurate results" (The Friction Wall -- accuracy doesn't matter if nobody completes it)
• "People will find the share button" (The Invisible Share -- they won't, they'll screenshot and crop your branding out)
• "The result content matters more than how it looks" (The Ugly Card -- on social feeds, visual presentation IS the content)
• "We can add sharing features later" (sharing is architecture, not a feature -- retrofitting viral loops is 5x harder)
• "AI-generated results are always better than templated ones" (The Random Disappointment -- LLM results are non-deterministic, which breaks the verification loop)

Red Flags

• No OG image configured for result pages
• Share rate measured but share-to-click rate not tracked
• Quiz has more than 10 questions without proven completion rate data
• Results page loads in >2 seconds (every second of latency = 7% fewer shares)
• No mobile-specific share UX (>60% of generator traffic is mobile)
• Same share text for every result type
• Generator works only in English but targets global audience

NEVER

• NEVER show the full result in the OG image/share card -- show enough to create curiosity, not enough to satisfy it
• NEVER randomize results without seeding from user input -- non-deterministic results kill the viral verification loop
• NEVER gate the PRIMARY result behind email/payment -- gate the DETAILED breakdown, not the headline result
• NEVER launch a generator without testing result distribution with 100+ simulated inputs
• NEVER skip mobile optimization -- generators are impulse content consumed on phones during social media browsing