Responsive search ads sit at the centre of a disagreement between Google’s own guidance and the largest independent dataset anybody has published. Both parties have an interest, both have published figures, and they point in different directions.
Knowing that before you optimise anything is worth more than either source alone.
This is for you if you have been chasing an Excellent ad strength rating and are not sure it is buying you anything.
What Google says about responsive search ads
Google’s guidance is specific and quantified.
Moving ad strength from Poor to Excellent is stated as worth 15% more clicks and conversions. A second responsive search ad in an ad group adds 6.6% more conversions, and a third adds 3.7%. Adding a logo and business name adds 8%.
Those are useful numbers and they come with an obvious commercial context. Google benefits when advertisers supply many unpinned assets and let the system combine them broadly, because that gives its matching more room to operate.
That does not make the figures wrong. It means they should not be the only source you consult, which is exactly what usually happens.
What twenty thousand accounts show instead
Optmyzr analysed responsive search ads across roughly 20,000 active Google Ads accounts in April 2026, and the headline finding runs against the guidance.
Ads rated Excellent had the worst cost per acquisition in the dataset, at $28.68. Ads rated Average had the best, at $12.43.
Within the same analysis, headlines under twenty characters produced $9.35 against $18.27 for longer ones. Sentence case beat title case by a factor of 3.7, at $7.46 against $27.47. Partial pinning performed best and full pinning worst.
A separate Optmyzr analysis across over a million ads and more than 22,000 accounts found no correlation between ad strength and conversion performance at all.
Optmyzr sells optimisation software and its dataset skews towards experienced practitioners running that software. Worth stating, and it remains the largest independent comparison available.
How to hold both sources at once
The disagreement is narrower than it first appears, and separating the claims helps.
Google’s figures about asset quantity are plausible and modest. A second advert adding 6.6% and a third adding 3.7% describes diminishing returns, which is consistent with the mechanism.
The contested claim is the ad strength indicator specifically. Google presents it as predictive of performance. Two independent analyses covering more than a million ads find it is not.
So you can accept the asset quantity guidance and decline the strength target without contradiction. Those are different claims and only one is disputed.
That is the practical resolution, and it lets you use both sources rather than picking a side.
What ad strength is actually good for
Treat it as a completeness check rather than a performance predictor.
When it flags that you have too few headlines, or that your assets repeat the same message, that is a genuine observation worth acting on. Thin asset sets do limit what the system can do.
When it says Excellent, that tells you the box is full. It does not tell you the advert will work, and the evidence says it may indicate the opposite.
So use it the way you would use a spellcheck. Helpful for catching omissions, silent on whether the writing is any good.
The failure mode is teams optimising towards the indicator, adding near-duplicate assets to satisfy it, and inflating the rating without improving anything real.
What to change in your responsive search ads instead of the rating
Four things, all supported by the independent data and all cheap.
Convert headlines to sentence case. Free, reversible, and the largest measured effect in that dataset.
Write shorter headlines on at least some assets, aiming under twenty characters rather than filling the character limit.
Unpin most of what you have pinned, keeping pins only where a legal or brand requirement genuinely demands a fixed position.
And supply distinct messages rather than variations of one. Quantity of genuinely different assets helps the combination engine. Quantity of near-identical assets inflates the strength indicator and gives the system nothing new to work with.
How many responsive search ads per ad group
Google’s own diminishing returns figures answer this reasonably well.
If the second adds 6.6% and the third adds 3.7%, then two is a sensible default for most ad groups, and a third earns its place where you have enough volume to read the difference.
Beyond that the returns are not published, and the practical constraint arrives anyway: each advert needs enough impressions to accumulate a readable history, and splitting a modest ad group across four adverts starves all of them.
The same arithmetic that governs creative count on Meta applies here. Divide your ad group volume by the number of adverts and ask whether each will produce enough to compare.
That question is more useful than any published recommendation about quantity.
Why this disagreement is worth knowing about generally
It is a clean example of something that recurs across paid media.
A platform publishes guidance that serves both the advertiser and the platform. An independent party with a large dataset finds the guidance does not predict what it claims to. Neither party is dishonest, and their interests differ.
The correct response is not cynicism about either. It is knowing that guidance from a platform is guidance from an interested party, and seeking a second source before optimising a whole account towards a single indicator.
That habit costs one search and it would have saved a great many accounts a great deal of effort chasing a rating.
What responsive search ads changed about the job
Worth stepping back, because the format altered what a search advertiser actually does.
Previously you wrote a complete advert and it ran as written. Now you supply components and the system assembles them, differently for different queries, without telling you which combination produced which result.
So the craft moved. It is no longer writing the best single advert. It is supplying components that combine well in any order, which is a genuinely different skill and one nobody was taught.
That explains why asset variety matters more than asset polish here. A beautifully written headline that only works in one position is worth less than three plainer ones that work anywhere.
It also explains why the pinning question is contested. Pinning restores the old control at the cost of the new mechanism, and partial pinning is the compromise the data supports.
What to change this week
Three steps.
Check whether anybody in your account is treating ad strength as a target. If so, that is the first thing to stop.
Then convert your headlines to sentence case and write shorter alternatives for at least half of them, since both are supported by the largest available dataset and cost nothing.
Then review what you have pinned, and unpin anything without a legal or brand reason.
Ad headlines covers the brevity and casing evidence in detail, fund the ad set and read what the algorithm chose is the position underneath, creative and CPM covers the equivalent auction mechanism, and Google Ads cost per lead covers the market backdrop. The case studies show what these accounts produced. To review your assets against both sources, book a teardown.