Search Ad Copy That Converts

Search ad copy is the single most-leveraged change you can make to a campaign that's already getting impressions. Better copy lifts CTR, which lowers CPC (via Quality Score), which lifts conversion volume at the same budget. In 2026 the format is almost entirely Responsive Search Ads (RSAs), which means writing 15 headlines and 4 descriptions that the algorithm combines into 43,680 possible ad permutations. This module covers what the data actually shows works, how to structure copy assets, pinning logic, ad strength, and the testing protocol that compounds over time.

1. From ETAs to RSAs — the format that ended manual control

Until 2022, Expanded Text Ads (ETAs) let you write a single ad with three headlines and two descriptions in fixed positions. You knew exactly what users would see. Google deprecated ETAs in June 2022, replacing them with Responsive Search Ads (RSAs) as the only ad format for Search campaigns.

RSAs work differently:

• You provide up to 15 headlines (30 character limit each) and 4 descriptions (90 character limit each).
• Google's algorithm combines headlines and descriptions into ad permutations at auction time.
• The algorithm uses signals (query, user history, device, time) to pick the best combination for each impression.
• You see aggregated performance, not individual ad-combination performance.

The trade-off: you give up manual control but gain machine-learning combination testing across thousands of variants. The data is consistent — brands using RSAs well outperform brands trying to recreate ETA-style manual control with heavy pinning.

2. Anatomy of an RSA

How they appear in SERP:

• 2-3 headlines display, separated by "|" or hyphen
• 1-2 descriptions display below
• Display URL shows your domain plus optional path segments
• Sitelinks, callouts, and other extensions display below or beside

3. The 15 headlines + 4 descriptions structure

Don't treat the 15 headline slots as random brainstorm output. Structure them by category so Google has variety to test across all the dimensions that matter.

Recommended headline mix (15 total)

Recommended description mix (4 total)

• Description 1: Expanded value prop with proof point (numbers, specifics).
• Description 2: Differentiation vs alternatives + secondary CTA.
• Description 3: Trust signal + offer/guarantee.
• Description 4: Use-case or audience-specific framing.

4. Headline categories with examples

On the average, five times as many people read the headline as read the body copy. When you have written your headline, you have spent eighty cents out of your dollar.

— David Ogilvy, Confessions of an Advertising Man (1963)

5. The first impression: position 1 and position 2

RSAs show 2-3 headlines per impression. The first two slots are typically the most visible and matter most for CTR. Google rotates which headlines appear there based on context.

The general pattern Google's algorithm prefers:

• Position 1: Most relevant to the query (often a keyword-match headline)
• Position 2: Most differentiating value prop or USP
• Position 3 (when shown): Brand or CTA

You can influence this with pinning (next section), but the strongest accounts trust the algorithm and write all headlines to be position-agnostic. Each headline should stand alone as a compelling phrase.

6. Pinning — when to use it, when not to

Pinning forces a specific headline (or description) to appear in a specific position (Position 1, 2, or 3 for headlines; Position 1 or 2 for descriptions). Pinning reduces the number of permutations Google can test, which can hurt RSA performance.

When pinning makes sense

• Brand control: Pin the brand-name headline to Position 1 for brand campaigns.
• Compliance / legal: Pin disclaimers or required language (financial, healthcare, regulated categories).
• Promotional accuracy: Pin the promo offer (e.g., "25% Off Through Sunday") so it doesn't accidentally not show.

When pinning hurts

• Pinning more than 1-2 headlines reduces RSA permutations dramatically.
• Pinning the same position with multiple headlines creates rotation only within that pin group (Google rotates among them but can't use them elsewhere).
• Heavy pinning gets you a "Poor" Ad Strength score — not catastrophic but loses some serving privileges.

7. Ad Strength score — what it measures and what to ignore

Google shows each RSA an Ad Strength rating: Incomplete, Poor, Average, Good, Excellent. The factors Google measures:

• Headline count — 15 is ideal; fewer reduces strength.
• Headline uniqueness — varied headlines without redundancy.
• Keyword inclusion — headlines containing target keywords.
• Description count — 4 is ideal.
• Asset pinning — over-pinning lowers strength.

What to do with Ad Strength

Aim for "Good" or "Excellent" for most ads — the score correlates loosely with performance. But:

• Don't sacrifice copy quality for the score. A "Good" ad with great copy outperforms an "Excellent" ad with mediocre copy.
• Don't obsess over moving from Good to Excellent — the marginal performance lift is small.
• Don't add weak headlines just to hit 15. Five strong headlines beats 15 mediocre ones.

Draft an ad in the builder below and watch the relevance score move — it shows what a higher Quality Score saves you on CPC versus a weaker competitor.

8. Patterns that win across categories

The proof-point pattern

Specific numbers outperform generic claims. Include verifiable proof points.

