Google replaced expanded text ads with responsive search ads in June 2022. Nearly four years on, RSAs are still the only standard Search ad format, and the typical agency account still runs them with five or six near-identical headlines, a couple of generic descriptions, and a "Poor" or "Average" ad strength score. The format changed. Most creative habits did not.
That gap is the opportunity. Agencies that actually treat RSAs as a machine learning problem, write assets that give the algorithm room to test, and stop pinning everything see better CTRs, lower CPCs, and more impression share. The ones treating RSAs like expanded text ads with extra fields stay stuck. Here is how to get it right.
What Are Responsive Search Ads?
A responsive search ad is Google's default Search campaign ad format. Instead of writing a single fixed ad with one headline and description, you supply up to 15 headlines (30 characters max) and up to 4 descriptions (90 characters max). Google's machine learning assembles the best combination for each auction, weighted by the search query, the device, the user's history, and other contextual signals.
At any given time, Google can show up to 3 headlines and 2 descriptions from your pool. The order and combination shift dynamically. Someone searching "affordable CRM for small business" might see a different combination than someone searching "best CRM software 2026," even if both queries hit the same ad group. That is the entire point of the format. Automated creative testing at a scale no human could match manually.
RSAs replaced expanded text ads (ETAs) in June 2022. Before that, you wrote individual ads with fixed headline and description combinations, then ran manual A/B tests to find winners. RSAs automate that. Google estimates that 15 headlines and 4 descriptions yield over 43,000 unique permutations, though in practice the algorithm narrows quickly to the combinations that work and ignores the rest.
For agencies, this shifts where the creative work happens. With ETAs, the job was writing the single best ad. With RSAs, the job is writing a varied set of assets so the algorithm has enough material to optimise. Different muscle.
Responsive Search Ads Examples
Here are five real RSA setups across industries. The point is not to copy them. It is to see what genuine asset variety looks like, the kind of variety that lets Google find combinations that win.
E-commerce product ad. Product specifics mixed with urgency and trust signals. Headlines: "Wireless Noise-Cancelling Headphones" / "From $149, Free Express Shipping" / "Rated 4.8 Stars by 2,000+ Buyers" / "Shop Now, 30-Day Returns" / "Premium Sound, Half the Price." Descriptions: "Shop our top-rated wireless headphones with active noise cancellation. Free express shipping on all orders over $100." / "Compare our range and find the perfect fit. Easy 30-day returns, no questions asked." Product name, price, social proof, CTA, value prop. Each headline holds up on its own and slots into any combination.
B2B SaaS ad. B2B headlines have to balance problem awareness with solution specifics. Headlines: "Stop Losing Deals to Slow Follow-Up" / "CRM Built for Sales Teams of 10-200" / "Trusted by 500+ B2B Companies" / "Start Your Free 14-Day Trial" / "Close 30% More Deals This Quarter." Descriptions: "Our CRM automates follow-up sequences so your reps spend time selling, not updating spreadsheets. No credit card required to start." / "See why mid-market sales teams are switching. Full pipeline visibility, automated workflows, and reporting that actually helps." Pain point, solution scope, social proof, CTA, outcome promise.
Local service ad. Local businesses live or die by geographic relevance and trust. Headlines: "Sydney's Top-Rated Plumber" / "Same-Day Emergency Callouts" / "Licensed & Fully Insured" / "Free Quotes, No Obligation" / "Serving the Inner West Since 2008." Descriptions: "Fast, reliable plumbing for homes and businesses across Sydney. Same-day service available for emergencies. Call now for a free quote." / "From blocked drains to full bathroom renovations. Licensed, insured, and backed by 500+ five-star reviews." Location, service type, urgency, credentials, longevity. Each gets a dedicated headline.
Legal services ad. Legal RSAs work best when they combine practice area specificity with accessibility signals. Most legal advertisers skip the second half. Headlines: "Experienced Family Law Solicitors" / "Free Initial Consultation Available" / "Over 20 Years in Family Law" / "Compassionate, Results-Driven Advice" / "Fixed-Fee Divorce Consultations." Descriptions: "Navigating separation or custody matters? Get clear, practical advice from solicitors who have handled thousands of family law cases." / "Book a free consultation to understand your options. Fixed-fee initial appointments available across Melbourne." Practice area, free consultation, experience, tone, pricing transparency.
