How to Test Google AI Search & Performance Max Assets

This guide walks you through testing Google’s AI‑powered features for Search campaigns and Performance Max asset groups. You’ll learn how to set up experiments, measure results, and interpret the data so you can make informed decisions about your ad spend.

What You’ll Need

• A Google Ads account with an active Search campaign or Performance Max campaign.
• At least 30 days of historical conversion data (or a clear conversion tracking setup).
• Access to Google Ads Editor or the web interface (for asset testing).
• A clear hypothesis: what specifically do you want to test (e.g., new ad copy, AI‑generated assets, or a new bidding strategy)?

Step-by-Step

1. Define your goal and metric. Decide what “success” means: conversions, ROAS, CTR, or CPA. Keep the same primary metric for both the test and control groups.
2. Set up conversion tracking. Ensure Google Ads is tracking the actions that matter (purchases, sign‑ups, leads). Use Google Tag Manager or the global site tag. Verify that conversions are recording correctly in the “Conversions” summary.
3. For Search campaigns – test AI‑generated ad variations. In your Search campaign, create a new ad group or use the “Ad variations” feature under “Ads & extensions.” Choose “Text ad” and let Google’s AI generate multiple headlines and descriptions. Alternatively, use the “Responsive Search Ads” (RSA) format and pin some assets to control placement.
4. Run a campaign experiment (for bidding or ad copy). Go to “Drafts & experiments” in the left menu. Create a draft of your campaign, then start an experiment. Choose a split (e.g., 50/50) and set a duration (at least 3–4 weeks). Make sure the experiment uses the same budget and targeting as the original.
5. For Performance Max – test asset groups. In your PMax campaign, create a new asset group (or duplicate an existing one). Change the assets: new headlines, images, videos, or even a different “final URL” for a specific landing page. Use the “Asset group report” to see which combinations perform best.
6. Use the “Experiments” tab for PMax. Google Ads now allows you to run experiments on Performance Max campaigns. Set up a “Performance Max experiment” by selecting the campaign you want to test, then create a new version with different assets or a different budget allocation.
7. Monitor performance. After the experiment starts, check the “Experiments” report daily for the first few days to ensure data is flowing. Look at the “Confidence” metric – wait until it reaches at least 90% before making decisions.
8. Analyze the results. Once the experiment concludes, Google will show you whether the test version outperformed the control. Look at the “Lift” percentage and the statistical significance. If the lift is positive and significant, apply the winning version to your campaign.
9. For asset testing, use the “Asset performance” report. In your PMax asset group, click on “Assets.” You’ll see ratings (Low, Good, Best). Replace any “Low” assets with new ones. For Search RSAs, use the “Ad strength” indicator – aim for “Excellent” by having 8–10 headlines and 4–5 descriptions.
10. Document your findings. Keep a log of what you tested, the dates, and the outcome. This helps you avoid repeating failed tests and builds a knowledge base for future optimizations.

Common Mistakes

• Ending the experiment too early. Statistical significance takes time. If you stop after a few days, you might make decisions based on noise.
• Changing multiple variables at once. If you test new ad copy and a new bidding strategy simultaneously, you won’t know which change caused the difference.
• Ignoring the “Confidence” metric. Google shows a confidence level – if it’s below 90%, treat the results as inconclusive.
• Not using a control group. Always keep the original campaign or asset group as a baseline. Without it, you can’t measure lift.
• Forgetting to check the “Learning period.” PMax campaigns go through a learning phase (usually 1–2 weeks). Don’t judge performance during that time.

Tips

• Use the “Drafts & experiments” feature for Search. It’s the most reliable way to test AI‑generated ad copy without affecting your live campaign.
• For PMax, test one asset type at a time. For example, change only the headlines, not the images and videos together.
• Leverage Google’s AI suggestions. In the “Recommendations” page, you’ll see suggestions like “Add more headlines” or “Use AI‑generated images.” Apply those that align with your hypothesis.
• Give the experiment enough budget. If your daily budget is too low, the experiment may not reach statistical significance. Aim for at least 50 conversions per variant per month.
• Use the “Asset group report” to spot underperformers. Replace assets with low CTR or conversion rates promptly.

Troubleshooting

• Experiment not showing results. Check that the experiment is “Active” and that the split is not 0/100. Also verify that your conversion tracking is firing – use the “Diagnostics” tool in Google Ads.
• PMax experiment not available. This feature is rolled out gradually. If you don’t see the option, try creating a draft of the campaign and then starting an experiment from the draft.
• Results show “No data.” Wait 24–48 hours. If it persists, check that the campaign is not paused and that the start date is in the past.
• Confidence level stuck at 50%. This means the experiment is underpowered. Increase the experiment duration or budget.
• Assets not showing in the report. Make sure you’ve given the asset group at least 7 days to collect data. Also, check that the asset is actually attached to the asset group (not just uploaded to the account).

Frequently Asked Questions

How long should I run a Google Ads experiment?

At least 3–4 weeks, or until you have 50–100 conversions per variant. This ensures statistical significance.

Can I test Performance Max against a Search campaign?

Yes, but it’s not a true experiment because the campaign types differ. Use the “Experiments” feature to compare PMax to a standard Search campaign, but know that the results may be influenced by other factors like audience targeting.

What is the “Learning period” in Performance Max?

It’s the first 1–2 weeks when the AI is exploring. Performance may fluctuate. Don’t make changes during this time.

Should I use “Ad variations” or “Drafts & experiments”?

Ad variations are for quick tests of ad copy within a single ad group. Drafts & experiments are for full campaign‑level changes like bidding strategy or budget.

How do I know if the AI‑generated assets are better?

Compare the asset group performance to a control group. If the AI‑generated assets have a higher conversion rate or ROAS, they’re better.

What does “Confidence” mean in Google Ads experiments?

It’s the statistical likelihood that the observed difference is not due to chance. Aim for 90% or higher.

Related

• GST Refund on Pharma Exports in India: How-To
• Ip command or nmcli? How to Choose
• BlueStacks Incompatible Device Error: Fixes