Run an automated AI UGC ad testing loop
Ads that already ran 30+ days are proven angles.
Figures as shared by the creator · source
From @ceo_vlad on X · Oct 2026
The workflow
ScrapeApifyAutomaten8nScriptsClaude
Nebula · adsInfiniteUGCTest $20/dayMeta Ads ManagerThe loop
ResearchCompetitor ads running 30+ daysProduceAI creators, voices, captionsAlso see his article post (17,029 views, 174 bookmarks).
See the end result
Where & with what
What they did
- 1The moveA daily loop: scrape winning competitor ads, script, generate UGC ads, test at ~$20/day, feed winners back.
- 2Step 2Apify + n8n: pull competitor Meta ads running 30+ days
- 3Step 3Claude: 9 scripts from the product doc; a human approves
- 4ResultHe says it runs for a $6M/month brand at about 25% margin, taking 1–2 hours a day.
Copy the prompt
You are my UGC scriptwriter. Product doc: [paste]. Proven competitor angles (ads running 30+ days): [paste research tab]. Write 9 UGC ad scripts, 25-40 seconds each: 3 problem-first, 3 testimonial-style, 3 "I tried X so you don't have to". For each: hook (first 2 seconds), body, CTA, and the angle it borrows. Plain spoken language, no claims we can't prove. Results sorter: Here are yesterday's ad results [paste CSV: ad, spend, CTR, CPA, ROAS]. Sort into WINNERS (CPA under [target] with at least [N] conversions), KILL (spend over [X] with no conversions), TOO EARLY (everything else). One line of reasoning each, then 3 new script angles based on the winners. (Our templates. System by @ceo_vlad.)
Quick answers
What did they do?
The poster describes a daily loop for a product brand: Apify and n8n pull competitor Meta ads that have run 30+ days into a research tab; Claude writes 9 UGC scripts from the product doc and a human approves them; InfiniteUGC (its Nebula agent) turns the scripts into ads with characters, voices and captions; a ~$20/day test campaign goes live in Meta Ads Manager; Claude sorts results into winners, kill and too early, with a Slack alert; winners feed the next day's scripts. A Google Sheet is the control room.
What were the results?
He says it runs for a $6M/month brand at about 25% margin, taking 1–2 hours a day. His article post shows 17,029 views and 174 bookmarks (X, Oct 2026).
What tools do you need?
Apify (web scraping platform with ready-made scrapers (incl. the meta ad library)); n8n (workflow automation that connects apps and ai models); Claude (ai assistant for writing, products and building small tools); InfiniteUGC (ai ugc ad generator: ai creators, voiceover, captions, long ads (its agent is called nebula)); Meta Ads Manager (create and run ads on facebook and instagram); Slack (team chat with alerts and app integrations); Google Sheets (free online spreadsheets).
How much does it cost?
You can start for $0: Apify, n8n, Claude, Meta Ads Manager, Slack, Google Sheets have a free plan. InfiniteUGC is paid or trial-based. Check each tool's site for current prices.
Is it allowed?
Yes. Meta may add an "AI info" label to AI-generated media. Source.
Full write-up & sources
What they did: The poster describes a daily loop for a product brand: Apify and n8n pull competitor Meta ads that have run 30+ days into a research tab; Claude writes 9 UGC scripts from the product doc and a human approves them; InfiniteUGC (its Nebula agent) turns the scripts into ads with characters, voices and captions; a ~$20/day test campaign goes live in Meta Ads Manager; Claude sorts results into winners, kill and too early, with a Slack alert; winners feed the next day's scripts. A Google Sheet is the control room.
Results: He says it runs for a $6M/month brand at about 25% margin, taking 1–2 hours a day. His article post shows 17,029 views and 174 bookmarks (X, Oct 2026).
Worth knowing: Ads that already ran 30+ days are proven angles. Testing many cheap AI variants and feeding winners back in turns creative into a system, not guesswork.
Source: @ceo_vlad on X · Oct 2026. Summarised in our own words; figures as shared by the creator. Updated .