Agency services ad. Agencies pitching their own services need headlines that signal specialisation and proof. Headlines: "Google Ads Management for SaaS" / "We Have Scaled 80+ B2B Accounts" / "Google Premier Partner Agency" / "Get a Free PPC Audit Today" / "Average 3.2x ROAS Across Clients." Descriptions: "Specialist Google Ads management for SaaS and B2B companies. We handle strategy, execution, and reporting so you can focus on product." / "See what is working and what is not. Request a free PPC audit and get actionable recommendations within 48 hours." Specialisation, track record, credentials, CTA, quantified results.
RSA Best Practices for Agencies
The gap between a mediocre RSA and a strong one usually comes down to seven practices that most agencies skip or half-implement.
Use all 15 headline slots. The algorithm needs variety to learn. Six or seven headlines in a 15-slot setup is like running an A/B test with a sample size of 20. There is nothing for it to draw on. More headlines mean more permutations, faster learning, and stronger long-term performance. Every additional headline gives Google exponentially more combinations to test.
Make every headline genuinely distinct. This is where most agencies fail. "Get Started Today," "Start Now," and "Begin Your Free Trial" are three headlines that say the same thing. Google's system flags it with a lower ad strength score, and rightly so. Each headline should cover a different theme: features, benefits, social proof, CTAs, pricing, urgency, brand name, or unique selling points. If you cannot articulate how a new headline differs from the existing ones, do not add it.
Include at least 3 keyword-rich headlines and 3 benefit-focused headlines. Keyword-rich headlines lift relevance signals and Quality Score. Benefit-focused headlines lift CTR. You need both. A working framework: three headlines that incorporate your primary keywords naturally, three that describe outcomes or benefits, two with social proof or credentials, two with CTAs, and the remaining slots for brand name, pricing, or urgency.
Write headlines that work in any combination. RSA headlines appear in random order unless pinned. "Headline A | Headline B | Headline C" needs to make sense regardless of which three Google selects, and in any order. Avoid headlines that only work as a sequence. "The Best CRM" followed by "For Growing Teams" as two separate headlines is a trap, because Google might never show them together.
Pin sparingly. Pinning locks a specific headline to a specific position. Pin your brand name to position 1 if brand visibility is non-negotiable. Beyond that, resist the urge. Every pin restricts the algorithm's ability to test, which is the whole point of using an RSA. If you pin all three positions, you have rebuilt an expanded text ad. Google's own data shows RSAs with no pins tend to outperform RSAs with multiple pins.
Include a clear CTA in at least one description. Descriptions give you 90 characters to make a case. At least one should contain a direct call to action. "Request a free quote." "Start your trial today." "Book a consultation." The others can elaborate on your value proposition, address objections, or add supporting detail.
Use keyword insertion for high-volume ad groups. Dynamic keyword insertion ({KeyWord:Default Text}) drops the user's search term into your headline. It works well for ad groups with many closely related keywords, because it creates a relevance signal that lifts both Quality Score and CTR. Use it in one or two headlines, not all of them. Lean on it too hard and your ads start sounding like form letters.
RSA Ad Strength: What It Means and How to Improve It
Google assigns every RSA an ad strength rating: Poor, Average, Good, or Excellent. The score updates live in the ad editor as you add or change assets. It is the most visible quality signal Google gives for RSAs, and it is widely misunderstood.
Ad strength is not a direct ranking factor. A "Poor" ad strength RSA can still win auctions and sit in top positions. Ad Rank comes from bid, Quality Score, and expected impact of ad extensions. What ad strength does correlate with is auction eligibility. Google has said that moving an RSA from "Poor" to "Excellent" is associated with about 12% more impressions on average. The mechanism is not ranking influence. It is reach, since ads with more diverse assets match a wider range of queries.
What drives the score. Four things. The number of headlines and descriptions provided (more assets, higher score). Headline variety (distinct themes versus repetitive messaging). Keyword relevance (whether your headlines incorporate the ad group's target keywords). Description quality (length, uniqueness, presence of a CTA). Google's system weighs all four in combination and gives specific suggestions alongside the score.
How to improve ad strength. Add unique headlines until you hit 15. Make sure they cover at least four distinct themes: features, benefits, CTAs, and social proof. Strip any headline that is a rewording of another. Drop your primary keyword naturally into at least two. Make both descriptions full-length, close to the 90-character limit, and meaningfully different from each other. If the score still says "Average," check whether your headlines make sense in isolation. Headlines that depend on context from other headlines tend to drag the score down.
