Ad Copy, Split Testing & AI for Google Ads

Summary

Aaron Young argues that systematic ad-copy research and split testing can improve Google Search campaign performance more reliably than simply supplying Google with 15 responsive-search-ad headlines and leaving selection entirely to its system.

Ad-Copy Framework

Each ad should combine four elements:

• Keyword focus to match the search.
• Brand mention to build recognition.
• Strong call to action to make the next step clear.
• Emotional trigger tied to the customer’s pain point or desired outcome.

Avoid generic claims such as “we’re the best.” Where possible, use specific, supportable facts—such as percentages, dollar savings, or guarantees—and highlight claims competitors cannot or will not make.

AI-Assisted Research and Writing Process

Young demonstrates using Claude to turn business and campaign information into headline and description ideas. The discovery process gathers:

• The offer or product and the problem it solves.
• The target audience, unique selling points, and primary keywords.
• Relevant claims, statistics, savings, or guarantees.
• Forbidden keywords and the landing-page URL.
• Existing ad copy and an exported ad report.

The ad report can be downloaded as a CSV from Google Ads and uploaded for analysis. Claude can also review the landing page and identify potential messaging angles.

The resulting copy is a starting point, not a proven winner. Select promising themes, refine them, and test them. Customer avatars or other audience research can further improve specificity. Gemini is mentioned as an alternative tool for generating additional ideas.

Split-Testing Playbook

• Run two ads per ad group; Young says third and fourth ads often receive little or no delivery.
• Change one core variable at a time so results can be attributed to that change.
• Test in layers. For example, after identifying dynamic keyword insertion pinned in position one as a stronger approach, the account was testing different emotional triggers next.
• Keep a record—such as in Google Sheets—of what has and has not been tested.
• When a test produces a winner, retain the learning and test another variable.
• To create the next test, copy the winning ad, paste it as a duplicate, pause the underperforming ad, and make one change to the new ad.
• Allow tests enough time and data before drawing conclusions; small samples can be misleading.

Metrics and Examples

• Client case study: The client’s CTR was below 1% before working with the team in late August. After ad-copy testing, it rose to above 10% by October. The presenter says conversion rate increased from about 1% to above 8%. He notes that CTR later declined, attributing this partly to the introduction of AI Mode and AI Overviews; conversion rate remains the more important outcome.
• Another test example: One ad had a 6% CTR and 9% conversion rate, compared with 8.6% CTR and 5% conversion rate for the other ad. The first ad had only 83 clicks, so the result was considered promising but not conclusive.
• Across the testing examples, the presenter describes improving conversion rates from roughly 2% to 5–9%, and CTR from well below 1% to around 7–8%.
• Google may show some responsive-ad headlines far less—or not at all—so having many assets does not ensure they will be tested evenly.
• The team says it can produce a collection of testable headlines in roughly 15–20 minutes, potentially supporting testing over the next three to six months.

Key Takeaway

Treat ad copy as an ongoing optimization process: research customer needs, write specific and differentiated messaging, test one change at a time, track results, and use conversion data—not headline volume alone—to guide decisions.

Presenter/source: Aaron Young, Define Digital Academy. Tools discussed: Claude and Gemini.