Ad Copy Testing: How to Write Google Ads That Convert

Why Ad Copy Still Wins or Loses Campaigns

You can have the best keywords, the smartest bidding strategy, and a beautifully optimized landing page — and an account full of mediocre ad copy will still bleed money. Copy is what stands between your bid and the user's eyeball. If it doesn't earn the click, none of the rest matters.

The Compounding Effect

A 20% lift in CTR doesn't just mean 20% more clicks. It means a higher Quality Score, which means a lower CPC, which means you can afford to bid more aggressively, which means higher position, which means even more clicks. Small CTR improvements compound into outsized account-wide gains.

Then there's the downstream effect. Better copy attracts better-qualified clickers. Better-qualified clickers convert at higher rates. Higher conversion rates feed Smart Bidding with cleaner signal, which improves bid quality further. The compounding effect from a single great ad can ripple through the entire account for months.

Why Most Ads Are Forgettable

Most search ads I audit could belong to any competitor in the category. "Best [Product] Online — Free Shipping — Shop Today" describes 10,000 brands. The user's eye glazes over. The click goes to whichever ad happens to be in position one or has the most assets. None of that is copy doing work — that's auction mechanics.

Creative as Leverage

According to Think with Google's creative effectiveness research, creative quality (not bid, not budget, not targeting) is the largest single driver of ad effectiveness in mature accounts. The implication: the time you spend writing better copy almost always returns more than the same time spent fiddling with bid adjustments. Pair this with the bidding mechanics in our bidding strategies guide and you have the two highest-leverage levers in PPC.

Anatomy of a Responsive Search Ad

Responsive Search Ads are now the only standard search ad format Google offers — Expanded Text Ads were sunset in 2022. RSAs work fundamentally differently from their predecessors: instead of writing a single ad, you provide a pool of headlines and descriptions and let Google's machine learning assemble the highest-performing combination per auction.

The Components

• Up to 15 headlines, each 30 characters maximum
• Up to 4 descriptions, each 90 characters maximum
• 2 path fields for the display URL, each 15 characters
• 1 final URL

At any given auction, Google selects up to 3 headlines and 2 descriptions to display, drawn from your asset pool. Different users may see different combinations based on what the algorithm predicts will perform best for their context.

Why More Variety Matters

The algorithm needs combinatorial variety to find winning combinations. An RSA with 4 headlines gives Google about 24 possible combinations of three headlines. An RSA with 12 headlines gives Google about 1,320 combinations. The 12-headline ad gives the algorithm 55x more combinations to test — and consistently outperforms.

The Ad Strength Indicator

Google rates each RSA as Poor, Average, Good, or Excellent. Aim for Excellent on every ad. Per Google's published data, ads rated Excellent see significantly higher CTR than ads rated Average. The criteria are unique headlines, varied messaging, included keywords, and avoidance of duplicates.

Headline Formulas That Earn the Click

The 30-character headline is the most expensive real estate in PPC. Three of them stack on top of your ad and they're the first thing the user reads. Every character has to earn its place.

Formula 1: Specific Benefit + Number

Numbers anchor attention because they break the visual pattern of pure text. "Cut Your Heating Bill 40%" outperforms "Lower Your Heating Bill" almost every time. "Save 4 Hours a Week" outperforms "Save Time."

Formula 2: Audience Naming

"For Marketing Teams." "For Solo Founders." "For Construction Crews." Naming your audience filters out wrong-fit clickers and signals strong relevance to right-fit ones. Counterintuitively, audience-named ads usually have lower CTR but dramatically higher conversion rate — and CTR alone is a vanity metric.

Formula 3: Question Headlines

"Tired of Slow Customer Support?" "Looking for Same-Day Delivery?" Questions invoke an automatic mental answer, which engages the user even before they read the rest of the ad. Use sparingly — overuse becomes gimmicky.

Formula 4: Social Proof

"Trusted by 12,000 Teams." "Rated 4.9 Stars by 8,000 Customers." "As Used By Toyota and Ford." Concrete numbers and named brands carry more weight than generic claims. "The Best Service" means nothing; "Rated #1 by Wirecutter" means something.

