Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Google Ads & PPC › Search Campaigns

Responsive Search Ads

How responsive search ads assemble headlines and descriptions, what ad strength means, when to pin, how to write assets, and how to test ads.
Guidebeginner11 min readUpdated

How a responsive search ad is assembled

Google documents the mechanism plainly: the advertiser enters up to fifteen headlines and up to four descriptions, and the system tests different combinations, learning which perform best for which searches. When the ad shows, it displays a selection of those assets, up to three headlines and two descriptions, with the number depending on the device and the space available; not every impression shows the full set. The headlines can appear in any order unless pinned, so each must make sense on its own and alongside any other.

The assembly is informed by the query. Google states that the system can show headlines that are more relevant to what the person searched, which is why an ad group's keyword theme should be reflected in several headlines rather than one: the more relevant assets there are, the more combinations can be relevant. It is also informed by what has already performed, so an ad that has served for weeks is a different ad from the one that launched, in the sense that the combinations it favours have narrowed.

Two consequences matter for the writer. There is no "the ad" to review; there are assets and the combinations report that shows which groupings served most. And the ad's relevance to the keyword, which is one of the three components Google rates in Quality Score, is a property of the pool as a whole: an ad group whose headlines never mention what its keywords are about will rate poorly however the pieces combine. The Quality Score guide covers the diagnosis; this page covers the writing.

Assets around the ad, such as sitelinks, callouts and images, are a separate system with their own guide on ad assets; they extend the ad but are not part of its headline pool.

Writing headlines and descriptions that combine

Because any headline can sit beside any other, the pool is written as a set of roles rather than a sequence. Most practitioners build it from a short list of message types, several of each, so that whichever three the system selects, the ad has a subject, a reason and a prompt.

RoleWhat it carriesExample (for "boiler servicing" ad group)How many
Keyword headlineThe ad group's theme in the words people searchBoiler Servicing in LeedsSeveral, with variations in phrasing
Benefit or proofWhy this provider: speed, guarantee, qualification, reviewsGas Safe Registered EngineersSeveral, each a different proof
Offer or price frameA reason to act now, stated without numbers if they changeFixed Price Servicing, No Call-Out FeeA few
Call to actionWhat to do nextBook Online in Two MinutesA few, varied in tone
BrandWho is speakingNorthern Heating Co.One or two
DescriptionA full sentence combining theme, proof and actionSame-week boiler servicing by Gas Safe engineers, with a written report and a fixed price. Book online today.All four, each self-contained

Rules that follow from the mechanism:

  • Each headline stands alone. No headline should depend on another to make sense, because the other may not be shown. A headline that reads "...and Free Delivery" is broken most of the time.
  • No two headlines say the same thing. Google's guidance asks for distinct headlines, and near-duplicates waste slots the system could have used to test a different message. Different phrasing of the same idea is fine; the same phrasing with one word changed is not.
  • Reflect the keywords, in several forms. Headlines that contain or paraphrase the ad group's keywords give the system relevant assets to choose for the query. This is where ad group structure and ad writing meet, and the match types guide explains why a tightly themed ad group makes this possible.
  • Match the landing page. The headlines promise; the page keeps the promise. A headline about a service the page does not mention lowers the chance of conversion (CVR) and, over time, the landing page component of quality.
  • Use the full lengths, varied. Headlines have a short character limit and descriptions a longer one. A mix of short and long headlines gives the system options for different devices; descriptions should use most of their length to carry the detail the headlines cannot.
  • Write in the reader's language. Sentence case or title case consistently; no exclamation marks in every line; no claims that policy forbids; British spelling for a British audience. When an AI tool drafts the pool, the same rules apply, and the AI-assisted ad copy testing guide covers tying drafts to a hypothesis and checking claims and lengths before launch.

The craft overlaps with writing titles and snippets for organic results, and the title tags and meta descriptions guide covers the same discipline from that side: a short promise that matches the query, and a longer line that earns the click.

