Google Ads has fundamentally changed: What used to be manual keyword bids and hours of optimization is increasingly being taken over by AI. Smart Bidding, Performance Max, automatic assets. The question is no longer whether to use AI, but how to control it correctly.
This guide shows what AI can do in Google Ads, where its limits are, and how to get the most out of automated strategies.
What Does AI in Google Ads Actually Mean?
Google's AI algorithms analyze hundreds of signals in real time for every auction:
- Device type and operating system
- Time of day and day of week
- Search query (semantic meaning, not just keywords)
- User location
- Demographic characteristics
- Past behavior on your website
- Remarketing status (has the user been on your site before?)
- Competitive landscape in the auction
- Seasonal signals
A human advertiser cannot manage this complexity manually. This is exactly where AI's advantage lies.
Smart Bidding Strategies Overview
1. Maximize Conversions
What it does: Spends as much budget as possible to maximize the number of conversions.
When to use:
- No cost targets defined
- Budget should be fully utilized
- As an entry into Smart Bidding (before enough data for Target CPA)
2. Target CPA
What it does: Keeps average cost per conversion at your defined target.
When to use:
- When you know what a lead/purchase can cost
- Google requires no minimum volume, you can start Target CPA even without conversion history. In my experience, though, it only steers reliably from around 30 conversions in 30 days, and Google itself recommends evaluating performance over 30 days with at least 30 conversions.
Best practice: Don't set Target CPA too aggressively. Start with actual historical CPA × 1.1.
3. Target ROAS
What it does: Optimizes bids for a specific return on ad spend.
When to use:
- E-commerce with conversion values
- When you know your break-even ROAS (worked example: at a 25% margin it is 1 / 0.25 = 400%)
- For Search and Shopping campaigns, Google requires at least 15 conversions in the last 30 days. From my practice, Target ROAS runs noticeably smoother with more conversions with values (more like 30 to 50 per month).
4. Maximize Conversion Value
What it does: Maximizes total conversion value within the budget.
Ideal for: Shops with different product prices. AI prioritizes high-value conversions.
Prerequisites for Successful Smart Bidding
AI is only as good as the data it works with. Smart Bidding strategies often fail because the foundations are missing.
Prerequisite 1: Correct Conversion Tracking
The AI optimizes for what you define as a conversion. Errors here cost a lot of money.
What must be set up correctly:
- Conversion tracking via Google Tag (not URL-based)
- De-duplication activated
- Correct conversion value (for e-commerce: dynamic value, not static)
- No micro-conversions as primary goal (e.g., no "50% scroll" as main conversion)
Prerequisite 2: Sufficient Conversion Volume
It helps to separate what Google technically requires from what works well in practice:
| Strategy | According to Google | Rule of thumb from my practice |
|---|---|---|
| Maximize Conversions | no minimum | can start right away |
| Target CPA | no minimum (evaluation over 30 days with at least 30 conversions recommended) | from around 30 conversions in 30 days |
| Target ROAS | Search/Shopping: at least 15 conversions in 30 days | more like 30 to 50 conversions with values in 30 days |
| Maximize Conv. Value | no minimum for Search/Shopping (Demand Gen: 50 conversions with value in 35 days) | requires clean, dynamic conversion values |
Table scrolls sideways
Prerequisite 3: Realistic Goals
A common mistake: Target CPA is set too low. The AI can't set bids high enough and the ad barely appears.
Rule of thumb: Start with historical CPA + 15–20%, then slowly reduce.
Performance Max: Full AI Automation
Performance Max (PMax) uses all Google channels (Search, Display, YouTube, Gmail, Discover, Shopping, Maps) in one campaign.
What PMax Does Well
✓ Reach across all Google channels ✓ Finds conversions you wouldn't have found manually ✓ Ideal for e-commerce with Google Shopping integration ✓ Works well with high conversion volume
What PMax Doesn't Do Well
✗ Transparency: Little insight into where budget flows ✗ Control: Barely possible to weight specific channels ✗ Brand protection: Tends to use budget for brand searches ✗ Small budgets: Often inefficient below €1,000/month
Best Practices for Performance Max
Asset quality is crucial:
- Fill the asset slots well: Google recommends at least 11 headlines (up to 15), 2 long headlines (up to 5), 4 descriptions (up to 5), images in landscape, square and portrait format, and at least one logo (up to 5)
- Upload video (otherwise Google often auto-generates poor quality videos)
- Set audience signals (your CRM list, website visitors)
Protect brand campaigns: Create a parallel standard search campaign for your brand name with a higher priority. PMax shouldn't "steal" your brand searches.
Don't neglect the search terms report: PMax now has its own search terms report (under "Insights and reports"). Check it regularly and exclude irrelevant terms with campaign-level negative keywords or negative keyword lists. For PMax, these apply to Search and Shopping inventory.
Automatically Created Assets and AI Max
Google automatically writes ad text based on your landing page and existing assets. Within AI Max, this feature is called text customization. Since 1 September 2026, Google has been gradually upgrading Search campaigns that use automatically created assets and/or campaign-level broad match to AI Max. The upgrade applies to individual campaigns, not the whole account; for campaigns with automatically created assets, text customization is switched on afterwards while final URL expansion stays off (Google Ads API team, 12 August 2026). I cover what the upgrade changes in your account and what to check in AI Max: What Changes in Your Account.
Recommendation: Review automatically generated text regularly. Google sometimes produces headlines that don't fit or aren't accurate.
Recommended Smart Bidding Rollout
Phase 1: Collect data (4–6 weeks)
→ Manual CPC or Maximize Clicks
→ Check and optimize conversion tracking
→ Build up conversions (benchmark from my practice: around 30)
Phase 2: First Smart Bidding activation
→ Switch to Maximize Conversions
→ Observe learning phase (7–14 days)
→ No interventions during learning phase
Phase 3: Goal-based bidding (after 4–6 weeks)
→ Switch to Target CPA with realistic goal
→ CPA target: historical CPA + 15%
→ Monitor weekly, optimize monthly
Phase 4: Advanced (after 3+ months)
→ Gradually lower Target CPA (max 15% every 2 weeks)
→ Implement seasonal adjustments
→ Evaluate portfolio strategies for campaign groups
Where Manual Control Is Still Needed
AI is powerful, but not omnipotent. These areas still require human attention:
- Brand strategy and messaging: AI optimizes for conversions, not brand consistency
- Negative keywords: Smart Bidding doesn't prevent irrelevant search terms
- Seasonal adjustments: for predictable events (Black Friday, trade shows), set manual seasonal adjustments
- New markets and products: no historical data yet, start with Manual CPC
- Very high budgets / competitive markets: excessive automation can lead to budget waste
Checklist: Optimizing AI & Smart Bidding
- Conversion tracking fully and correctly set up
- Enough conversion data for the chosen strategy (rule of thumb: around 30 in 30 days)
- Realistic Target CPA based on historical data
- Learning phase indicators understood (don't intervene!)
- Performance Max asset groups with complete assets
- Brand campaign protects own brand keywords
- Negative keywords regularly updated
- Auto-created assets reviewed monthly
Conclusion
AI in Google Ads is not a threat to advertisers. It's a powerful tool that, when used correctly, brings enormous efficiency gains. The art lies in giving AI the right framework: correct tracking, sufficient data, realistic goals.
Those who control AI rather than blindly trusting it will achieve the best results.
Want to optimize your Smart Bidding strategy? Talk to me now.
