AI Performance Marketing Stack: Meta + Google Ads on $500/mo

The core problem with ad campaigns under $1,000/mo is iteration speed. Large teams test 50–100 creative variants per week; small businesses launch 3–5 and wait a month for results. AI closes this gap. One marketer with the right tool stack now generates, tests, and optimizes creatives faster than a five-person team could two years ago.

This guide covers a real stack, prompts, budget allocation, and an optimization system for Meta Ads and Google Ads at a $500/mo budget.

AI Stack for Performance Marketing: What and Why

The stack has four layers. Each one solves a distinct problem in the ad creation and optimization funnel.

┌──────────────────────────────────────────────────────────┐
│              AI Performance Marketing Stack               │
├──────────────┬──────────────┬─────────────┬──────────────┤
│  Research    │  Creative    │  Copy       │  Analytics   │
│  & Strategy  │  Production  │  Generation │  & Optimize  │
├──────────────┼──────────────┼─────────────┼──────────────┤
│ Claude/GPT   │ Midjourney   │ Claude/GPT  │ Platform     │
│ SpyFu        │ Canva AI     │ + frameworks│ built-in     │
│ Meta Library │ RunwayML     │             │ + Sheets/AI  │
└──────────────┴──────────────┴─────────────┴──────────────┘

Research & Strategy. Claude or GPT-5.4 for competitor analysis, hypothesis generation, and audience segmentation. Meta Ad Library (free) for dissecting competitor creatives. SpyFu or SEMrush (free tier) for Google Ads keywords.

Creative Production. Midjourney or DALL-E 3 for static images. Canva AI (Pro at $13/mo) for format adaptation. RunwayML (free tier) for short videos. On a $500/mo ad budget, spending $200+ on tools is wasteful — the exact breakdown is below.

Copy Generation. Claude or GPT-5.4 with PAS, AIDA, and BAB frameworks for ad copy. Prompts follow in the next section.

Analytics & Optimize. Built-in Meta and Google analytics. Google Sheets with formulas for consolidated dashboards. AI for reading the data and generating optimization hypotheses.

Stack Cost

Tool	Purpose	Cost/mo
Claude Pro / ChatGPT Plus	Research, copy, analysis	$20
Canva Pro	Creative adaptation	$13
Midjourney Basic	Image generation	$10
Meta Ad Library	Competitive analysis	Free
Google Keyword Planner	Keywords	Free
Google Sheets	Dashboards, reports	Free
Total		$43/mo

The remaining $457 goes to ad spend. At $500/mo, every dollar sitting in a tool subscription is a dollar not buying data you can actually learn from.

Allocating the $500/mo Budget: Meta vs. Google Ads

There’s no universal split. Where you put the money depends on your product type, the length of your buying cycle, and where your audience actually hangs out.

Three Allocation Models

Model 1: E-commerce / impulse purchases (70/30 Meta/Google)

• Meta Ads: $320/mo — visual creatives sell the product directly
• Google Ads: $137/mo — brand keywords + retargeting via Performance Max

Model 2: B2B / SaaS (30/70 Meta/Google)

• Meta Ads: $137/mo — awareness and retargeting
• Google Ads: $320/mo — search ads targeting high-intent queries

Model 3: Local business (50/50)

• Meta Ads: $228/mo — geo-targeted ads
• Google Ads: $229/mo — Local Search Ads + Google Maps

The First Two Weeks Rule

The first 14 days are an algorithm learning phase. Meta Ads needs at least 50 conversions per week to exit the Learning Phase. At $10/day with a $5 cost per conversion, that’s mathematically impossible — you’re not going to get there.

The fix: optimize for micro-conversions. Instead of purchases, target add-to-cart. Instead of form submissions, target form clicks. This gives the algorithms enough signal to actually learn something.

Week 1-2:  Optimize for micro-conversions (clicks, views)
Week 3-4:  Switch to macro-conversions (leads, purchases)
Week 5+:   Optimize by ROAS/CPA

AI Prompts for Generating Ad Creatives

Prompt for Competitor Analysis via Meta Ad Library

Context: I'm running ads for [product/service] with a $500/mo budget.
Target audience: [description].

Here are 5 competitor ads from Meta Ad Library:
[paste copy and creative descriptions]

Analyze:
1. What hooks (first 3 seconds / first line) are being used
2. What creative format dominates (static/video/carousel)
3. What CTAs are used
4. What pain points/desires are being leveraged
5. What's not covered — where is the gap I can occupy

Output: a table with the analysis + 3 hypotheses for my first split test.