Formula 5: Urgency or Scarcity

"Sale Ends Friday." "Only 24 Spots Left." "Free Setup Today Only." Use urgency carefully and only when it's true. Fake urgency erodes trust fast and may violate Google's policies.

Formula 6: Direct Offer

"Try Free for 30 Days." "Get a Quote in 60 Seconds." "$199 Whole-Home Install." Direct offers work because they tell the user exactly what they get if they click. No mystery, no hesitation.

Building Your Headline Library

Mix formulas across your 15 headlines. Don't write 12 benefit-focused headlines — vary the angles so Google has different lever types to test. A typical strong RSA has 3 benefit headlines, 3 audience-named, 2 social proof, 2 offer-based, 2 question, 1 brand-focused, and 2 keyword-mirroring headlines.

Description Lines: The Underused Lever

Headlines get all the attention. Descriptions are an afterthought for most advertisers. That's a mistake — descriptions are where you handle objections, add proof, and seal the click.

Description Length and Pacing

Each description gets 90 characters. Use them. Short descriptions look thin and underperform longer ones in head-to-head tests. Aim to fill at least 80 of the 90 available characters per description.

What Goes in a Description

Good descriptions usually combine three elements: a benefit, a proof point, and a call to action. For example: "Custom roof inspections from licensed pros. 4.9-star rated by 2,000+ Texas homeowners. Get a free estimate today."

That's a benefit (custom inspections from pros), proof (4.9-star rating, 2,000 homeowners, Texas-specific), and a CTA (free estimate today). All in 89 characters.

Description Variety

Provide 3-4 description variants per RSA, each emphasizing a different angle. One on price, one on speed, one on quality, one on guarantee. Let Google test which resonates with which audience segment.

Common Description Mistakes

• Repeating the headline wastes the slot. Descriptions should add information, not echo it.
• All-CTA descriptions ("Visit our site! Click now! Learn more!") feel desperate and underperform.
• Buzzword stuffing ("Best-in-class enterprise-grade solutions") signals that you have nothing concrete to say.
• Too short. A 40-character description on a 90-character allowance looks lazy and underperforms.

The strongest descriptions read like they were written by a thoughtful human who understands the audience. The weakest read like they were written by a committee trying not to offend anyone.

Pinning, Asset Variety, and RSA Strategy

RSAs let you "pin" individual headlines or descriptions to specific positions, forcing them to appear in slot 1, 2, or 3. Pinning is powerful — and over-pinning is the most common reason RSAs underperform.

How Pinning Works

Pin a headline to position 1 and Google will only show that headline in the first position, never in positions 2 or 3. If you pin nothing else, Google still has freedom to test the other slots. Pin every headline to a specific position and you've effectively converted your RSA back into a single static ad — no algorithmic optimization, no testing, no learning.

When to Pin

Pin only when there's a hard requirement. Examples: legal disclaimers that must appear in a specific position, branded keywords that must always be present, regulatory phrases. For most accounts, the right amount of pinning is zero or one pin per RSA.

Pinning to Multiple Positions

You can pin a headline to multiple positions ("position 1 or 2") which preserves more flexibility while still anchoring the headline somewhere prominent. This is usually better than rigid single-position pinning.

Asset Variety Discipline

Google's RSA validator flags ads with too-similar headlines as "Low Variety." Treat that warning seriously. Headlines that are minor variations of each other ("Free Shipping," "Free Delivery," "No Shipping Fees") give the algorithm nothing to test. You want angles, not synonyms.

Multiple RSAs per Ad Group

Best practice in 2026 is 1-2 RSAs per ad group. More than that fragments learning and starves each ad of impressions. If you want to test fundamentally different messages, run them as separate experiments rather than as parallel RSAs in the same ad group. Search Engine Journal has documented the diminishing returns of running 3+ RSAs per ad group.

A Real Methodology for Ad Copy Testing

RSAs have changed copy testing dramatically. You can't run clean A/B tests at the headline level the way you used to with Expanded Text Ads, because Google rotates and assembles assets dynamically. But you can still test methodically — you just have to test at a higher level.

