AI Max campaigns require Smart Bidding — you can't run AI Max with manual CPC or enhanced CPC. But not all Smart Bidding strategies are equal, and choosing the wrong one for your campaign's current state is a common cause of poor AI Max performance.
The Three Core AI Max Bidding Strategies
Maximize Conversions: The AI spends your entire daily budget to get as many conversions as possible, without a specific CPA or ROAS target. No performance floor — the AI will spend what it takes.
Target CPA (tCPA): The AI aims for a specific cost per conversion. It will forgo volume if necessary to hit the target CPA. Good when you have a fixed acquisition cost requirement.
Target ROAS (tROAS): The AI aims for a specific return on ad spend (revenue/cost ratio). Good for e-commerce where conversion values vary by product and you want to maximize revenue efficiency.
When to Use Each Strategy
Use Maximize Conversions when:
- Starting a new AI Max campaign (no conversion history)
- You want maximum volume and have an unconstrained budget
- Your CPA target is flexible and you just want to see what the AI can achieve
- You have fewer than 30 conversions in the last 30 days (the AI needs volume to calibrate)
Use Target CPA when:
- You have a fixed cost-per-lead or cost-per-acquisition requirement
- You have at least 30 conversions in the last 30 days (better with 50+)
- Your conversion values are constant (every lead = same value)
- You're running lead generation where all leads have similar quality
Use Target ROAS when:
- You have dynamic conversion values (e-commerce with varying order values)
- You want to maximize revenue, not just conversion count
- You have at least 50 conversions in the last 30 days (better with 100+)
- You're running AI Max for Shopping (https://yositeup.com/blog/google-ai-max-shopping-replacing-performance-max-2026)
The Learning Phase and Bidding Strategy Selection
AI Max campaigns have a learning phase after launch or after significant changes. The learning phase is longer and more sensitive with tCPA and tROAS than with Maximize Conversions.
During the learning phase:
- Performance is unpredictable — costs may be higher, CPA/ROAS may be off target
- Avoid making changes (budget, targets, assets) during the first 2 weeks
- The AI is calibrating its model — changes restart the learning phase
Recommended launch sequence:
- Launch with Maximize Conversions
- Let it run for 2-4 weeks (collect 30-50 conversions)
- Switch to tCPA or tROAS once the AI has enough data
- Expect another shorter learning phase after the switch
Setting Your tCPA or tROAS Target
Common mistake: setting targets too aggressively too soon. The AI will under-bid to hit an unrealistic target, causing impression loss and low conversion volume.
For tCPA: Calculate your current actual CPA (if you have data), then set your target 10-20% higher than current CPA. Gradually lower the target over 2-4 weeks as the AI improves.
For tROAS: Calculate your current actual ROAS, set your target 10-20% lower than current ROAS initially. Gradually raise the target over time.
Never set targets below your actual business viability threshold at launch — give the AI room to operate while still being directionally correct about your economics.
The July 2026 ToS and Bidding Transparency
The July 2026 ToS (https://yositeup.com/blog/google-ads-tos-july-2026-ai-automation-what-changed) includes a transparency provision for AI Max bidding: Google must provide auction-level explanations for significant bidding decisions on request. This is a compliance provision, not something you interact with directly — but it means Google's AI Max bidding model is auditable.
If you suspect the AI is making bidding errors (consistently bidding far from your target), you can request auction-level data through the Google Ads API for diagnostic purposes.
The June 2026 Deletion and Bidding Model Reset
After the June 2026 data deletion (https://yositeup.com/blog/google-ads-reporting-data-deleted-june-2026), AI Max campaigns had to rebuild their bidding models from the remaining data. This had a different impact depending on your bidding strategy:
Maximize Conversions: Least affected — the model just needs to learn "spend budget efficiently," which is a lower-information task
Target CPA: Moderately affected — without Q2 2026 conversion data, the CPA model had less information to calibrate against
Target ROAS: Most affected — ROAS optimization requires knowing the full distribution of order values. Losing Q2 data (which may have had seasonal value patterns) degraded model quality more
If you saw tROAS performance degrade in late June / early July 2026, the data deletion is likely a contributing factor. The model rebuilds over time — performance typically recovered within 4-6 weeks.
Switching Between Strategies Mid-Campaign
You can switch bidding strategies on an active AI Max campaign without losing all data. However:
- Each strategy switch triggers a learning phase restart
- Make switches deliberately, not reactively to short-term performance noise
- Avoid switching during peak periods (the AI needs stability when it most matters)
From Maximize Conversions to tCPA: The AI already has conversion data; the switch just adds a constraint. Learning phase is shorter (~1-2 weeks).
From tCPA to tROAS: Requires that your conversion tracking includes conversion values. If you switch to tROAS but all conversions have value = 0, the AI has no signal.
Portfolio Bidding for AI Max
If you run multiple AI Max campaigns, you can group them in a portfolio bid strategy. All campaigns in the portfolio share target achievement — if one campaign is above target, another can be below target and still hit the portfolio goal.
Portfolio bidding for AI Max works for campaigns targeting the same business goal with similar conversion types. Don't group AI Max for Search and AI Max for Shopping in the same portfolio unless they're truly substitutable in terms of conversion value and goal.
DSA to AI Max Bidding Continuity
When migrating DSA to AI Max for Search (https://yositeup.com/blog/google-ads-dsa-ai-max-migration-february-2027): if your DSA ran on tCPA or manual CPC, you'll need to set up a new bidding strategy for AI Max. DSA historical data can inform your initial tCPA target (use DSA's average CPA from the last 90 days), but expect a learning phase after migration.
Top comments (0)