A/B Ad Copy Testing | Which Message Actually Works

Ad copy testing is about determining which message genuinely works by measurement rather than guesswork. A well-built test informs not only that campaign but your landing page copy and overall positioning.

What testing means in the RSA era

Responsive search ads already test headlines internally. So the classic form of writing two ads and racing them has largely changed.

Today testing happens on two levels. First at asset level: reading the asset report to see which headline inside the RSA performs. Second at campaign level: comparing two genuinely different approaches through separate ads or the experiments feature.

The single-variable rule

For a test to work, only one thing can change at a time. If you change the headline, the landing page and the bid together, you will not know why the result came out as it did.

The rule sounds tedious and is often broken. But when broken, the test stops producing results and merely passes time.

What is worth testing

Small wording changes rarely produce measurable differences. What is worth testing is a different claim or a different frame.

Comparing a price emphasis against a speed emphasis is meaningful; testing two synonyms for “fast delivery” is not. Likewise, addressing a different audience, naming a different problem or using a different call to action are real tests.

How long to wait

Ending tests early is the most common mistake. The variation ahead in the first days can fall behind as data accumulates.

How long to wait depends on your traffic: you need enough data to say the result was not chance. In low-traffic accounts that runs into weeks, and sometimes fixing obvious problems produces results faster than testing.

The measure is conversions, not clicks

Copy with a high click-through rate does not always produce more sales. Headlines that provoke curiosity without delivering get clicks and then lose people on the page.

So I evaluate tests on conversions and cost per conversion. Click-through rate is an intermediate indicator; the final measure is always the result.

How do we start? First I read the asset report on your existing ads to see which messages work. I send the real alternatives to test and the timeline in writing. You can reach me from the contact page.

Frequently Asked Questions

How many ads should be in one ad group?

Few. Too many split traffic and none gathers meaningful data. Usually one main RSA plus one alternative for comparison is enough.

What do I do once there is a winner?

Remove the loser and continue with a new alternative. Testing is not a one-off; creative fatigue means it has to be continuous.

Should I use the experiments feature or run it manually?

In accounts with sufficient budget and volume, experiments give cleaner results because traffic is split in a controlled way. In small accounts manual comparison stays practical.