Test at the Ad Group Level

The cleanest unit of test is the entire RSA: messaging direction A versus messaging direction B. Build two RSAs in the same ad group, give them dramatically different angles (not minor variations), and let them run for a meaningful sample size.

Use Drafts and Experiments

Google's built-in Experiments feature lets you split traffic between an original campaign and a test campaign with controlled differences. Use it for major copy direction tests rather than ad-group-level tests when the change involves multiple ads.

How Long to Run a Test

What to Measure

Don't call winners on CTR alone. CTR can rise while conversion rate falls, leaving total conversions flat or worse. Measure conversion rate, conversions per impression, cost per conversion, and (where possible) revenue per impression. The right primary metric depends on your campaign goal.

Statistical Significance

Use a free split test calculator or platforms like Optmyzr to check statistical significance. The rule of thumb: if the p-value isn't under 0.05, the difference might be noise. Calling winners early is the most common mistake in copy testing and produces a long string of "wins" that don't replicate.

Analyzing Results Without Fooling Yourself

Most copy test analyses are wishful thinking dressed up in numbers. The discipline is to ask hard questions before declaring a winner.

Question 1: Did You Hit Sample Size?

If you didn't reach your target sample size, you don't have a result. You have an indication that may or may not hold up. Resist the urge to call the test early because you're excited about the trend.

Question 2: Did Anything Else Change?

If the test ran across a major change — a budget increase, a Smart Bidding adjustment, a product launch, a holiday — the result is contaminated. You can't isolate the effect of the copy change from the effect of the other change.

Question 3: Is the Effect Big Enough to Matter?

A statistically significant 0.3% lift in CTR might be real but isn't worth implementing. The cost of changing copy across an account is real (operational time, learning resets, stakeholder confusion). Save your changes for meaningful lifts — typically 10% or better on the primary metric.

Question 4: Does It Replicate?

A single winning test is interesting. A pattern of winning tests across multiple ad groups is a finding. Before you rebuild your account around a single test result, run the same change in a second ad group and see if the pattern holds.

What to Do With Losers

Losing variants aren't failures — they're information. Why did they lose? What does that tell you about the audience or the offer? The accounts that improve fastest are the ones that learn from losing tests as much as from winning ones.

Pair copy testing with the broader signal hygiene practices in our click fraud guide and the diagnostic frameworks in our competitive analysis guide to make sure your test data is trustworthy.

Common Ad Copy Mistakes

A short list of the mistakes I see most often in account audits — and what to do instead.

Mistake 1: Generic Claims

"Best quality." "Industry-leading." "World-class." These phrases have been so overused they're invisible. Replace them with specifics: "30% faster than the industry average," "Built by engineers from Apple and Google," "Used by 8 of the 10 largest law firms in Texas."

Mistake 2: Feature Lists Instead of Benefits

Features describe what the product is. Benefits describe what the user gets. "256-bit encryption" is a feature. "Bank-level security so your data stays yours" is the benefit. Always translate features into benefits before they hit the headline.

Mistake 3: Ignoring the Search Intent

Ads should mirror the query intent. A user searching "buy red running shoes size 10" is in transactional mode and should see transactional copy ("Red Running Shoes Size 10 — In Stock — Free Shipping"). A user searching "best running shoes for plantar fasciitis" is in research mode and should see comparison copy ("Top 5 Shoes for Plantar Fasciitis — Doctor Reviewed"). Same advertiser, different ads.

Mistake 4: Forgetting Mobile

Mobile users see your ad on a smaller screen with less surrounding context. The first headline matters more than ever. Front-load your most compelling angle and don't bury the lead.

Mistake 5: Ad Policy Violations

Excessive capitalization, exclamation marks in headlines, trademark violations, and unverifiable claims will get your ads disapproved. Read Google's ad policies regularly — they change.

Mistake 6: Set It and Forget It

Even the best ad copy goes stale. Refresh your top RSAs every quarter, even if performance is steady. The act of refreshing keeps your team thinking about the offer and prevents the slow drift toward complacency.

For competitive insight on how the best advertisers in your category structure their copy, Sentinel's Google Ads Clicker Bot is a useful starting point.