For anyone writing Google Ads copy in 2026, Responsive Search Ads (RSAs) are not one option among several — they are the format. The older expanded text ads are long retired, and the RSA is what every search campaign runs. That makes RSA copywriting a core skill, and yet most accounts treat it as an afterthought: a handful of headlines typed in once, a couple of pinned assets for control, and no ongoing iteration. The result is mediocre Ad Strength, generic combinations, and click-through rates that leave money on the table in every auction.
This tutorial is practical and copy-paste ready. We cover how RSA assets actually work in 2026, when to pin and when to let the algorithm decide, what Ad Strength really measures and how to raise it without resorting to filler, a categorized library of 100+ headline templates you can adapt immediately, four description frameworks that convert, how to read the asset reporting Google gives you, and how to use AI to generate headline candidates the right way. By the end you will have both the understanding and the raw material to build genuinely high-performing search ads.
First: over-pinning. Pinning feels like control, but every pin removes combinations Google can test and almost always drops Ad Strength. Reserve pinning for genuine requirements — disclaimers, mandatory brand mentions, non-negotiable offers — and pin multiple assets to the same slot when you must. Second: writing 15 headlines that all say the same thing. The algorithm needs genuinely different angles to optimize; fifteen rewordings of one benefit give it nothing to test. The fix for both is the same discipline this guide teaches — provide varied, distinct, on-message assets and let the machine learning do its job within sensible guardrails.
How Responsive Search Ads work in 2026
A Responsive Search Ad is a single ad made of interchangeable parts. You supply the parts; Google assembles and serves the combinations most likely to perform for each query.
The components:
- Up to 15 headlines, each a maximum of 30 characters. Google shows up to three headlines in a given served ad (sometimes two, depending on space and device).
- Up to 4 descriptions, each a maximum of 90 characters. Google shows up to two descriptions in a served ad.
- The display path, two optional fields of 15 characters each appended to your display URL.
How assembly works: for each auction, Google's machine learning selects a combination of headlines and descriptions predicted to maximize performance for that specific query, user, and context. Over time and across many impressions, the system learns which assets and combinations perform best and serves them more often. This is why providing more varied assets helps — it gives the algorithm a richer space to optimize within.
Why variety matters more than volume: the algorithm can only test combinations that are genuinely different. If ten of your fifteen headlines express the same benefit in slightly different words, the effective testable space is small. Distinct angles — a benefit headline, a social-proof headline, an offer headline, a question headline — give the system meaningfully different messages to match against different searchers and intents.
The relationship to ad groups: RSAs inherit their effectiveness from ad group structure. Because Google rewards relevance, the headlines that perform best echo the keyword theme of the ad group and the searcher's intent. A tightly themed ad group lets you write headlines that are specifically relevant; a sprawling ad group forces generic headlines that please no particular query. Structure first, copy second. For the keyword-theming foundation, see our match types guide and our existing RSA copywriting method guide.
What Google controls vs what you control: you control the asset pool, pinning, and quality; Google controls which combinations serve and in what order (unless pinned). The skill is supplying excellent raw material and setting just enough constraint — not trying to manually engineer every served combination, which both fails and forfeits the algorithm's advantage.
Pinning strategy: when to control and when to let go
Pinning is the single most misused RSA feature. It lets you fix an asset to a specific position (Headline 1, 2, or 3; Description 1 or 2), and many advertisers reach for it instinctively to keep control. That instinct usually hurts performance.
What pinning does to the ad:
- A pinned asset only ever appears in its pinned position.
- Other unpinned assets compete for the remaining unpinned positions.
- The more you pin, the fewer combinations Google can test — and the lower your Ad Strength.
When pinning is justified:
- Legal or compliance requirements — a mandatory disclaimer, regulated-industry phrasing, or required terms that must always appear.
- Brand mandates — a brand name or trademark that must always be present or always sit first.
- Non-negotiable offers — a headline offer that must appear in every served ad for a campaign period.
The right way to pin when you must: pin two or three assets to the same position rather than a single asset. This guarantees something from your required set always appears in that slot while still giving Google a choice among them. For example, if a benefit must always lead, write three benefit headlines and pin all three to Headline 1 — the algorithm picks the best of the three for each query.
A practical rule: start with zero pinning. Only add a pin when you can name the specific requirement it satisfies. If the answer is just I want control, that is not a reason to pin — that is a reason to write better headlines so the combinations Google chooses are all good ones. The goal is an asset pool where every combination works, which removes the need for control in the first place.