Prompt for Ad Copy Generation (Meta Ads)

Role: Senior performance copywriter with 10 years of direct response experience.

Product: [name, description, price, key benefit]
Audience: [demographics, pain points, desires, objections]
Format: Meta Ads (Primary Text up to 125 characters for preview,
Headline up to 40 characters, Description up to 30 characters)

Generate 5 ad variants, each using a different framework:
1. PAS (Problem → Agitation → Solution)
2. AIDA (Attention → Interest → Desire → Action)
3. BAB (Before → After → Bridge)
4. Social Proof Lead (start with a result/testimonial)
5. Direct Offer (straight to the point, no preamble)

Requirements:
- First line of Primary Text = a hook that stops the scroll
- Specific numbers instead of abstractions ("3 days" instead of "fast")
- One CTA per ad
- No exclamation marks or clickbait

Prompt for Google Search Ads

Role: Google Ads specialist, certified in Search and Performance Max.

Product: [description]
Target keywords: [list of 5-10 keywords]
Landing page: [URL]
Unique selling proposition: [USP]

Generate a Responsive Search Ad:
- 15 headlines (up to 30 characters each):
  - 5 with keywords
  - 5 with benefits/USP
  - 3 with numbers/facts
  - 2 with CTAs
- 4 descriptions (up to 90 characters each):
  - 2 focused on benefits
  - 1 with social proof
  - 1 with a direct CTA

Pinning: indicate which headlines to pin to positions 1 and 2.
Negative keywords: suggest 10-15 negative keywords for this ad group.

Prompt for Visual Creative Generation (Midjourney)

For static Meta Ads, three formats consistently perform: a lifestyle shot with the product, a before/after, and an infographic with numbers. Midjourney prompt:

Product lifestyle photograph, [product] in use by [persona description],
natural lighting, shot on iPhone, authentic feel, no text overlay,
16:9 aspect ratio for Meta feed placement --ar 16:9 --v 6.1 --style raw

The --style raw parameter strips out the “AI gloss.” Natural-looking photos beat stock illustrations on CTR in almost every test I’ve run.

Campaign Structure: Meta Ads at $320/mo

With a limited budget, a simple structure beats segmentation every time. The more campaigns you run, the thinner the budget spreads and the longer the learning phase drags on.

Recommended Structure

Account
└── Campaign 1: Advantage+ Shopping / Conversions
    ├── Ad Set 1: Broad Targeting (no interest targeting)
    │   ├── Ad 1: Static — PAS framework
    │   ├── Ad 2: Static — Social Proof
    │   └── Ad 3: Video — Problem/Solution
    └── Ad Set 2: Lookalike 1% (if data is available)
        ├── Ad 1: Static — Direct Offer
        └── Ad 2: Carousel — Features

Broad targeting. At $10/day, narrow interest targeting chokes your auction volume. Meta Advantage+ finds converting audiences better than manual targeting at this budget level. Broad is cheaper and learns faster.

3–5 ads per ad set. Fewer and you don’t have enough data to pick a winner. More and you’ve split the budget so thin that nothing reaches a meaningful sample size.

Format: 60% static, 40% video. Static creatives are cheaper to produce and faster to test. Video is for scaling what you’ve already proven works.

Campaign Structure: Google Ads at $137/mo

Recommended Structure

Account
├── Campaign 1: Search — Brand Keywords ($40/mo)
│   └── Ad Group: Brand terms + variations
│       └── RSA: 15 headlines / 4 descriptions
├── Campaign 2: Search — High Intent ($67/mo)
│   ├── Ad Group 1: [Product] + buy/price/order
│   └── Ad Group 2: [Problem] + solution/how to
└── Campaign 3: Performance Max ($30/mo)
    └── Asset Group: all formats

Brand keywords are non-negotiable. Even at $137/mo. Competitors bid on your brand name, and brand CPC sits at $0.10–$0.50 typically. That’s $1–2/day to protect ground you’ve already earned.

High-intent keywords only. For a deeper dive into AI-assisted keyword research, see the dedicated guide. When the budget is tight, skip anything that doesn’t signal purchase intent. “What is a CRM” is expensive and won’t convert. “CRM for small business pricing” — that person is shopping.

