Responsive search ads: how to write headlines and descriptions, and how the ad strength rating relates to performance

Responsive search ads are Google’s default text format for search, built around supplying many headlines and descriptions so the system can assemble combinations per auction. This guide focuses on the mechanics that affect what shows, how pinning and ad strength behave, and how to read the current asset reporting when you are writing and maintaining RSAs in live accounts.

How responsive search ads assemble and show combinations

Per Google’s documentation, each responsive search ad can contain up to 15 headlines and 4 descriptions, and you must provide at least 3 headlines and 2 descriptions to create one. In any given impression, Google can show up to 3 headlines and up to 2 descriptions, with a minimum of 1 headline and 1 description required to serve. The second and third headlines and the second description are not guaranteed to appear in every impression.

Google’s help pages describe several enhanced flexibility behaviors for RSAs. Headline text may appear at the start of the description line, which means a headline asset can function as description copy in some combinations. Up to 2 unused headlines from the same ad can serve as link-based assets in the former sitelink space, and unused headlines or descriptions from another ad in the same ad group can also serve as link assets. Only active assets are eligible for this link-based use, and sensitive verticals are excluded from this behavior.

These mechanics mean that when you write RSA assets you are writing for multiple roles. A headline might appear as H1, H2, or H3, be reused at the start of a description, or serve as a link-based asset if it is not selected in the main combination. Descriptions can appear as D1 or D2, and unused descriptions can also be pulled as link assets from other RSAs in the ad group. You control which assets are eligible by enabling or pausing them, and by which RSAs you keep active in the ad group.

Character limits and minimums to publish an RSA

Google’s RSA limits are strict character counts per field. Headlines are limited to 30 characters each, descriptions to 90 characters each, and the 2 path fields to 15 characters each. For double-width languages such as Korean, Japanese, and Chinese, each character counts as 2 toward these limits, per Google’s help pages.

To publish a responsive search ad you must provide at least 3 headlines and at least 2 descriptions, within the overall maximum of 15 headlines and 4 descriptions. An RSA also requires a final URL, an associated ad group, and keywords in that ad group, since Google documents that an RSA will not serve without keywords. If any of those elements are missing, ad strength will show as Incomplete and the ad cannot serve.

How pinning behaves, what it costs, and when to use it

Pinning in RSAs restricts which assets can show in each position. Google’s help pages state that content pinned to Headline 1, Headline 2, or Description 1 always shows when the ad serves, while pins to Headline 3 and Description 2 are not guaranteed to appear. If you have must-appear text such as legal disclaimers, Google directs you to place and pin that text in H1, H2, or D1.

Google also documents the cost side of pinning. Pinning reduces the matchable combinations that the system can test, is not recommended for most advertisers, and can affect ad strength. If you pin every position, unpinned assets never show, because the system is constrained to the pinned text in each slot.

To mitigate this, Google’s guidance is to pin 2 or 3 unique assets per position when you must use pinning, and to avoid identical text pinned into the same slot. For example, if compliance requires a specific phrase in H1, you can create 2 or 3 different headlines that all contain that phrase and pin them to H1, which preserves some variation while satisfying the fixed-text requirement. This addresses the misconception that pinning is harmless control; Google’s documentation is explicit that it reduces combinations and can lower ad strength.

Pinning also does not guarantee every pinned position will show in all cases. Only H1, H2, and D1 pins are documented as always showing, so the belief that pinning guarantees your text shows in any slot is incorrect. H3 and D2 pins are eligible but not guaranteed in every impression.

What ad strength measures and how it relates to the auction

Ad strength for RSAs is a dynamic rating that runs from Incomplete through Poor, Average, Good, and Excellent, with suggested action items. Google’s help pages describe it as a feedback tool that evaluates asset diversity, relevance to the ad group’s keywords, and the system’s ability to form effective combinations. It is not a measure of creative quality in a general sense, but a coverage and variation check against Google’s documented factors.

Google states explicitly that ad strength is not used to calculate Ad Rank, Quality Score, or auction wins, and that it does not directly influence serving eligibility. This corrects the misconception that ad strength affects your auction results as a scoring input. The documented relationship is that advertisers who improved RSA and sitelink ad strength from Poor to Excellent saw 15% more conversions on average, based on Google internal data from an August 2025 range, which is an observed performance uplift rather than an auction mechanic.

The factors Google lists for higher ad strength include asset-keyword relevance, with the guidance that a keyword should fit entirely within one headline and that keywords longer than 30 characters should be placed in a description. Ad strength also considers whether you are close to the maximum of 15 headlines and 4 descriptions, whether your assets are unique and diverse rather than repetitive, and whether you have at least 6 sitelinks across account, campaign, or ad group levels to reach Good or better. Incomplete status can also appear when required elements such as final URL, ad group, or keywords are missing.

