What are responsive search ads (RSA)?

A responsive search ad is not a finished ad but a store of parts. The advertiser writes between 3 and 15 headlines of up to 30 characters each and between 2 and 4 descriptions of up to 90 characters, plus two path fields of 15 characters that only dress up the visible URL. None of those parts has a reserved place in the final result. Google assembles the ad at auction time and picks which headlines and which descriptions appear, and in what order.

The change in craft is considerable. The advertiser stops writing a message and starts writing an inventory of messages that must hold up in any combination. A headline that depends on the previous one to make sense falls apart as soon as the system separates or reverses them. That is the real difference from the earlier format, the expanded text ad, where every line held a guaranteed position and the copy always read the same way.

Since 30 June 2022 this format has been the only search ad type that can be created or edited in standard Search campaigns. Expanded text ads that already existed kept serving and can be paused, resumed or removed, but not created anew. Anyone opening a search campaign today is not choosing between formats, because this is the only one left.

When a query triggers a keyword in the ad group, the system selects at least one headline and one description from the available pool and presents them in varying orders and combinations. The served ad can show up to three headlines and up to two descriptions, depending on the device and the space available on the results page. Every impression can be a different combination, so the same ad rarely looks identical twice in a row.

Pinning is the only direct brake on that distribution. Pin a headline to position one and that text always appears there, with no other headline able to take the slot. It makes sense for a legal obligation, for the brand name or for a condition that must never be missing. The price is arithmetic: every pinned position wipes out at a stroke all the combinations that would have used a different asset in that place. Pinning three headlines to three positions turns the responsive format into a static ad with extra steps.

There is a rarely used middle path. Several assets can be pinned to the same position, so the system keeps choosing among them inside that slot and part of the rotation survives. On top of that come campaign-level text assets, up to three headlines and two descriptions that apply to all active responsive ads in the campaign and can be scheduled with a start and end date, useful for time-limited promotions without touching every ad one by one.

Ad strength is the indicator that diverts the most budget in this format. It sits next to the ad with five levels, from incomplete to excellent, and a good share of teams reads it as a performance grade. It is not. It measures the relationship between the assets supplied and the keywords the group targets, the variety of the text and the number of combinations available. It is a completeness check on the ad, not a sales forecast.

Google puts it plainly: ad strength is not used to calculate Ad Rank, Quality Score or auctions won. Nor does it judge the quality of the landing page or the appeal of the offer, which is where a large part of the commercial result is decided. An ad rated excellent can bill less than one rated average whose message lines up better with what the company actually sells.

Optimising towards the indicator carries a measurable cost. To climb a level, generic headlines get added, pins that protected a mandatory detail get released, and the message that was converting gets diluted. The sensible decision is made with conversions, revenue and margin per asset combination, leaving the strength bar as what it is, a checklist to use while writing.

• Write every headline as a self-contained unit that reads correctly on its own and in any position of the ad.
• Cover distinct angles across the headlines: price, delivery time, warranty, geographic scope, customer type. Restating the same idea in other words produces no useful combinations.
• Pin only what is mandatory by law or by brand policy, and consider pinning several assets to the same position to keep some rotation.
• Keep each ad group tightly themed, because concrete headlines are only possible when the incoming queries resemble one another.
• Check the asset and combination reports to see which text is actually being served, instead of inferring it from what you wrote.
• Use campaign-level text assets for dated messages such as a seasonal promotion, and switch them off once it ends.

• Pinning nearly every asset to win back the control of the old format, which shrinks the possible combinations to a handful and cancels the point of the ad.
• Treating ad strength as a results forecast and changing profitable copy purely to move from good to excellent.
• Filling all fifteen headlines with near-identical variants of the same sentence, so the system rotates between messages that say the same thing.
• Writing headlines that form one continuous sentence and turn into nonsense once the system reorders them or serves only one of them.
• Confusing the path fields with the real URL and promising there a section of the site that the landing page does not contain.