Performance Max. Run it at $1/day. PMax touches all Google placements (Search, Display, YouTube, Gmail, Maps). At this scale it’s an extra data collection channel, not a primary driver.

Metrics: What to Track and When to Act

Three levels of metrics, three different cadences: daily monitoring, weekly analysis, monthly review.

Daily Metrics (5 minutes each morning)

Metric	Meta Threshold	Google Threshold	Action on Deviation
Spend	>20% deviation from daily plan	>20% deviation	Check limits
CPM	>$15 (B2C), >$30 (B2B)	N/A	Refresh creatives
CPC	>$2 (B2C), >$5 (B2B)	>$3 (B2C), >$8 (B2B)	Check relevance
CTR	<1%	<3% (Search)	Test new hooks

Weekly Metrics

Metric	Target	Formula
CPA (Cost Per Acquisition)	Depends on margin	Total Spend / Conversions
ROAS	>2.0 for e-com	Revenue / Ad Spend
Frequency	<3.0 over 7 days (Meta)	Impressions / Reach
Quality Score	>6 (Google)	Built-in metric

Monthly Metrics

Metric	What It Shows
Blended CAC	Total acquisition cost across all channels
LTV:CAC ratio	Should be >3:1 for sustainability
Attribution delta	Gap between platform data and actual sales

AI Prompt for Metrics Analysis

Here's my ad campaign data for the past week:

Meta Ads:
- Spend: $80, Impressions: 12,000, Clicks: 180, Conversions: 4
- Top ad: [name] — CTR 2.1%, CPA $12
- Worst: [name] — CTR 0.4%, CPA $35

Google Ads:
- Spend: $35, Impressions: 2,800, Clicks: 95, Conversions: 3
- Avg CPC: $0.37, Quality Score avg: 7

Budget: $500/mo. Goal: CPA < $15. Current margin: $40 per sale.

Analyze the data and provide:
1. What's working and why (specific hypotheses)
2. What's not working and what three actions to take
3. Budget reallocation proposal for next week
4. Three new A/B test hypotheses

Campaign Optimization: Weekly Cycle

At $500/mo, you’re running short iterations. Every week: one meaningful test, one call to scale or cut.

Weekly Optimization Cycle

Monday: analysis. Pull last week’s data. Feed it into Claude/GPT with the analysis prompt above. Find the winner and the dead weight.

Tuesday: creative refresh. Pause anything with CTR below 0.8% (Meta) or 2% (Google Search). Generate 2–3 replacement variants using the prompts above, built around elements from the winners.

Wednesday: launch. Upload the new creatives. Check that budgets are distributed correctly. Make sure no campaigns are stuck in a Learning Phase.

Thursday–Friday: monitoring. Daily threshold checks. Touch nothing — the algorithms need time to stabilize.

Saturday–Sunday: passive data collection. Seriously, don’t touch anything. Weekend traffic behaves differently from weekday traffic, and you want that extra data for Monday’s analysis.

Scaling Rules

At $500/mo, scaling means reallocation — not budget increases.

The 70/30 rule. 70% of the budget goes to ads that are already performing (CPA below target). 30% funds new hypotheses. Revisit the split each week.

The kill rule. An ad has burned through 2x the target CPA with zero conversions — pause it. Don’t wait for statistical significance that won’t come at $10/day.

The scale-up rule. An ad holds a CPA under 50% of target for 7 days — push its budget share up by 20%. Don’t go higher than that in one move; Meta resets the Learning Phase if you do.

Automating Routine Tasks with AI

Weekly Report in 10 Minutes

Skip the manual data pull from two platforms. Export CSVs and upload them to Claude.

1. Meta Ads Manager → Export → CSV (last 7 days)
2. Google Ads → Reports → Download CSV
3. Upload both files to Claude with this prompt:

Analyze two CSV files: Meta Ads and Google Ads for the past week.

Output:
1. Summary table: channel, spend, conversions, CPA, ROAS
2. Top 3 ads by CPA (both channels)
3. Ads to pause (CPA > $X or CTR < Y%)
4. Budget recommendation for next week
5. Three test hypotheses

Format: markdown tables, no filler.

Automating Creative Generation

A workflow for creating 10 ad variants in 30 minutes:

1. Analyze winners (5 min). Break down what’s working in your best ads: hook, format, CTA, visual style.

2. Generate copy (10 min). Give Claude the winner data and ask for 5 variations across different frameworks.

3. Generate visuals (10 min). Midjourney: 4 image variants. Canva AI: resize to formats (1080×1080 for feed, 1080×1920 for stories, 1200×628 for Google Display).