Ad strength: what it measures and how to raise it
Ad Strength is Google's rating of your RSA's asset quality and diversity, shown as Poor, Average, Good, or Excellent with specific improvement suggestions. It is not a direct auction ranking factor, but it correlates with performance and serves as a useful build checklist.
What Ad Strength evaluates:
- Headline quantity — are you using enough of the available headlines?
- Headline diversity — are your headlines distinct, or repetitive?
- Keyword inclusion — do headlines include relevant, popular keywords from the ad group theme?
- Asset uniqueness — are you avoiding near-duplicate assets?
How to raise it without filler:
- Add more distinct headlines until you have at least 8-10 strong ones, ideally more. But add real angles, not padding.
- Increase diversity by spanning the categories — benefit, feature, social proof, offer, urgency, question, brand. Diversity is what the meter most rewards and what genuinely helps performance.
- Include the keyword theme in several headlines so relevance is clear, without keyword-stuffing every one.
- Remove duplicates — if two headlines say nearly the same thing, replace one with a new angle.
The trap to avoid: chasing Excellent by adding generic headlines that satisfy the meter but dilute your message. Google itself notes that improving Ad Strength from Poor toward Excellent is associated with more conversions on average — but that association assumes the added assets are genuinely good. Padding the meter with weak, off-brand, or generic headlines can hit Excellent while lowering real performance. The correct target is the highest Ad Strength you can reach with assets you would be proud to show — usually Good or Excellent with disciplined, varied copy.
Ad Strength is best treated as a build checklist, not a scoreboard. The accounts that obsess over hitting Excellent often do it by stuffing in generic headlines that drag down the message — and their conversion rate quietly suffers. The accounts that win write a dozen genuinely distinct, on-brand headlines across clear categories, land at Good or Excellent naturally, and then judge the ad by conversions. Let Ad Strength guide the build; let the conversion data settle the argument.
Think of Ad Strength as ensuring you have not under-built the ad. Once it says Good or Excellent and your copy is genuinely strong, stop optimizing the meter and start optimizing for conversions through asset reporting.
100+ headline templates by category
Below is a copy-paste library of headline templates organized by category. Replace the bracketed placeholders with your specifics, keep within 30 characters, and ensure each can grammatically combine with others. Use several categories per ad group so the algorithm has diverse angles to test.
Benefit-led (lead with the outcome):
- '[Outcome] in [Timeframe]'
- 'Get [Benefit] Without [Pain]'
- 'Cut [Cost/Time] by [X]%'
- 'Achieve [Goal] Faster'
- 'More [Desirable Result], Less [Hassle]'
- 'The Easy Way to [Outcome]'
- 'Stop [Pain Point] for Good'
- 'Double Your [Metric]'
- 'Save Hours on [Task]'
- '[Benefit], Guaranteed'
- 'Finally, [Desired Outcome]'
- 'Make [Hard Thing] Simple'
- 'Boost [Metric] Fast'
- 'Reach [Goal] Sooner'
- 'Less [Pain], More [Gain]'
Feature-led (lead with what it does):
- 'Built-In [Feature]'
- 'All-in-One [Category] Tool'
- 'Works With [Integration]'
- 'No-Code [Capability]'
- 'Real-Time [Feature]'
- 'Automated [Process]'
- 'Powered by [Technology]'
- 'Includes [Notable Feature]'
- 'Connect [Tool] in Minutes'
- '[Number] Features in One'
- 'Drag-and-Drop [Capability]'
- 'Enterprise-Grade [Feature]'
- 'Customizable [Element]'
- 'Sync Across [Platforms]'
- 'Native [Integration] Support'
Social proof (lead with credibility):
- 'Trusted by [Number]+ [Audience]'
- 'Rated [X]/5 by [Source]'
- 'Join [Number]+ Customers'
- '[Notable Customer] Uses Us'
- 'Loved by [Audience]'
- 'As Seen in [Publication]'
- '[Number]+ 5-Star Reviews'
- 'The [Category] Leader'
- 'Why [Audience] Choose Us'
- 'Award-Winning [Category]'
- 'Recommended by [Authority]'
- '[Number] Teams Trust Us'
- 'Top-Rated [Product]'
- 'Backed by [Credential]'
- 'Industry-Leading [Metric]'
Offer and promotion (lead with the deal):
- 'Save [X]% Today'
- 'Free [Trial/Demo/Sample]'
- '[X] Months Free'
- 'Start Free, Upgrade Later'
- 'No Setup Fees'
- 'Free Shipping on [Threshold]'
- '[X]% Off [Product]'
- 'Try Free for [Days] Days'
- 'Get [Bonus] Free'
- 'New Customer Discount'
- 'Limited-Time [Offer]'
- 'Money-Back Guarantee'
- 'Free Consultation'
- 'Bundle and Save [X]%'
- 'No Credit Card Required'
Urgency and scarcity (lead with timing):
- 'Ends [Day/Date]'
- 'Only [X] Left'
- 'Last Chance for [Offer]'
- 'Sale Ends Soon'
- 'Limited Spots Available'
- 'Offer Expires [Timeframe]'
- 'Today Only'
- 'Don't Miss [Event/Offer]'
- 'While Supplies Last'
- 'Book Before [Deadline]'
- 'Final Days of [Sale]'
- 'Hurry, [Offer] Ends'
Question (engage the searcher):
- 'Need [Solution] Fast?'