This means ad strength is a structured checklist for coverage, not a Quality Score proxy. The misconception that ad strength and Quality Score are directly linked is not supported by Google’s documentation, which only states the feedback role and the internal uplift statistic.

How to read RSA asset performance without Low/Good/Best labels

Google has deprecated the Performance label column that used Low, Good, Best, and Learning for RSA assets. For dates on or after June 5, 2025, the ad-level asset report instead provides full per-asset statistics such as impressions, clicks, conversions, conversion value, and cost, per Google’s help pages. This corrects the misconception that asset performance labels tell you what is working, because those labels are no longer the active reporting mechanism.

There are two key caveats in Google’s documentation for this asset reporting. First, asset metrics are non-summable, because each asset that appears in a served ad registers its own impression. If one impression shows 3 headlines and 2 descriptions, each of those 5 assets receives an impression count, so summing impressions across assets will exceed the ad’s total impressions. Second, ratio metrics such as click-through rate at the asset level are directional only, since the impression counting method differs from ad-level reporting.

Google’s optimization tip in the help pages is to consider replacing assets with zero impressions after several weeks. This is a practical use of the new metrics: assets that never serve despite being eligible are not contributing to combinations and can be swapped for new variations. The non-summable nature of the data means you should evaluate assets comparatively within the RSA rather than trying to reconcile them to campaign totals.

Writing RSA assets using Google’s documented guidance

Google’s RSA writing guidance focuses on how assets combine, not on generic copy formulas. The help pages instruct you to write the first 3 headlines as if they will appear together, since up to 3 headlines can show in one impression. This means H1, H2, and H3 should make sense as a coherent message in any order, without relying on one specific sequence.

Google also recommends including at least one keyword in your headlines, and states that a keyword should fit entirely within a single 30-character headline. For keywords that exceed 30 characters, the guidance is to place them in a description instead, which has a 90-character limit. Text that matches the user’s query may be auto-bolded in the ad, which is another reason to align headline text with the ad group’s keywords.

Ad strength favors maximizing unique headlines, so Google’s documentation encourages you to avoid repeating the same phrases across multiple headline slots. The system evaluates uniqueness and diversity, and repetitive assets can lower ad strength by reducing the number of distinct combinations it can test. With text customization features enabled, Google notes that ad strength counts Google-generated assets alongside yours, which affects how many total variations the system sees.

For sitelinks and link-based assets, Google’s ad strength guidance includes having at least 6 sitelinks across account, campaign, or ad group levels to reach Good or better. Since unused headlines and descriptions can serve as link-based assets in the former sitelink space, writing clear, destination-relevant text for those assets supports this requirement. If you need the broader account structure context that makes keyword-in-headline writing feasible, see the reference on themed ad group structure.

How many RSAs per ad group and how to test them

Google’s limits allow up to 3 enabled responsive search ads per ad group, and the help pages recommend implementing at least 2 RSAs with Good or Excellent ad strength in each ad group. Google’s internal data shows that adding a second RSA in an ad group is associated with a 6.6% average increase in conversions at a similar cost per conversion, and adding a third RSA is associated with a further 3.7% increase. This corrects the misconception that more RSAs per ad group is always better, since there is both a hard cap of 3 and diminishing documented uplift.

In practice, this means you should plan for 2 RSAs per ad group as a baseline, each built to reach Good or Excellent ad strength using the asset and sitelink guidance. A third RSA can be added when you have a distinct set of assets to test, recognizing that the incremental conversion gain Google cites is smaller than the gain from the second RSA. Within each RSA, you test at the asset level by supplying up to 15 unique headlines and 4 unique descriptions, then monitoring the per-asset metrics in the ad-level asset report.

Testing under this model relies on replacing under-serving or zero-impression assets rather than rotating whole ads on even serve. Since asset metrics are non-summable and directional, you compare assets within the same RSA and ad group, and you keep campaign-level performance (such as ROAS or cost per conversion) as the final decision metric. For how the relevance you build with this RSA craft shows up in diagnostic components, see the Quality Score reference.

Temper And Forge’s practice is to treat ad text as one of the auction-time quality levers that can be worked when costs rise or impression share is lost to rank, and to treat the visible Quality Score number as noise. The practical posture is to write for the themed ad group, use pinning only where compliance requires fixed text, and read per-asset RSA metrics directionally in the same way as Performance Max assets, with campaign ROAS as the real decision driver and ad strength as a coverage checklist rather than a score to chase.