Ad Strength is Google Ads' rating of a responsive search ad's asset set — Poor, Average, Good or Excellent. It grades whether you gave the ad-assembly machine enough parts: enough headlines and descriptions, enough distinct messages, enough keyword coverage. It is a completeness meter, not a quality verdict, and by Google's own description it sits outside the auction entirely.
That framing matters because the label appears where a KPI would appear — in a column, next to every ad — and so it gets managed like one. Accounts optimised to please the label fill up with fifteen interchangeable headlines and a proud Excellent badge. The label improved; the ads did not.
How Ad Strength works
Ad Strength is evaluated from the assets you supply, before anything serves. Four inputs move it:
- Quantity. A responsive search ad accepts up to 15 headlines and 4 descriptions; more assets mean more possible combinations, and the rating rewards that.
- Variety. Headlines that say different things score better than five rephrasings of one thing.
- Keyword relevance. Including your important keywords in headlines raises the rating.
- Pinning. Fixing assets to positions constrains assembly, so pins push the label down.
Google states the rating is guidance: it is not a component of Ad Rank and does not change your cost per click. Independent practitioner analyses have repeatedly found only a weak relationship between the label and conversion outcomes. An Excellent ad is not reliably a better ad — it is a better-fed machine.
Which is why a Poor rating is sometimes the correct rating. An ad that must lead with a pinned legal disclaimer will never grade well and should not try. A deliberately narrow ad, written to make one exact promise, trades the label for control on purpose. The tie-breaker is the ad's own record: if it converts at acceptable cost, the label loses the argument.
How an agent audits Ad Strength account-wide
The Google Ads API exposes the rating for every ad, so an agent with account access can do what the interface never quite offers: pull every responsive search ad with its Ad Strength, pin status and conversion history in a single pass. A useful prompt — list every ad rated Poor or Average alongside its conversions and pinned assets, then sort them into "genuinely thin, needs real variants" versus "deliberately constrained, leave alone". That is an afternoon of clicking done in a minute, and it turns the label back into what it was meant to be: a screening signal, not a target.
The trap is the obvious instruction. Tell an AI to "improve Ad Strength account-wide" and it will comply — by generating filler variants that say nothing new, which is precisely the failure the metric was supposed to prevent. The better brief names the input rather than the score: cover each distinct selling point once, keep every headline sayable on its own, and let the rating land where it lands. Drafted variants should arrive as reviewable changes, so nothing goes live on the strength of a label.
Ad Strength only makes sense against the format it grades — see responsive search ads for how assembly works, writing RSAs with AI for the drafting workflow, and the AI-assisted account audit for where the label belongs in a full review.