- 'Struggling With [Problem]?'
- 'Looking for [Category]?'
- 'Ready to [Goal]?'
- 'Tired of [Pain Point]?'
- 'Want [Outcome]?'
- 'Why Settle for Less?'
- 'Is [Old Way] Costing You?'
- 'Searching for [Solution]?'
- 'What If [Outcome]?'
- 'Still Using [Old Method]?'
- 'Curious About [Product]?'
Brand and trust (lead with who you are):
- '[Brand] — Official Site'
- '[Brand]: [Category] Done Right'
- 'The Original [Product]'
- '[Years]+ Years of [Expertise]'
- 'Made in [Place]'
- 'Secure and [Compliance]'
- '[Brand] You Can Trust'
- 'Certified [Standard]'
- 'Your [Category] Partner'
- 'Built for [Audience]'
- '[Brand] Quality, [Benefit]'
- 'Trusted [Category] Since [Year]'
That is over 100 templates across seven categories. The craft is not using all of them — it is selecting eight to fifteen per ad group, customizing each to your specific value proposition and keyword theme, and spanning multiple categories so the algorithm tests genuinely different messages.
Description frameworks that convert
Descriptions (up to four, 90 characters each, two shown per served ad) expand on the headlines and carry the persuasive heavy lifting. Write all four using distinct frameworks so each stands alone and the pair Google shows always works.
Framework 1 — Benefit + Proof: lead with the primary outcome, then substantiate it. Pattern: '[Achieve outcome] with [product]. [Proof point or credibility marker].' Example: 'Cut reporting time in half with automated dashboards. Trusted by 5,000+ teams.'
Framework 2 — Offer + Call to Action: lead with the deal, end with a clear next step. Pattern: '[Offer]. [Benefit]. [Action verb] today.' Example: 'Start free for 14 days. No card required. See results before you pay. Sign up now.'
Framework 3 — Feature + Differentiation: lead with a capability, then explain why it beats the alternative. Pattern: '[Key feature] that [differentiator]. [Reason to choose].' Example: 'All-in-one platform that replaces five tools. One login, one bill, zero busywork.'
Framework 4 — Trust + Credibility: lead with reassurance for risk-aware buyers. Pattern: '[Credibility marker]. [Risk reducer]. [Supporting detail].' Example: 'Enterprise-grade security and 99.9% uptime. Money-back guarantee. Set up in minutes.'
Description writing rules:
- Each description must stand alone, because Google may show any two in any order.
- Do not duplicate your headlines — descriptions should reinforce and expand, not repeat.
- Include the keyword theme where it reads naturally, supporting relevance.
- Lead with value, end with direction — the strongest descriptions front-load the benefit and finish with a next step.
- Use the full 90 characters when you can do so without padding; longer descriptions occupy more space and convey more.
The four-framework approach guarantees variety: whichever two descriptions Google pairs, the searcher sees a benefit, an offer, a differentiator, or a trust signal — never two of the same. This is the description-level equivalent of spanning headline categories, and it lifts both Ad Strength and real performance.
Asset reporting: reading what Google tells you
Once an RSA has accumulated data, Google provides asset-level reporting that tells you which headlines and descriptions are working. Reading it correctly is how you optimize without guessing.
The performance ratings: each asset is rated Low, Good, or Best (or Pending/Learning while data accumulates). These are relative ratings within the ad — Best assets contribute most to performance, Low assets least.
How to use the ratings:
- Replace Low-rated assets with new variations, ideally on the angles that produced Best-rated assets. Do not just delete Low assets and shrink the pool — swap in fresh attempts so the testable space stays rich.
- Identify winning angles by noting which categories your Best assets come from. If social-proof and offer headlines dominate, write more in those veins.
