Apple requires in-use screenshots · How an AI wrapper made $130K

The Mobile Takeaway
For people who ship apps
Thursday, November 12, 2026

Apple now requires that more than half of your screenshots show the app in use, a rule they'll check for every device and localization.

Bogdan Dragomir explains how he scaled an AI wrapper to $130,000 in four months by pairing niche landing pages with $30,000 in monthly Meta Ads spend.

Sponsor Presented by FunnelFox
Three founders who started with no head start: Jake Castillo (ex-Cal AI), Ernesto Lopez (Covenant Studios) and Steve P Young (App Masters). Where apps lose revenue, when paid ads start paying off, and whether to build one app or ten.
Save your seat → Free. Everyone who registers gets the recording.

Platform

• Apple spells out the 'more than half' rule for screenshots Apple Developer, App Review Help (Guideline 2.3.3 reference page)
Apple's new Guideline 2.3.3 reference page: App Review "will approve only those submissions where more than half of screenshots show the app in use," assessed per device and per localization (two of three pass; two of four don't). The rest may use marketing materials or splash screens; overlays are fine if they don't obscure "the primary app functionality." An in-use image can span multiple screenshots.
• Binance warns of unpatched iOS flaw allowing cross-app data access 9to5Mac
Binance warned iPhone users about an iOS flaw that could let malicious apps access sensitive data stored by other apps.

Monetization

Sub Club by RevenueCat artwork▶ YouTube · 41 min
Sub Club by RevenueCat
Key takeaway: Flo finds that 40% of its subscriptions occur more than six months after install, so paywall conversion is a long-term lifecycle rather than a single event.
• Flo manages 80 million monthly active users by tracking evolving needs over years, frequently converting freemium users at the 3, 6, or 12-month marks.
• Synergy between teams allowed Flo to turn a high-performing Meta ad narrative about "enhancing sex life" into a dedicated onboarding flow that became a major revenue driver.
• An experimental perimenopause onboarding step initially caused a 10% conversion drop; after rewriting the logic to handle "undetermined" phases without error messages, it resulted in a 23% win.
• Mike Charniatsou warns against "cargo culting" features like Duolingo’s streaks, as these mechanics cannot manufacture motivation if a user does not already value the core action.
• Popularity does not guarantee onboarding success; Flo’s "Partner Mode" has 2 million active pairs but failed to improve conversion when inserted into the initial user flow.
• David Barnard cites BoldVoice: removing the paywall’s "X" button spiked conversions, but refunds spiked massively too, a net negative drain on revenue.
Why it matters: For apps with long-term utility, optimizing for immediate paywall wins can cannibalize lifetime value and trigger refund spikes.
Worth it: Yes, because it provides rare data on long-tail conversion and warns against common "best practice" traps like hard paywalls and borrowed gamification.
themobiletakeaway.com
Apptivate (Remerge) artwork▶ YouTube · 33 min
Apptivate (Remerge)
Key takeaway: Match your offerwall rewards to the specific UA source and user intent to avoid overpaying for low-value engagement.
• Users acquired through purchase-oriented campaigns are often 2x to 3x more expensive; pairing them with ad-based games in an offerwall is an inefficient use of high-intent traffic.
• Dynamic reward models should adjust based on the media source, showing ad-watching milestones to ad-revenue users and high-monetization events to purchase-oriented users.
• Mixing different user goals—like purchase intent and ad-revenue—in a single UA campaign confuses platform algorithms and breaks downstream monetization models.
• Sustainability in rewarded traffic requires knowing the exact LTV at specific milestones; rewarding a user $3.00 for reaching a level that only generates $2.00 in value is a common trap.
• Incentivized users behave more aggressively and churn faster than organic users, meaning standard LTV prediction models will produce dangerously optimistic results.
• To prevent fraud in rewarded campaigns, always track events via server-to-server (S2S) integration rather than relying on client-side app signals.
Why it matters: Small app businesses using offerwalls can burn through UA budgets quickly if they do not segment rewards by the cost and behavior of the incoming traffic.
Worth it: Yes, because it provides a concrete framework for optimizing rewarded traffic which is often misunderstood as low-quality filler.
themobiletakeaway.com
• What the Apple Ads team told one indie dev @adriamatz on X
After a call with the Apple Ads team, Adrià Martinez shared their advice for indie apps: if a keyword gets few impressions, bid higher ("Too low and you don't even enter the auction"). Paying more per tap can still lower cost per install. Start with two campaigns, one protecting your brand and one broad for discovery, and use Custom Product Pages per keyword.
• RevenueCat recommends blended metrics to solve attribution discrepancies RevenueCat Blog
RevenueCat suggests founders accept data gaps between Meta and MMPs by using blended metrics to avoid measurement paralysis when scaling.

Business

Starter Story artwork▶ YouTube · 14 min
Starter Story
Key takeaway: Bogdan Dragomir generated $130,000 in revenue in under four months by combining a simple AI ebook-generation wrapper with aggressive Meta Ads spending and niche-specific landing pages.
• The platform, getebook.ai, reached its first $100,000 in four months and currently maintains a monthly recurring revenue of $7,391 with 151 active subscribers.
• Dragomir spends $20,000 to $30,000 per month on Meta Ads, maintaining profit margins between 40% and 50% by constantly testing new creatives to prevent performance dips from ad fatigue.
• The growth strategy relies on building multiple funnels for different angles, such as specific landing pages for Amazon KDP authors featuring royalty and self-publishing cost calculators.
• The founder, who does not code, built the initial version using the AI tool Lovable but eventually rebuilt the application using Claude Code after the first codebase became an unmanageable mess.
• Monthly infrastructure costs include roughly $1,300 for OpenAI and Gemini image generation APIs, $200 for Claude.ai, and $20 for Supabase.
• The pricing model uses a tiered subscription from $29 to $59 per month, supplemented by upsells for one-time credits and tutorials to increase customer lifetime value.
• Dragomir warns that while AI allows for rapid development, it also allows founders to build a mess just as fast, requiring founders to eventually understand the logic they are shipping.
Why it matters: This case study demonstrates that for AI wrappers, marketing expertise and funnel optimization often outweigh technical complexity in achieving rapid profitability.
Worth it: Yes, because it provides a transparent breakdown of the ad spend and funnel tactics required to scale a high-margin AI utility.
themobiletakeaway.com

Every issue