Google Serves High Ad Strength RSAs More Often, But They Convert Worse

By Priya Raman

Google now directs more impressions to whichever responsive search ad in an ad group carries the higher ad strength rating — and that ad posts the better conversion rate only 45% of the time. A new study of more than 5,900 Google Ads accounts by Brad Geddes, co-founder of Adalysis, puts hard numbers on a question PPC managers have argued about since RSAs became the default format.

The study follows up on Adalysis's 2022 ad strength research, giving the industry a four-year view of how Google's ad serving has drifted. Geddes previewed the findings in a LinkedIn post last week, calling ad strength "one of Google Ads' most misunderstood metrics."

Where Google's Position and the Data Collide

Google's Help Center describes ad strength as a guide for ad creation. It says the rating "doesn't directly influence your ad's serving eligibility" and plays no part in calculating Ad Rank or Quality Score. Ginny Marvin, Google's Ads Product Liaison, wrote in a March 2024 post on X that ad strength "isn't used directly in the auction & shouldn't limit your testing."

Geddes's data speaks to a different claim. "Not an auction input" and "doesn't affect which of your ads gets shown" are separate assertions. The second one is where the numbers bite.

What Happens When Two Ads Go Head to Head?

The study isolates ad groups with two active RSAs of different ratings, sharing the same keywords, bids and budget. In those groups:

• The higher-rated ad received more impressions 63% of the time.
• An Excellent ad beat a Poor ad on impressions in 78% of ad groups.
• Ads two ratings apart won 66% of the time; even Good versus Average went to the higher-rated ad 59% of the time.
• The higher-rated ad had the higher CTR in 52.5% of ad groups — barely better than a coin flip.
• The higher-rated ad had the higher conversion rate in only 45%, meaning the lower-rated ad converted better in 55% of ad groups.
• On conversions per 1,000 impressions (CPI, in Adalysis's terminology), the lower-rated ad won 54.3% of the time.
• The lower-rated ad had the lower CPA in 51.5% of ad groups. Across all ad groups, higher-rated ads ran an 11.8% higher CPA — a $100 CPA on a lower-rated ad corresponds to $111.80 on its higher-rated rival.

Geddes's warning for anyone running ad tests: Google "will serve the higher-rated ad more often, even if it has a worse conversion rate or CPI."

Four Years of Drift

Comparison with the 2022 study shows the direction of travel:

• More impressions to the higher-rated ad: 56.8% of ad groups in 2022, 63% now.
• Higher CTR: 48.5% in 2022, 52.5% now.
• Higher conversion rate: 45.8% in 2022, 45.1% now.

Google's picks align with clicks slightly better than four years ago, but not with conversions. Geddes concludes that Google "has increased its reliance on ad strength to determine what ad to show in an auction over the years."

Why the Lower-Rated Ad Often Wins

Geddes attributes most of the conversion edge to pinning. In most mixed-strength ad groups, one ad was pinned and the other was not. Pinned ads earn lower ratings and post higher conversion rates. The logic is familiar to any practitioner: you pin a price, a disclaimer or a qualifier to filter the click, not chase it. Ad strength rewards variety and keyword coverage, and a pinned ad written to pre-qualify buyers works against both.

Among unpinned ads, the picture reverses. CTR rose with every step up in ad strength, as did CPI; conversion rates rose slightly at each step except Poor to Average. For unpinned ads, chasing a higher rating generally pays off. Ad strength predicts performance well when you leave Google's system alone, and poorly when you deliberately override it.

How Many Headlines Excellent Actually Takes

Only a third of unpinned ads in the dataset reached Excellent. Without pinning, an ad can reach Good with as few as seven assets; Geddes recommends at least 11 different headlines for Excellent. Volume alone does not do it: only 44% of ads with 15 or more headlines reached Excellent, and one ad with 38 assets was still rated Poor.

Pinning raises the bar. The minimum asset count to reach Good jumps to 10. Pinning two positions caps an ad at Good and often leaves it at Poor or Average — stricter than 2022. If you pin all three headline positions, plan on about 13 assets. Pinning three or more different assets to the same position still leaves room for Excellent; pinning only one or two caps the ad at Good. Headline one carries more weight than headline two, which carries more than headline three.

What to Change in Your Accounts

Geddes's practical guidance for practitioners:

• Sort RSAs by pinning status before reading any rating. Tag each ad as unpinned, partly pinned or fully pinned. Judge each group on its own.
• Push unpinned ads to Excellent with variety, not volume: at least 11 headlines mixing keywords, benefits, features, calls to action and authority statements. If an unpinned ad sits at Average with 15 headlines, repetition is the likely culprit — rewrite near-duplicates before adding more.
• Rebuild pinned ads rather than abandoning them: remove the pins, build to Excellent, then re-pin and keep adding assets until the rating reaches at least Average, or Good with a single pinned headline.
• Judge tests on CPI and CPA, and give mismatched tests more time. It is fine to run ads with different ratings in one ad group, but expect the lower-rated ad to collect far fewer impressions, which stretches the time to statistical significance. If a lower-rated, pinned ad wins on CPA, keep it running.

Geddes closes his study with a line every PPC manager should keep: "often, the human ability to create compelling ads will outperform the machine's ability to serve the ads properly." The caveat is that the machine still decides how often your best ad gets seen. Google tells advertisers what the rating is not; Geddes's 5,900 accounts tell them what it does.