- Do not over-react to a single Low rating early on — ratings need accumulated data, and an asset can be Low simply because it has been served less.
The combinations report shows which assembled combinations Google served most and how they performed. This reveals the messages and pairings that resonate, which informs both your RSA optimization and your broader messaging.
Optimization discipline:
- Change a few assets at a time so you can attribute the effect of each change. Rewriting an entire ad confounds everything and resets learning.
- Give changes time — after swapping assets, allow the ad to re-accumulate data before judging the new versions.
- Run on a cadence — monthly review for most accounts, more frequently only for very high-volume ones. Constant tinkering prevents stable winners from emerging.
A note on data thresholds: low-volume ad groups may never generate enough impressions for confident asset ratings. In those cases, lean more on first-principles copywriting (the categories and frameworks above) and account-level patterns rather than waiting for asset-level statistical signals that will not arrive. Reserve heavy data-driven RSA optimization for your higher-volume ad groups where the reporting is meaningful.
AI-assisted headline generation done right
RSAs reward volume and variety of high-quality assets — exactly what AI is good at producing quickly. Used well, AI is a drafting accelerator that lets you populate the template categories fast and maintain a backlog of fresh angles. Used badly, it floods your ads with generic, off-brand, or non-compliant copy.
Where AI helps:
- Generating candidate headlines at volume — feed it your value proposition, keyword theme, and the category templates above, and it will produce dozens of variations in seconds.
- Breaking writer's block — when you have eight good headlines and need five more distinct angles, AI surfaces options you might not have considered.
- Adapting for seasonality and offers — quickly spin up urgency and promotional variants for a sale period.
- Maintaining an idea backlog — keep a running list of AI-generated candidates to test in your monthly optimization cadence.
The non-negotiable editing pass. AI output requires human review before it goes live, for four reasons:
- Generic phrasing — AI skews toward safe, bland headlines that could describe any advertiser. Edit for specificity and your distinct value.
- Unsupported claims — AI will happily generate 'the best' or 'number one' or invented statistics. Remove anything you cannot substantiate; unverifiable superlatives also risk disapproval.
- Brand voice — AI does not know your tone. Edit so the copy sounds like your brand, not like generic ad copy.
- Character limits and grammar — verify the 30-character headline and 90-character description limits and confirm assets combine grammatically.
A practical workflow: use AI (Google's own asset suggestions plus an external LLM) to generate 30-50 candidate headlines across the seven categories. Then human-edit down to the 12-15 strongest, customized and substantiated, before launching. Our guides on AI for RSA testing and rotation and ChatGPT prompts for Google Ads provide ready prompt structures for this.
The principle: AI drafts, humans decide. Never paste AI headlines live unreviewed. The combination of AI's volume and a human's judgment about accuracy, brand, and compliance is what produces a deep pool of genuinely good assets — which is exactly what the RSA format needs.
30-day RSA build and optimization plan
The HowTo schema above is the day-by-day. Here is the strategic framing across the four weeks.
Week 1 — Foundation and raw material. Audit ad group theming (RSA quality depends on it), gather the research that fuels good headlines (keyword themes, converting search terms, benefits, proof points, offers, competitor angles), and draft headlines across all seven categories using the template library. By the end of week 1 you should have a deep, varied asset pool per ad group.
Week 2 — Build and launch. Write the four framework-based descriptions, set pinning only where genuinely required, reach Good or Excellent Ad Strength without filler, add a full set of extension assets, and launch two RSAs per ad group where volume allows. By the end of week 2 your ads are live and complete.
Week 3 — Let it learn. Resist editing. Monitor only for disapprovals and serving issues. Let the algorithm accumulate the data it needs to rate assets and find winning combinations. Premature edits here are the most common way teams sabotage their own RSAs.
Week 4 — Read and refine. Open asset reporting, identify Low performers and winning angles, replace Low-rated assets with new variations on what works, and codify a monthly optimization cadence. Build an AI-assisted backlog of headline ideas so you always have fresh angles to test.
Beyond day 30, RSA optimization is a steady-state practice: monthly asset review, periodic refreshes for seasonality and new offers, and continuous testing of new angles against your winners. The copywriting craft compounds — the more you learn about which messages resonate in your category, the better every subsequent ad becomes. Pair strong RSAs with strong landing pages (see our landing pages for conversion guide) and a healthy Quality Score (see our Quality Score guide) and the search ad becomes a genuine growth engine rather than a checkbox.