One note on chasing "Excellent." A "Good" ad strength score is fine for most ad groups. Chasing "Excellent" sometimes pushes you to add headlines that dilute your message, or weak CTAs that exist purely to satisfy the variety requirement. If you are at "Good" with strong headlines and the actual performance metrics (CTR, conversion rate, CPA) are healthy, do not water down the ad to chase a label. The score is a guide, not a target.
Testing RSAs at Scale
RSAs do their own internal testing by rotating combinations. That does not remove the need for structured testing at the ad group and campaign level. Here is how to actually test RSAs when you are running dozens or hundreds of ad groups.
Run at least 2 RSAs per ad group. Google explicitly recommends this. With two RSAs in an ad group, Google splits impressions and gradually favours the stronger one. The simplest form of RSA testing, and it needs zero manual analysis. Just make sure the two RSAs have meaningfully different headline sets, not minor variations on the same themes.
Use experiment campaigns for big creative tests. If you want to test a different messaging angle (benefit-led versus problem-led, for example), do not run both in the same ad group. Use Google Ads experiments to set up a campaign-level split with controlled traffic allocation. You get statistically valid results without the noise of Google's uneven impression distribution within ad groups.
Measure at the ad group level, not the individual RSA level. Google does not distribute impressions evenly between RSAs in the same ad group. The RSA with more impressions is not necessarily "better." It may simply have been handed more opportunity. Judge RSA performance through ad-group-level metrics (CTR, conversion rate, CPA) over time. Comparing individual RSA stats head to head is mostly noise.
Refresh underperforming headlines every 4 to 6 weeks. The asset detail report (Ads & Assets > Assets > View asset details) shows performance labels for individual headlines and descriptions. Assets labelled "Low" have been tested enough, and they keep losing. Swap them out and give the algorithm another 4 to 6 weeks to learn. Do not replace everything at once. Two or three at a time keeps the algorithm's learning on your strong assets intact.
Use the asset detail report to identify winners. The same report labels each headline and description as "Best," "Good," "Low," or "Learning." "Best" headlines are your proven performers. Use them as starting points when you build RSAs for new ad groups in the same account or for similar clients. Over time, you build a library of validated headlines by industry and offer type. For a broader framework on scaling creative and bid tests across accounts, see ad optimisation strategies that scale.
RSAs and Performance Max
Here is something most agencies overlook. The text assets you write for RSAs in your Search campaigns feed straight into Performance Max. When you build a PMax asset group, the headlines and descriptions you provide are functionally identical to RSA assets. Google's system uses the same machine learning to assemble text combinations for PMax ads across Search, Display, YouTube, Discover, Gmail, and Maps.
That means RSA quality has a multiplier effect. Strong headlines written for a Search RSA improve that ad and every PMax ad pulling from the same pool. Weak headlines drag both down.
The practical move is to keep a master headline library for each client. Best-performing headlines and descriptions, organised by theme (features, benefits, CTAs, social proof, pricing). When you build a new Search RSA, you pull from the library and add campaign-specific variants. When you build a PMax asset group, same source. Consistency across campaign types, and every headline improvement benefits the entire account.
The library approach also speeds up onboarding. Instead of handing a junior media buyer a blank brief, you hand them the client's validated headline library and ask them to adapt and extend it. The quality floor sits higher because they are building on proven assets, not starting from zero.
For agencies running Performance Max budget pacing alongside Search campaigns, the library also creates a feedback loop. Headlines that perform well in Search RSAs (where you actually get granular reporting) tell you which text assets to prioritise in PMax (where reporting is murkier). Search becomes the testing ground. PMax becomes the scaling channel.
RSAs are not complicated. They do require a different creative discipline than the ETA format that came before. The agencies that treat RSAs as a machine learning system, feed it diverse high-quality inputs, and let the algorithm do its thing consistently beat the agencies that treat RSAs as slightly longer text ads. Write distinct headlines. Cover multiple themes. Pin only when you must. Test in a structured way. Feed your winners into PMax. That is the whole strategy.
Try Pace free to manage RSAs, budgets, and pacing across Google, Meta, TikTok, and LinkedIn from a single dashboard.
