Google Ads Responsive Search Ads: Best Practices Guide (2026)

Google Ads Responsive Search Ads (RSAs) let you supply up to 15 headlines and 4 descriptions that Google's machine learning mixes and matches at auction time to show the most relevant combination to each searcher. The best practices for RSAs are to write all 15 distinct headlines covering benefit, feature, brand, social proof, and call-to-action angles; fill all 4 descriptions; pin sparingly and only for compliance reasons; rotate underperforming assets every 2-4 weeks; and treat Ad Strength as a directional signal, not the goal.

Since Google retired expanded text ads in 2022, RSAs have been the default search ad format in every Google Ads account. But "responsive" does not mean "hands-off." This guide covers the tactical best practices, the mechanics underneath the ad format, and the common mistakes that waste budget. For the broader bidding strategy layer that pairs with RSA creative, see our Google Ads bidding strategies guide.

TL;DR: RSA Best Practices at a Glance

• Write 15 distinct headlines: Every headline slot should carry a unique message. Avoid near-duplicates.
• Fill all 4 descriptions: Empty description slots leave impressions on the table.
• Pin only for compliance: Pinning more than 1-2 headlines per position cripples Google's optimization engine.
• Check asset performance weekly: The asset performance table tells you exactly which headlines and descriptions need replacing.
• Iterate every 2-4 weeks: Rotate out "Low" performers, keep "Best" and "Good" assets, add fresh variants.
• Ad Strength is directional, not the target: "Excellent" does not guarantee conversions. Relevance and conversion volume matter more. Our deep dive on what Ad Strength actually measures explains why it is a diagnostic, not a performance metric.

What Are Responsive Search Ads (Rsas)?

Responsive Search Ads are Google's machine-learning-powered search ad format. Instead of writing one static ad, you provide a pool of assets - up to 15 headlines and 4 descriptions. At each auction, Google's system selects and assembles a combination predicted to perform best for that specific query, device, and user signal. The ad can display up to 3 headlines and 2 descriptions in any order, which means Google tests thousands of permutations without you creating each one manually.

RSAs replaced expanded text ads (ETAs) as the default format in June 2022. Unlike ETAs, where you pinned headlines to specific positions and controlled every combination, RSAs shift the optimization burden to Google's algorithms. Your job shifts from writing perfect fixed ads to supplying a rich, diverse asset pool and monitoring performance signals.

How Rsas Differ from Expanded Text Ads

ETAs let you write exactly 3 headlines and 2 descriptions in fixed positions. You controlled the sequence and knew which combinations would serve. RSAs break that mold:

RSA vs Performance Max: Two Different Machines

This is a common point of confusion. RSAs operate inside Search campaigns on the Google Search Network only. Performance Max (PMax) is a cross-network campaign type that serves ads across Search, Shopping, Display, YouTube, Discovery, Gmail, and Maps using a single asset group. RSAs compete in keyword-triggered auctions. PMax competes wherever Google's algorithms find a conversion opportunity, often without keyword targeting at all.

A Search campaign with RSAs gives you keyword-level control and search-term visibility. PMax gives Google broader latitude across channels but offers far less transparency into where spend goes. Most advertisers run both - RSAs for high-intent search coverage, PMax for demand generation and remarketing across surfaces. For guidance on PMax budget and bidding, see our Performance Max budget and bidding guide.

RSA Asset Best Practices: How to Write 15 Headlines and 4 Descriptions

The single most important variable in RSA performance is asset diversity. If your 15 headlines all say roughly the same thing, Google has nothing to optimize - the machine-learning layer becomes a random shuffle. Structure your headlines across five categories:

• Benefit headlines (3-4): Lead with the outcome the customer gets. "Save 10 Hours a Week on Reporting," "Ship Orders 40% Faster."
• Feature headlines (3-4): Describe what the product does. "Real-Time Inventory Sync Across Channels," "One-Click Integration With Shopify."
• Brand headlines (2-3): Use your company name or tagline. "[Brand] | Trusted by 5,000+ Teams," "Official [Brand] Partner."
• Social-proof headlines (2-3): Cite reviews, ratings, or recognition. "Rated 4.8 Stars on G2," "Gartner Magic Quadrant Leader 2025."
• CTA headlines (2-3): Direct action prompts. "Start Your Free Trial," "Get a Custom Demo Today," "See Plans and Pricing."

Descriptions should extend the headline message with supporting detail. Use all 4 slots. Each description should work alongside any headline combination, so avoid call-and-response pairs that only make sense together (e.g., "Want faster shipping?" as a headline paired with "We deliver in 2 days" as a description - those must always co-occur to parse, which Google cannot guarantee).

For keyword-level strategy that feeds into the ad copy itself, reference our keyword match types guide.

Pinning Headlines and Descriptions: When and Why

Pinning forces a specific headline or description to appear in a designated position (e.g., Headline 1 always shows your brand name). Google's RSA engine optimizes best when assets are unpinned and free to recombine. Pinning should be used minimally and for two reasons only:

• Compliance: Regulatory requirements demand specific language in a fixed position (financial disclaimers, legal notices, trademark language).
• Brand consistency: A single brand-identifier headline pinned to position 1 or 3 is reasonable if brand presence in every impression is a hard requirement.

Over-pinning is the most common RSA mistake. Pinning 3+ headlines collapses the combinatorial space and reduces the ad to a glorified ETA - but without the control ETAs gave you. Google's system cannot optimize when it has few degrees of freedom. If you pin more than 2 assets total, you are likely leaving performance on the table.

Ad Strength: What "Excellent" Actually Means

Ad Strength is Google's internal rating (Poor, Average, Good, Excellent) that estimates how well your RSA asset pool covers diverse searcher intents. It is a coverage score, not a performance promise. An "Excellent" rating means you have enough asset variety for the system to test combinations. It does not mean the ad copy is compelling, on-brand, or conversion-optimized.