If you would like AI-driven optimization that watches your RSA asset performance and surfaces the changes worth making — so your team writes the copy and the system handles the monitoring — SteerAds runs a free 14-day audit on your Google and Microsoft Ads accounts.
Sources
Official and third-party sources consulted for this guide:
- support.google.com/google-ads — Responsive Search Ads documentation
- support.google.com/google-ads — Ad Strength and asset best practices
- optmyzr.com — Optmyzr RSA and ad strength studies
- searchengineland.com — Search Engine Land RSA coverage and analysis
- support.google.com/google-ads — asset reporting and ad performance ratings
FAQ
How many headlines and descriptions should I write for a Responsive Search Ad?
Provide as many distinct, high-quality assets as you genuinely have: up to 15 headlines and 4 descriptions per RSA. Aim for at least 8-10 strong, varied headlines and all 4 descriptions to give Google's machine learning enough combinations to optimize, which also lifts ad strength. Quality beats quantity — 10 distinct, on-message headlines outperform 15 where five are filler or near-duplicates. Vary the angle across assets: benefit-led, feature, social proof, offer, urgency, question, and brand. Repetitive headlines that say the same thing differently waste asset slots and give the algorithm nothing to test.
Should I pin headlines in Responsive Search Ads?
Pin sparingly and deliberately. Pinning forces an asset into a fixed position, which reduces the combinations Google can test and almost always lowers ad strength. Pin only when something must always appear or always sit first — a legal disclaimer, a required brand name, a specific compliance phrase, or a non-negotiable offer. The recommended pattern when you must pin is to pin two or three assets to the same position so Google still has choices within that slot. For most campaigns, leaving headlines unpinned and letting the algorithm optimize produces better performance than tightly controlling every position.
Does Ad Strength actually affect performance?
Ad Strength is a directional indicator of asset quality and diversity, not a direct ranking factor, but it correlates with results: Google's own data associates improving Ad Strength from Poor to Excellent with meaningfully more conversions on average. Treat it as a useful checklist — it nudges you toward more headlines, more variety, more keyword relevance, and less duplication, all of which genuinely help. Do not chase Excellent at the expense of message quality, though. An ad with Good strength and tight, on-brand copy can outperform an Excellent ad stuffed with generic headlines added only to satisfy the meter. Use Ad Strength as a guide, judge by conversions.
How do I write RSA headlines that improve Quality Score?
Ad relevance is one of the three Quality Score components, and RSA headlines are where you earn it. The highest-leverage tactic is mirroring the searcher's intent and including the keyword theme of the ad group in several headlines — which is why tightly themed ad groups matter so much. Write headlines that match what the query is asking, address the specific need, and lead with the benefit. Pair relevant headlines with a landing page that continues the same message. Generic headlines that could apply to any advertiser hurt relevance; specific, intent-matched headlines that echo the query lift it. See our Quality Score guide for the full picture.
Can I use AI to write Responsive Search Ad headlines?
Yes, and it is one of the best uses of AI in paid search — but as a drafting accelerator, not an autopilot. AI tools (including Google's own asset suggestions and external LLMs) generate large volumes of headline variations quickly, which is exactly what RSAs need. The discipline is editing: AI output skews generic, occasionally makes unsupported claims, and does not know your brand voice or compliance constraints. Use AI to generate 30-50 candidates across categories, then human-edit for accuracy, brand fit, character limits, and policy. Never paste AI headlines live unreviewed — unverifiable claims and off-brand phrasing are common and can cause disapprovals.
What is the difference between RSA assets and assets like sitelinks?
RSA assets are the headlines and descriptions inside a single responsive search ad — the components Google mixes and matches to assemble the ad. Assets like sitelinks, callouts, structured snippets, images, and prices are account- or campaign-level extensions that appear alongside the ad to add information and take up more space. Both are now called assets in Google's 2026 terminology, which causes confusion. For a complete ad, you need strong RSA headlines and descriptions and a full set of extension assets. They work together: RSA assets make the core ad, extension assets make it bigger and richer in the results.
How long should I wait before judging or editing an RSA?
Give a new RSA at least two to four weeks and a few thousand impressions before drawing conclusions, longer for low-volume accounts. Asset performance ratings (Low, Good, Best) need accumulated data to be meaningful, and editing an ad resets its learning. When you do optimize, change one thing at a time: replace Low-rated assets with new variations rather than rewriting the whole ad, so you can attribute changes. Avoid the temptation to constantly tinker — frequent edits prevent the algorithm from ever finding stable winning combinations. Set a cadence (monthly is reasonable for most accounts) and make deliberate, measured changes.