---
title: Using AI to scale app ad creatives (without fake testimonials)
slug: ai-tools-for-app-ad-creatives
summary: Where AI helps app ad production in 2026 — permutations of real-human ads (ElevenLabs voice/translation, Suno music), AI statics, LLM-brainstormed extreme hooks, Veo 3 / Higgsfield video — and why fully synthetic "testimonials" carry FTC and backlash risk.
domain: mobile-apps
tags: [user-acquisition, ad-creative, ai, ugc, compliance]
related: [ad-creative-volume-and-hit-rate, ugly-ads-and-brand-caution, ugc-app-ad-format-captions-green-screen, product-magic-moment-as-ad-creative, ai-generated-ad-creative]
sources: [yt:UYIgu02h8cs, yt:7JtttjbTRAQ, yt:jzlPf100vy4, yt:EyoYQU6JyV8]
updated: 2026-10-24
---

# Using AI to scale app ad creatives

## TL;DR
- Main win is **volume of permutations**: Runa went from tens to **400+ concepts/month** using AI tools.[^1]
- Safest pattern (Phil Carter): keep a **real human** in the ad and use AI to multiply it — ElevenLabs to translate one spokesperson into ~30 languages, Suno to test background tracks.[^2]
- Be careful with **wholesale AI video ads** (e.g. Veo 3 end-to-end): quality still uneven, consumers react badly when they detect AI, and fake testimonials are regulated.[^2]
- In the US, the FTC rule (2025) bans **fake testimonials** — an AI person claiming "I used Runa to crush my 5K" is a violation risk (raised by David Barnard).[^3]
- AI is strongest for **hooks**: Alex would have an LLM brainstorm openings at the extremes of emotion, then render them with Veo 3 and edit in CapCut — hooks you couldn't film in a creator's apartment.[^4][^5]
- Also: AI **statics** (Built With Science runs many), and LLMs as a brainstorming partner over user reviews and competitor ads.[^6][^7]

## Details
**Production stack mentioned (2026):** ElevenLabs (voice, dubbing), Suno (music), Veo 3 (Google video generation), CapCut (editing), Higgsfield (AI UGC — Yuliya Lennox, an ex-employee, says it produced a convincing UGC avatar from just an app logo, no prompt).[^2][^5][^8]

**Workflow Alex would use now:** scan the Meta Ads Library for inspiration → have a top LLM write scripts/concepts aimed at awe, weirdness or fear → generate the hook visuals with Veo 3 → assemble with real demo footage in CapCut.[^5]

**Brainstorming.** David Barnard suggests mining your reviews, user feedback and competitor ads with an LLM to iterate faster — not only generating assets.[^7]

**Not a replacement for judgement.** Yuliya warns that AI assistants tend to recommend pretty, on-brand 15-second ads — the opposite of what wins (see [[ugly-ads-and-brand-caution]]).[^9]

## Disagreements / caveats
- Phil (cautious on wholesale AI) and Alex/Yuliya (bullish on AI hooks and AI UGC) differ on how far to go; the regulatory line is *impersonating real users/testimonials*, not AI visuals per se.
- FTC rule details and enforcement change — check current guidance (and EU/UK consumer law) before running AI people.
- Tools named are fast-moving and some were promoted by guests with ties (Yuliya/Higgsfield).

## See also
- [[ad-creative-volume-and-hit-rate]] — why volume matters
- [[product-magic-moment-as-ad-creative]] — lead with AI features themselves
- [[people/phil-carter]], [[people/alex-screens-design]]

[^1]: yt:UYIgu02h8cs @ 39:36 — Phil Carter
[^2]: yt:UYIgu02h8cs @ 42:12 — Phil Carter
[^3]: yt:UYIgu02h8cs @ 40:24 — David Barnard
[^4]: yt:7JtttjbTRAQ @ 24:19 — Alex (Boost App Social)
[^5]: yt:7JtttjbTRAQ @ 29:47 — Alex (Boost App Social)
[^6]: yt:EyoYQU6JyV8 @ 13:45 — Ethan Ethier
[^7]: yt:UYIgu02h8cs @ 42:57 — David Barnard
[^8]: yt:jzlPf100vy4 @ 43:20 — Yuliya Lennox
[^9]: yt:jzlPf100vy4 @ 06:39 — Yuliya Lennox