Ad variations are a Google Ads experiment that tests one change to your ad text across many search ads at once, such as swapping a headline or rewording a phrase, by splitting traffic between the original ads and the changed versions. Instead of editing hundreds of ads by hand and guessing whether the edit helped, you get a side-by-side comparison of the two.
How ad variations work
You build a variation from the Experiments area of your account. First you choose the scope: the whole account, selected campaigns, or a filtered set of ads, for example every ad whose headline contains “Free Quote”. Then you describe the change. The common options are find and replace (swap “Free Quote” for “Fixed Price Quote”), updating a specific headline or description, or changing the final URL. At the time of writing (October 2026) ad variations work on Google search campaigns and apply to responsive search ads.
You set a start date, an end date and the share of traffic the variation receives, usually half. Google then serves the original ads to one group of eligible auctions and the changed ads to the other. Because both versions run in the same weeks, against the same competitors and the same seasonal swings, the comparison is far fairer than looking at last month against this month.
The results table compares clicks, click-through rate, conversions, cost per conversion and other metrics for the two arms. Google marks a difference when it believes the result is statistically meaningful rather than noise. When the test ends you can apply the variation, either by updating the original ads or by creating new ads and pausing the originals, or you can discard it and keep things as they were.
Why it matters
Most accounts I look at have dozens or hundreds of ads that were written once and left. Small wording decisions add up across all of them: whether you mention price, whether you say “London” or the borough, whether you lead with speed or with trust. Ad variations let you settle those questions with your own customers’ behaviour rather than opinion, and roll the winner out everywhere in a few clicks.
For a UK business the useful tests are often about the offer and the reassurance. A retailer might test “Free UK Delivery” against “Next-Day Delivery”. A trades business might test “Gas Safe Registered” in a headline against a price-led message. A professional services firm might test “Free Initial Consultation” against naming the specific problem it solves. Each of these is a single, clear change that the tool is built to measure.
Common mistakes
- Testing too many things at once. If you change the headline, the description and the URL together, a win tells you nothing about which change caused it.
- Running a test on low-traffic ads. A few hundred impressions a week will not produce a trustworthy answer; widen the scope or pick a busier part of the account.
- Ending the test early because one side looks ahead after a few days. Leads swing; decide the run length in advance and include at least two full weeks.
- Judging on click-through rate alone. A more eye-catching headline can draw more clicks from people who never buy. Judge on conversions and cost per conversion where volume allows.
- Forgetting that pinned headlines still apply. If the phrase you are testing sits in an unpinned headline that rarely shows, the variation will collect little data on it.
How to act on it
Start with a written hypothesis: what you are changing, why you think it will help, and which metric decides it. Pick a phrase that appears in many ads so the test has enough volume, and set the split to 50%. Leave it alone until it reaches its end date. If you need to test bids, budgets or bidding strategies rather than wording, use a campaign experiment instead, because ad variations only change the ads.
Keep a simple log of each test, the result and what you applied. Over a year that log becomes a record of what your buyers respond to, which is useful well beyond Google Ads: it informs landing pages, emails and even sales scripts. If you are new to testing in general, the principles in A/B testing apply here too.