Responsive search ads and Ad Strength: writing ads that win the auction

How responsive search ads actually work

A responsive search ad is not one fixed ad. You write up to 15 headlines of 30 characters each and up to 4 descriptions of 90 characters each, and Google's machine learning assembles them into combinations for every auction. A served ad shows up to three headlines and two descriptions, picked to fit the device, the query, and the user. Over time the system learns which combinations earn clicks for which searches and serves more of what works. You are not writing one message; you are writing a parts kit and letting the auction choose the assembly.

This is why redundancy hurts you. If five of your headlines all say a near-identical version of the same thing, Google has fewer distinct combinations to test, and the ad shown for a given query is weaker than it could be. Quantity and diversity are the two levers you control directly. Google's own data shows the simple act of adding a second RSA to an ad group lifts conversions by 6.6% at a similar cost per conversion (Google Ads Help, 2024), because more raw material gives the system more room to find a winner.

What Ad Strength measures, and whether it matters

Ad Strength grades four things: how relevant your headlines and descriptions are to your keywords, how many assets you provide, how unique those assets are, and whether you have enough sitelinks at the account, campaign, or ad group level. It is a first-impression score built on what has worked for advertisers in aggregate. Crucially, Google is explicit that Ad Strength does not affect Ad Rank or Quality Score. A Poor rating does not mean your ad is deprioritized in the auction.

So does it matter? Partly. Google reports that advertisers who improve Ad Strength for their RSAs and sitelinks from Poor to Excellent see 15% more conversions on average (Google Ads Help, 2024), and Excellent ads tend to qualify for more impressions because the system predicts they will be relevant. But the Optmyzr analysis of roughly 20,000 accounts found no correlation between Ad Strength and actual results: Average-rated ads posted the lowest CPA at $12.43 while Excellent ads posted the worst at $28.68 (Optmyzr, 2026). Treat Ad Strength as a prompt to add assets and reduce repetition, not as the scoreboard. Your conversion data is the scoreboard.

The practical read: get out of Poor, aim for Good or Excellent by adding distinct assets, and then stop optimizing for the meter. Do not pin yourself into a worse-performing ad just to turn the label green.

Writing headlines that are distinct, not redundant

The goal is 15 headlines that say genuinely different things, because diversity is what gives Google useful combinations. Keyword stuffing the same phrase into every slot no longer helps and can hurt. A kitchen remodeler should not write five variants of "Kitchen Remodeling Services." It should spread across angles: the service ("Kitchen Remodeling Services"), the proof ("Licensed & Insured in Pasadena"), the outcome ("The Kitchen You've Been Picturing"), the offer ("Free In-Home Estimate"), and the urgency ("Booking Spring Projects Now"). Each headline carries a different reason to click, so any combination Google assembles still reads as a coherent ad.

Cover the three jobs every ad creative has to do. First, match the keyword and intent so the searcher sees their own words reflected back. Second, state the offer and a clear call to action so they know what happens next. Third, differentiate with proof, geography, or a guarantee so you stand out from the four other ads on the page. Plan your 15 headlines as a grid across those jobs rather than writing them one at a time, and you avoid the redundancy that flattens Ad Strength and wastes combinations.

One quiet finding worth knowing: the Optmyzr study reported sentence case headlines hit a $7.46 CPA versus $27.47 for Title Case, a 3.7x difference (Optmyzr, 2026). It is a single study, not gospel, but it is a cheap thing to test.

When to pin, and when to leave it alone

Pinning locks a headline or description to a fixed position so it always shows. Pin sparingly. Every pin removes a degree of freedom from Google's testing, and pinning a single asset to one slot can cut your testing potential by roughly 75% because that position can no longer rotate. Reserve pinning for things that genuinely must appear: a legal disclaimer, a regulated claim, your brand name in position 1, or a specific offer you have to control.

The fix when you do need to pin is to pin two or three assets to the same position rather than one. That keeps the slot under your control while still letting Google test which of your approved variants performs best. The Optmyzr data found partial pinning actually delivered the best results, a $13.68 CPA against full pinning and no pinning (Optmyzr, 2026), which lines up with the logic: constrain what you must, free what you can. Pin your CTA to position 2 or 3 where users expect the offer, leave position 1 open for the system to match intent, and never pin so heavily that you collapse the ad down to one fixed combination.

Use every asset, not just the ad itself

The ad text is only part of what shows on the results page. Assets, formerly called extensions, expand your ad's footprint, push competitors down, and give people more ways to act. Sitelinks send users to specific pages like Pricing, Reviews, or a Free Quote form. Callouts add non-clickable trust phrases like "Licensed and Insured" or "24/7 Emergency Service." Structured snippets list specifics like service types or brands. Image and location assets make local ads more tangible. These are also why Ad Strength asks for six or more sitelinks: assets are part of the creative, not an afterthought.

For a local or service business, assets often carry the conversion. A homeowner searching at 11pm for an emergency plumber clicks the "Call Now" or "Book Online" sitelink more readily than the headline. Add at least four sitelinks, four callouts, and a couple of structured snippet sets to every campaign, and supply them at the ad group level where the offer is specific enough to warrant it. The more relevant assets you provide, the larger and more useful your ad becomes, and the more of the page you own against the competition.

Testing and iterating your ad copy

Responsive search ads test themselves at the combination level, but you still steer the test. Use the asset detail view, where Google labels each asset Low, Good, or Best, to find your dead weight. Swap out every Low asset for a fresh angle and leave the Best ones alone. Do this on a cadence, not constantly, since ads need impressions to accumulate a verdict. Then run the combinations report to make sure no assembled ad reads awkwardly or contradicts itself, which is the main risk of a loosely written kit.

For bigger swings, run two RSAs per ad group and let Google's ad rotation, or a formal experiment, surface the winner on the metric you care about. Judge on cost per conversion and conversion rate, not on Ad Strength or even click-through rate alone, since a high-CTR ad that attracts the wrong clicks is a tax, not a win. Give each change two to four weeks and enough impressions before you call it. Iteration is the whole point of the format: you are not writing the perfect ad once, you are feeding the system better raw material over time.

How WellBuilt writes Google Ads creative

WellBuilt treats RSAs as a managed, ongoing discipline rather than a one-time setup. We start by mapping your 15 headlines across the jobs an ad has to do, keyword and intent, offer and CTA, and proof and differentiation, so the assets are distinct and Google has real combinations to test. We pin only what compliance or brand genuinely requires, and when we pin we use multiple variants per position so testing keeps running. We build out sitelinks, callouts, structured snippets, and image assets at the right level, because the assets often do the converting.

From there it is iteration on a schedule. We review the asset detail report, replace Low-rated assets, check the combinations report for anything that reads wrong, and judge ads on cost per lead and conversion rate rather than the Ad Strength meter. We will push for Good or Excellent strength where it costs nothing to get there, and we will tell you plainly when chasing the label would make a worse-performing ad. What we do not do is promise specific lift numbers from copy changes; we test, measure against your conversion data, and keep the combinations that earn their place.