Pinning: when it helps and what it costs

Pinning fixes an asset to a position: a headline pinned to position one always shows there, a description pinned to position two always shows there. Google documents that pinning limits the combinations the system can test, and advises using it only when a message must appear. The cost is real: each pin removes every combination that would have placed something else in that slot, and the learning that depends on variety slows. A fully pinned ad is an expanded text ad in disguise and forfeits the format's main advantage.

Pinning is justified in a few cases. Regulated industries may need a disclaimer or a line of mandated wording in every impression. Brand campaigns usually want the brand name first. A promotion that the landing page is built around should appear every time the ad does. Where any of these applies, the lighter technique is to pin several assets to the same position: three different brand headlines pinned to position one, for instance, keeps the brand first while letting the system choose which phrasing. The documentation notes that ad strength takes pinning into account, so a heavily pinned ad will rate lower; that is the rating reflecting reduced variety, not a judgement on the copy.

What pinning does not do is improve performance by itself. Advertisers who pin to make the ad "read properly" are optimising for how the ad looks to them in the preview, and the preview is one combination out of many that will never all be seen by the same person.

Ad strength: what the rating measures

Ad strength is a rating from Incomplete through Poor, Average and Good to Excellent, shown as the ad is built. Google describes it as a measure of the relevance, quality and diversity of the ad's assets, with suggestions attached: add more headlines, include popular keywords in headlines, make the headlines more distinct, make the descriptions more distinct, unpin assets. It is calculated from the assets themselves, before and regardless of serving.

That makes it a checklist rather than a verdict. An ad with few headlines, several near-duplicates and nothing that mentions the keywords will rate poorly, and the suggestions will be correct. An ad that follows the guidance will rate well, and that tells you the pool is complete, not that the message is persuasive. Google states that ad strength is not a factor in the auction; the auction uses the quality of the ad as predicted at auction time, which depends on how the assets perform, not on the rating.

Use it this way: build the pool to Good or better as a floor, because a poor rating usually does mean the system has too little to work with; stop chasing Excellent if reaching it would mean adding headlines that say nothing or unpinning a line that must appear; and then judge the ad by its click-through, conversion rate and cost per result. The CTR calculator is a quick way to compare ads on the first of those when the reports show raw clicks and impressions.

Testing ads: variations, experiments, what to read

The old test, two fixed ads rotated evenly and compared, no longer fits the format, because each responsive search ad is already running its own test inside. Testing now happens at three levels.

Within the ad. The asset report labels each headline and description with a performance rating once it has enough data (Learning, then Low, Good or Best), and the combinations report shows which groupings served most. Replace Low assets with new messages; keep Best ones; leave Learning ones alone until they have data. This is the routine, monthly in a busy ad group and quarterly in a quiet one.

Between ads in an ad group. Two responsive search ads with different angles, one leading on price and one on quality, say, give the system a choice. Google's ad rotation setting defaults to optimising for the better performer, so the comparison is not a controlled test, but the reports still show which angle earns the impressions and conversions. An even rotation is available where a fair split is needed.

Across the campaign. Ad variations apply a change (a replaced phrase, a new description) across many ads at once and report it as a controlled split; custom experiments split a campaign's traffic between the original and a trial version. These are the proper tools for a controlled A/B Test of a message, a landing page or a structure, because they hold everything else constant.

Whatever the level, read conversions and cost per result before click-through. A headline that wins clicks and loses conversions has selected the wrong people; a lower click-through with a better conversion rate is often the better ad. Where the campaign's purpose sits in the marketing funnel decides the reading: an awareness campaign can judge on clicks, a lead campaign cannot.

Illustrative example — not a real result

Replacing low-rated assets in a solicitor's ad group

A conveyancing firm's "conveyancing quote" ad group has run one responsive search ad for three months on £90 a day. The asset report rates two headlines Best ("Fixed Fee Conveyancing Quote", "No Move, No Fee"), seven Good, three Low and three still Learning. The Low headlines are generic ("Trusted Local Solicitors", "Expert Legal Advice", "Contact Us Today"). The firm replaces them with three specific ones: "Quote in Under an Hour", "Rated 4.9 From 600 Reviews" and "Online Case Tracking Included". Over the next six weeks the ad's click-through moves from 6.1% to 7.4% and the cost per quote request falls from £31 to £27 on the same budget. The firm records the change, waits for the new headlines to leave Learning, and schedules the next review for three months later rather than touching the ad again.