Chasing "Excellent" at the expense of message quality is a trap. A "Good" RSA with sharp, differentiated headlines will often outperform an "Excellent" RSA padded with generic filler. Treat Ad Strength as a diversity check - fill the slots, vary the angles, and then focus on asset-level performance data to decide what works.

How RSA Asset Selection Works (Auction-Time Signals)

At each auction, Google's model evaluates the available assets against real-time signals: the search query, the user's device type, location, time of day, audience list membership, and recent browsing behavior. It predicts which headline-description combination is most likely to generate a click or conversion for that specific impression and assembles the ad on the fly. This is why asset diversity matters - the model can only choose from what you give it. A narrow asset pool gives the model few options regardless of how good the signals are.

RSA Reporting: The Asset Performance Table

The asset performance table (found in the Ads & Assets section of Google Ads) is the primary feedback loop for RSAs. Each headline and description receives one of four ratings:

• Best: Performing significantly above the ad group average. Keep and protect these.
• Good: Performing at or near average. Retain but monitor.
• Low: Underperforming. These are replacement candidates in the next iteration cycle.
• Pending/Learning: Not enough data yet. Give these 2-4 weeks before judging.

The table also shows impression counts per asset, so you can see which headlines are actually getting served. If an asset has zero impressions after a month, it may be pinned in a way that excludes it, or it may be so far from what the model selects that it never surfaces.

Iteration Cadence: How Often to Rotate Assets

A steady, data-informed rotation produces better results than either set-and-forget or constant tweaking. A practical rhythm:

• Weekly check: Review the asset performance table for any "Low" assets that have crossed 2,000+ impressions.
• 2-4 week rotation: Replace 1-3 "Low" performers with fresh variants. Keep "Best" and "Good" assets in place.
• Add, don't just replace: If you have fewer than 15 headlines or 4 descriptions active, fill empty slots before removing anything.

Don't rotate so aggressively that assets never accumulate enough data to receive a rating. A headline that has served only 200 impressions may look "Low" due to noise, not actual underperformance.

A/B Testing Limitations for Rsas

RSAs do not support native A/B testing where you split traffic between Ad A and Ad B and declare a winner. The ad itself is a combinatorial machine - it is already testing permutations internally. To compare creative strategies, use Google Ads ad variations (which let you test a specific change - e.g., a new headline set - across an ad group) or campaign experiments (which duplicate a campaign and split traffic at the campaign level). Both approaches add rigor but require enough volume to reach statistical significance. For a deeper look at automation strategy across your Google Ads account, see our automation guide.

Common RSA Mistakes That Waste Budget

• One headline duplicated 15 times: The system has no variation to test. This is the same as serving a single static ad with extra steps.
• Pinning everything: Pinning all headlines to fixed positions destroys the responsive mechanism and leaves Google with no optimization room.
• Ignoring asset reporting: Without reviewing the performance table, "Low" assets stay live for months, dragging down the ad's average performance.
• Empty description slots: Leaving description lines unfilled reduces the number of possible ad assemblies and misses the chance for longer, more persuasive copy in the extra headline space that sometimes extends into description territory.
• One ad group, one RSA: Google recommends at least one RSA per ad group but a single RSA per ad group with a narrow asset pool often underperforms. In high-volume ad groups, testing multiple RSAs with different creative angles can surface what resonates.

To swap dynamic text inside these ads by user or date, our Google Ads ad customizers guide explains the feed and function setup.

To make match types work inside your ad copy automatically, read our dynamic keyword insertion guide.

Frequently Asked Questions

How Many Headlines Should I Write for a Responsive Search Ad?

Write all 15. Each headline should be distinct and fall into one of five categories: benefit, feature, brand, social proof, or call to action. A full, diverse asset pool gives Google's machine learning the most combinations to test and maximizes your chance of finding high-performing permutations.

When Should I Pin Headlines in an RSA?

Pin only when compliance requires specific language in a fixed position (e.g., legal disclaimers) or when brand-identifier presence in every impression is non-negotiable. Over-pinning removes the degrees of freedom Google's optimization engine needs and turns the RSA into a static ad.

Does "Excellent" Ad Strength Mean My RSA Will Perform Well?

Not necessarily. Ad Strength measures asset diversity and coverage, not copy quality or conversion potential. A "Good" RSA with sharp, differentiated messaging often outperforms an "Excellent" RSA padded with generic filler. Use Ad Strength as a coverage check, not a performance target.

How Often Should I Change RSA Assets?

Check the asset performance table weekly. Rotate out "Low" performers every 2-4 weeks and replace them with fresh variants. Keep "Best" and "Good" assets. Don't rotate so fast that assets never accumulate enough impressions to get rated.

Can I a/B Test Responsive Search Ads?

RSAs do not support native A/B testing between two fixed ads. Use Google Ads ad variations to test a specific creative change across an ad group, or campaign experiments to split traffic at the campaign level. Both require enough impression volume to reach statistical significance.

Key Takeaways

• RSAs are machine-learning-powered search ads that combine up to 15 headlines and 4 descriptions at auction time.
• Write all 15 headlines across five categories - benefit, feature, brand, social proof, and CTA - for maximum combinatorial coverage.
• Pin sparingly. Over-pinning cripples Google's optimization and wastes the RSA advantage.
• Use the asset performance table as your primary optimization signal; replace "Low" performers every 2-4 weeks.
• Ad Strength is a diversity check, not a performance guarantee. Good copy beats full slots.
• RSAs are Search-only; Performance Max is cross-network. Run both with different strategies.