4. Assemble and upload (5 min). Pair copy with visuals. Upload to Meta Ads Manager and Google Ads.

That’s 4–6 hours of designer and copywriter work, done. For deeper coverage of ad copy frameworks (PAS, AIDA, BAB, FAB, 4U), see AI Copywriting for Ads. For a broader approach to AI-assisted content generation, see the AI Content Repurposing guide.

Common Mistakes on Budgets Under $500/mo

Too many campaigns. Five campaigns at $3/day means none of them will exit the Learning Phase. Cap it at 2–3.

Chasing the final conversion on day one. At $10/day, the algorithm isn’t going to see 50 purchases per week. Start with intermediate conversions, then switch to final ones once you have enough data.

Touching creatives every day. Meta needs 3–5 days to evaluate an ad. Google Search needs 7–14 days for Quality Score to settle. Every change you make resets the clock.

Ignoring negative keywords. In Google Ads at $4/day, one irrelevant click is 25% of your daily budget gone. Check the Search Terms Report weekly and keep adding negatives.

Copying big competitors. A brand spending $50,000/mo can afford brand awareness. You can’t. At $500/mo every dollar has one job: direct response. No awareness, no brand plays.

Launch Checklist: First 30 Days

Week 0 (prep):
□ Install Meta pixel and Google Ads tag
□ Set up conversions (macro + micro)
□ Analyze 10 competitors in Meta Ad Library
□ Collect 20-30 keywords in Keyword Planner
□ Choose a budget allocation model
□ Generate the first batch of creatives (5 static + 2 video)

Week 1-2 (launch):
□ Launch 2 campaigns (Meta + Google)
□ Micro-conversions for algorithm learning
□ Daily monitoring of threshold metrics
□ DO NOT change creatives

Week 3 (first optimization):
□ Analyze 2 weeks of data (prompt above)
□ Pause underperforming ads
□ Launch 3 new variants
□ Switch to macro-conversions (if enough data)

Week 4 (scaling):
□ Reallocate budget using the 70/30 rule
□ Identify best channel — shift 10-15% of budget there
□ Prepare creatives for next month
□ Record baseline metrics for comparison

$500/mo is a workable performance marketing budget — if the structure is right. AI cuts creative production costs from $500–2,000 down to $43/mo and compresses a two-week testing cycle to two days. The bottleneck isn’t tools. It’s the discipline to run the weekly optimization cycle without skipping. Every missed analysis cycle is $100+ spent on ads you already know aren’t working.

---

Need help setting up a performance marketing stack? I help startups build marketing systems at belov.marketing and develop AI solutions at belov.works.

FAQ

Should you run Meta and Google Ads simultaneously from day one, or start with one platform?

Start with both simultaneously. The reason is data velocity: at $500/mo split 50/50, each platform gets enough signal by week 3–4 to identify a clear winner. Starting with one platform for 30 days then switching delays this comparison by a month and wastes budget on a single channel that may not be optimal for your product type. The exception: if your product is purely impulse-purchase (e.g., physical goods under $50), Meta alone for the first 30 days is defensible — the intent is in the creative, not in the search query.

How do you know when a Meta ad’s performance drop is creative fatigue versus audience saturation?

Creative fatigue shows up as declining CTR while CPM stays flat. The ad is still being shown to the same audience, but fewer people are clicking — the creative has lost novelty. Audience saturation shows up as rising CPM while CTR holds — you’ve exhausted the cheap inventory and are now bidding into a smaller, more expensive pool. Frequency above 3.0 over 7 days usually signals saturation. Creative fatigue typically happens within 2–3 weeks for a small audience; the fix is new creatives. Saturation requires expanding the audience or pausing and letting the frequency reset.

Can AI-generated Midjourney images outperform real product photography in Meta Ads?

For most B2C products, no. Real photographs taken on an iPhone in natural light outperform AI-generated imagery on CTR because they read as authentic rather than produced. The --style raw parameter in Midjourney reduces the AI gloss, which helps, but the uncanny valley effect is still detectable to users. AI-generated images work fine for abstract services, SaaS interfaces, and B2B use cases where lifestyle photography isn’t the primary creative driver. Test both in the same ad set and let the data decide — don’t assume either way.