Ads for different intents: search vs brand vs competitor

The same format serves three quite different jobs, and the pool should be written for the job.

Generic (non-brand) search. The person does not know the business. Headlines carry the keyword theme, the proof that this provider is credible, and the reason to choose it; descriptions carry detail. Variety matters most here, because the system is matching many different queries to the same pool.

Brand search. The person already knows the business and is navigating. The brand name should appear first, usually pinned, and the rest of the pool should carry what the person might want next: the offer of the moment, the reasons to buy direct, the sitelinks that get them to the right page. Brand ads are the one place where a high click-through is nearly automatic, so judge them on what the clicks do afterwards.

Competitor search. The person was looking for someone else. Google's trademark policy limits using a competitor's name in ad text, and most accounts should not try; the ad instead states what this business offers that the comparison shopper would value, in neutral terms, and the headline that mentions the competitor is absent. Expect a low click-through and a higher cost per click than brand terms, and set the budget accordingly. The match types guide explains why competitor keywords need exact match. Whether a competitor campaign is worth running at all, and how to judge it on new customers rather than clicks, is the subject of the guide to bidding on competitor names.

How the author structures these three kinds of campaign in practice is described on the Google Ads expertise page.

For a design that tests ad copy and assets against a proper control, the search ad asset experiment covers the setup, the metrics and how to read the result.

When an ad rates Poor, the ad strength diagnosis decides from the ad's results whether the rating points to a real gap or can be left alone. Because ad strength is so often mistaken for the keyword's Quality Score, the comparison of Quality Score and ad strength sets out what each rating measures and which one to act on. When the ads serve but few people click them, the low click-through rate diagnosis checks the searches, the position, the assets and the competition before the headlines take the blame.

Common mistakes

  • Writing one ad and padding it. Fifteen headlines of which ten are rewordings of the same two ideas give the system nothing to test.
  • Pinning everything. Control of the preview at the price of the format's learning; the ad becomes a fixed ad with a worse rating.
  • Chasing Excellent. Adding empty headlines or unpinning mandatory lines to move a rating that is not an auction factor.
  • Judging on click-through alone. The combination that wins clicks may lose conversions; read cost per result.
  • Headlines that promise what the page does not deliver. Lower conversion now and a weaker landing page rating later.
  • Replacing Learning assets. They have not had time to earn a rating; replacing them restarts the clock.
  • Ignoring the combinations report. It shows what people actually saw; the preview shows what you imagined they would.

Frequently asked questions

Does a low ad strength rating mean the ad will perform badly?

Not necessarily. Google describes ad strength as a measure of how well an ad's assets follow its guidance on quantity, diversity and keyword relevance, and it is calculated before the ad serves. An ad with pinned headlines or a narrow message can rate Average and still convert well, and an Excellent ad can still be beaten. Treat it as a completeness check, and judge performance from the ad's results.

How many responsive search ads should an ad group have?

Google allows a small number of enabled responsive search ads per ad group and recommends at least one. In practice, one well-built ad per ad group is enough to begin with, and a second with a different angle gives the system something to choose between once there is traffic. More than that in a low-traffic ad group splits the data so thinly that nothing can be learned.

Should I pin headlines to keep control of the message?

Pin only what must appear: a legal line, a brand name in a brand campaign, an offer that the landing page depends on. Each pin removes combinations from the pool, and Google notes that pinning can reduce the ad's performance by limiting what it can test. Pinning several assets to the same position, so the system can still choose among them, keeps most of the control at a fraction of the cost.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Also related

Sources

  1. About responsive search ads — Google Ads Help
  2. Google Ads Help — Google Ads Help

Related