Flight log / Transmission 012

Responsive search ads best practices for 2026

You write 15 headlines and 4 descriptions. Google shows at most 3 and 2 of them, in whatever order it likes. The craft is now in the deck, not the hand.

RSAs flipped the job of a search copywriter. You no longer write an ad. You write ingredients, and Google decides which ones end up on the plate for each query. That makes the inputs, the guardrails and the way you read results matter more than any single headline.

This guide covers the current specs, when pinning is worth its cost, what Ad Strength does and does not do, how to read asset reporting now that the old labels are gone, and how AI Max text customization changes the picture.

What are the responsive search ad specs in 2026?

Google also publishes its own lift figures for more ads: advertisers with one RSA who add a second see a 6.6% increase in conversions, and going from 2 to 3 RSAs adds an average of 3.7%, according to Google Ads Help. Those are Google's averages across advertisers, not a promise for your ad group, but the direction is consistent: one RSA per ad group leaves room on the table.

How do responsive search ads actually work?

You supply up to 15 headlines and 4 descriptions. At auction time Google assembles a combination it predicts will perform best for that search, device and context. Any headline can appear next to any other headline, in any order, unless you pin it.

That last part is the whole game. Every asset has to make sense on its own and next to every other asset. A headline that only works as the second half of a sentence will eventually show up as the first half of a different one.

How should you write RSA headlines?

Google's own guidance is to include at least one of your keywords in your headlines and to make as many headlines as possible unique. The practical version: give each headline a job. Here's a role map for a hypothetical SOC 2 compliance tool, with every example under 30 characters.

Aim for two or three headlines per role across your 15, then check that no two say the same thing in different words. Near-duplicates don't add combinations; they just dilute the data.

Descriptions follow the same rule with more room. Each of the 4 should stand alone: one leads with the outcome, one with proof, one with the offer, one with the qualifier. Description 1 is the one most likely to show, so keep it the strongest standalone line.

Should you pin headlines in responsive search ads?

Pinning locks a headline or description to a position. Google is clear about the tradeoff: pinning one asset prevents others from showing in that position, which reduces the number of combinations and can lower Ad Strength. Google's help page notes that Headline 1 and Description 1 content usually appears when the ad serves, while Headline 3 and Description 2 are not guaranteed.

The middle path is Google's own suggestion: pin 2 or 3 unique headlines to the same position so the system can still rotate between them. In most B2B accounts, that is where pinning earns its keep: a topic headline locked to position 1 with two or three variants, everything else free.

One new wrinkle. If AI Max for Search is on and final URL expansion picks a different landing page, Google says pinned RSA assets aren't used for that impression. Pins are guardrails, not guarantees, once the system is allowed to choose the page.

What does Ad Strength actually measure?

Ad Strength is a pre-launch feedback rating: Incomplete, Poor, Average, Good or Excellent. Google says it looks at whether your assets include relevant keywords, how much variety your headlines and descriptions give the system to combine, and whether you have enough sitelinks, with 6 or more improving the rating.

What it isn't matters more. Google states that Ad Strength is not used to calculate Ad Rank, Quality Score or auction wins and doesn't directly affect serving eligibility. It's a checklist for asset diversity, not a performance score.

Google does publish a correlation: advertisers who improve Ad Strength for their RSAs and sitelinks from Poor to Excellent see 15% more conversions on average. That's a cross-advertiser average from Google, and part of it is simply that Excellent ads tend to have more, better assets. Use the rating as a floor, not a target.

How do you read RSA asset reporting now?

The old performance labels (Best, Good, Low) are gone from the RSA asset report. Google deprecated the column and replaced it with full performance statistics per asset, available for dates on or after June 5, 2025: impressions, clicks, conversions, conversion value and cost. There's also a combinations report showing which assets were shown together.

Google attaches two caveats worth taking seriously:

• Ratio metrics are directional. Google calls asset-level CTR and CPA directional indicators only, because every impression includes several assets at once.
• Totals won't reconcile. Asset impressions, clicks and costs won't sum to ad-level numbers, since each asset in an impression logs its own metrics.

So read the report for preference, not proof. Impression share between assets tells you what the system favors. Conversions per asset tell you which messages show up in converting combinations. Neither tells you an asset caused anything on its own.

A workable routine: once an ad group has meaningful volume, replace the 2 or 3 headlines with the fewest impressions, keep their role, change the angle, and log what you swapped and when. Over a quarter that becomes a real mission log of what your buyers respond to.

How should you test responsive search ads?

Ad variations are the most underused option. Google supports them for RSAs and lets you apply a single change across an account, a campaign or a custom scope, set the share of traffic that sees the modified ads, and pick an end date. Google recommends variations for a single broad change and custom experiments for several simultaneous changes.

Whatever the method, judge on conversions per impression or cost per qualified lead, not CTR. A headline that promises more than the page delivers can win on CTR and lose on pipeline. Which is why the most important RSA test is often not about the ad at all but about whether the click lands somewhere that continues the same message. Our message match guide covers that side.

How does AI Max text customization affect RSAs?

Text customization is the AI Max feature that writes extra headlines and descriptions for your RSAs. It's the successor to what Google used to call automatically created assets. Google's documentation says it draws on your landing page content, your existing ads and your keywords, combining snippets pulled from the page with generative AI.

• It sits outside your limits. Generated assets don't count toward the 15 headlines and 4 descriptions.
• It competes with your copy. Google says customized assets only serve when predicted to beat the assets you wrote.
• It refreshes. Assets are reviewed and refreshed as needed at least every 48 hours.
• You can see it. The "Added by" column and asset reports separate generated assets from yours, and you can remove ones you don't want.
• Controls are partial. Text guidelines, which let you exclude terms and set tone, are still described as a feature in testing on Google's text customization page. Check whether your account has them before you count on them.

Timing matters. Google's announcement of the AI Max upgrade said automatic upgrades begin in September 2026, and campaigns that used automatically created assets upgrade with text customization enabled by default. Dynamic Search Ads got a reprieve: Google later pushed the DSA auto-upgrade to February 2027. Automatically created assets did not, and stayed on the September 2026 schedule. If you haven't looked at the "Added by" column this month, you may already be running generated copy.

The implication is uncomfortable and useful: the page your ad points at now shapes the ad itself. A specific, query-relevant page gives the system specific material. A generic homepage gives it generic material. That also feeds back into ad relevance and landing page experience, which we break down in how to improve Quality Score without touching your bids.

Where this breaks

• Low-volume B2B ad groups. Many niche B2B keywords never generate enough impressions for asset data to mean much. Consolidate before you test.
• Unknown weighting. Google doesn't publish how Ad Strength weights its inputs or how often generated assets displace yours.
• Attribution blur. Asset conversions are shared across every asset in the combination, so no single headline gets clean credit.
• Moving target. AI Max defaults are still rolling out. Check settings per campaign rather than assuming last quarter's configuration.

RSA best practices checklist

1. At least 2 RSAs per ad group, each with 10 to 15 distinct headlines and 4 descriptions.
2. Every headline has a role: topic, outcome, qualifier, proof, objection, offer.
3. At least one keyword-led headline, but not 15 of them.
4. Pin only when wording is mandatory, and pin 2 or 3 variants to a slot.
5. Ad Strength at Good or better, treated as a floor.
6. 6 or more sitelinks available at account, campaign or ad group level.
7. Asset report reviewed monthly; low-impression assets replaced and logged.
8. AI Max text customization status checked, and the "Added by" column reviewed.
9. Each ad group sends traffic to the most relevant page you have.

For how RSAs fit into the rest of a B2B search program, see the B2B PPC playbook.

FAQ