Ads With AI (And When To Not Run Them)

Paid ads are the fastest way to scale a working funnel and the fastest way to burn cash on a broken one. The first test: are you converting organic traffic into customers? If not, ads make you lose money faster. Only run ads when the funnel works for free traffic first.

1. 1Landing page converts at least 5% of targeted organic traffic
2. 2You know your CAC target based on LTV (see that lesson)
3. 3You have at least 2 different proven creative hooks from organic
4. 4You have analytics wired up (conversion tracking on the CTA)
5. 5You can afford to lose your full ad budget and learn from it

Miss any of those and don't run ads yet. Fix the funnel first.

Where to run ads, 2026

The 'AI makes variants' superpower

The biggest leverage AI gives you on ads is variant generation. Where a founder in 2020 might test 3 ad creatives, you can now generate and test 30 in a day. AI drafts hooks, writes body copy, generates image prompts for Midjourney/Seedream, and scripts short videos. You then run small-budget tests and let the data pick winners.

The variant generation prompt

"Product: [paste positioning]. Audience: [paste].

Generate 10 ad creative variants, each with:
1. Hook (first 3 seconds or top line) — 10 distinct angles: pain, curiosity, status, fear of missing out, contrarian, social proof, before/after, question, specific result, story opener
2. Body copy (2-3 sentences)
3. CTA button text (2-3 words)
4. Image concept for Midjourney (1 sentence)
5. Platform fit (which platform this works best on)

Brand voice: [paste]. Do not use AI-smell words. Concrete specifics over abstract claims. Assume a $100 daily test budget — make each variant distinct enough that we can learn from the winner."

The small-budget test structure

1. 1Week 1: spend $50-$150 testing 5-10 variants at $10-$20 each
2. 2Kill any variant with CAC 2x your target — immediately
3. 3Scale the top 2 variants at $50-$100/day
4. 4Week 2: make 5 new variants based on what worked
5. 5Keep a spreadsheet: creative | spend | conversions | CAC | status

When to not run ads

• You're under $2k MRR — organic growth has more to teach you
• Your landing page converts under 3% from cold traffic
• Your LTV is unclear or negative at small scale
• You can't track conversions reliably (especially post-iOS 14)
• You're still figuring out positioning — ads amplify broken positioning

What 'good' looks like

A good ad program: spends at 3x LTV/CAC or better, kills weak creatives within 48 hours, runs 5-10 live variants at any time, and has a weekly review where you kill, scale, and rewrite. Ads are never 'set and forget' — they're a weekly gardening habit. If you're not willing to do the weekly work, don't start.