A quick primer
Google's official position on this is clear: current best-practice guidance recommends keeping at least 2 RSAs with a Good/Excellent rating per ad group for coverage and reliability, but historical and practical data paint a more nuanced picture. Google's own ad help notes that advertisers who add a second RSA to an ad group that had only one (for a total of 2) see, on average, a 6.6% increase in conversions at a comparable cost per conversion — a real, measurable lift from going from one to two, but that doesn't mean more RSAs is always better.
There's an important practical nuance the industry emphasizes: consolidating data into one RSA speeds up learning — splitting the same resources across multiple ads means each gets less data, and the system needs longer to figure out which combinations perform best. That's a direct consequence of how RSAs work: the system tests combinations within one ad's asset pool, and the fewer impressions each ad gets, the slower and less statistically reliable the learning is for each of them individually. That's exactly why Google's earlier guidance leaned toward a single RSA per ad group as the default consolidation strategy, before later shifting the recommendation toward a minimum of two — a compromise between data reliability per ad and having a built-in comparison baseline.
What to check before you touch anything
- The ad group's current traffic — is there enough volume that splitting it between multiple RSAs won't stretch data accumulation out unreasonably long.
- Whether there's a specific hypothesis worth comparing (a different message, a different audience, different landing pages) — if so, a second RSA is justified as a controlled experiment, not as a "just-in-case" safety net.
- How saturated the single existing RSA already is (close to 15 headlines and 4 descriptions, with maximum functional diversity) — if the asset pool isn't maxed out yet, expanding one ad usually helps more than adding a second one with partly overlapping content.
- Whether existing RSAs already diverge in the semantics/audience they serve — if so, the question isn't "one or several," it's "is the ad group even structured correctly."
- How many RSAs are currently in the group — Google technically allows up to 3 per group, but that doesn't mean using the maximum is always warranted.
Possible approaches
- If the asset pool isn't fully built out (fewer than 15 headlines/4 descriptions with sufficient functional diversity) — first saturate one RSA to the limit rather than creating a second as a workaround for adding more text.
- If the asset pool is already saturated and there's a specific hypothesis to test (a fundamentally different message or USP) — add a second RSA specifically as a controlled test of that hypothesis, accepting that it will temporarily slow data consolidation.
- If the ad group is low-frequency (not much traffic to begin with) — favor a single, fully-loaded RSA: splitting data across multiple ads is especially costly for learning speed and reliability on low traffic.
- If the ad group is high-frequency and there's bandwidth for regular monitoring — 2 RSAs can be a reasonable balance between reliability (single consolidation) and having a comparison baseline, in line with Google's current guidance.
- Either way, add a third RSA to a group only for a clear, deliberate reason (testing a fundamentally different audience segment, for example) — by default, it's more likely to just fragment the data further than to add value.