• "Save 30 Minutes Daily" > "Save Time"
• "Used by 12,000 Teams" > "Trusted by Many"
• "Ships in 24 Hours" > "Fast Shipping"

The deflection pattern

Acknowledge a common objection and reframe.

• "No Credit Card Required"
• "Setup in Under 2 Minutes"
• "Cancel Anytime"
• "No Long-Term Contracts"

The audience-fit pattern

Speak to a specific buyer persona.

• "For Solo Marketers"
• "Built for Series A Startups"
• "Designed for Plus Size"

The contrast pattern

Position against an obvious alternative.

• "Faster Than Mailchimp"
• "Half the Cost of Salesforce"
• "Beyond Yoga — Made in USA" (positioning against import competitor)

The first-person voice pattern

Direct address often outperforms third-person.

• "Find Your Perfect Fit" > "Customers Find Perfect Fit"
• "Save Hours Every Week" > "Users Save Hours"

9. Dynamic Keyword Insertion (DKI) and ad customizers

Dynamic Keyword Insertion

DKI inserts the matched keyword into the headline at serve time. Syntax: {KeyWord:Default Text} — if the keyword fits the character limit, it's used; otherwise the default appears.

When DKI works

• Product-search campaigns where you've clustered closely-related keywords.
• You want the headline to mirror the query for relevance.
• You've audited all keywords for capitalization and grammar (DKI uses the keyword as-is).

When DKI hurts

• Your keyword list has typos or odd capitalization.
• Your keywords are long-tail and the inserted phrase reads awkwardly.
• You have brand-protection sensitivities (DKI of competitor brand terms is forbidden by trademark policy).

Ad customizers

Ad customizers insert business data (inventory count, price, location-specific text) into ads. Examples:

• {COUNTDOWN("2026/12/25 23:59:59")} — counts down to a deadline
• {LOCATION(City)} — user's city
• Business Data feed: insert price, stock count, or any custom field

10. Compliance and policy

Trademark policy

• You cannot use a competitor's registered trademark in ad copy without permission.
• You CAN bid on competitor brand keywords (the trademark owner can't prevent that).
• Brand owners can file trademark complaints with Google to remove offending ads.

Restricted categories

Categories with extra policy requirements: financial services, healthcare/pharma, alcohol, gambling, political ads, employment. These often require:

• Pre-approval / certification
• Required disclaimers (can be pinned)
• Restricted geo targeting
• Specific landing-page requirements

Common policy disapprovals

• Excessive capitalization — "SAVE NOW" gets disapproved; "Save Now" passes.
• Misleading content — claims that can't be substantiated.
• Inappropriate symbols — multiple punctuation marks, special characters.
• Unsupported claims — medical claims, financial guarantees.

11. The testing protocol

RSA-level testing

You can run multiple RSAs in the same ad group. Google rotates them and reports performance separately. Best practice:

• Run 2-3 RSAs per ad group during active testing.
• Each RSA should have a clear hypothesis (different angle, different value prop, different audience).
• After 2-4 weeks (or 1000+ impressions per RSA), pause losers and add new variants.

Headline-level rotation

Within an RSA, Google shows performance metrics per asset: "Best", "Good", "Low", "Pending". The workflow:

1. Every 2-4 weeks, check asset reporting.
2. Replace "Low" headlines with new variants.
3. Add headlines in new categories (e.g., if all your social proof is similar, add audience-fit variants).
4. Never remove all "Best" assets at once — you'll restart Learning.

Google Experiments (A/B testing)

Google Ads Experiments lets you split a campaign 50/50 with structural changes. Use for testing:

• Different RSA strategies (creative-led vs benefit-led).
• Different bid strategies.
• Different audience signals.
• Different landing pages.

12. Mobile vs desktop copy considerations

• Mobile shows shorter ad displays (often only 2 headlines and 1 description visible above the fold).
• Headlines that work mobile-first: shorter, more direct, more action-oriented.
• Set "Final mobile URL" for mobile-specific landing pages (especially if your desktop site is not mobile-optimized).
• Check ad previews on mobile in Google Ads UI; the truncation is real.

13. Localization and language

Google Ads supports per-campaign language settings. Best practice for multi-language accounts:

• One campaign per language (don't mix English and Spanish keywords in the same campaign).
• Translate, don't auto-translate. Native speakers catch nuance and cultural fit.
• Test culturally-specific value props. "Free shipping" matters more in some markets; "Same-day delivery" matters more in others.

14. Sitelinks, callouts, structured snippets, image extensions

Ad assets (formerly extensions) increase ad real estate and CTR. Use them all.

Sitelinks (account, campaign, or ad-group level)

• 4-8 sitelinks per campaign typical.
• Each sitelink: text (25 chars) + 2 description lines (35 chars each).
• Use to surface secondary destinations (pricing, features, customers, support).

Callouts

• Short non-clickable text snippets (25 chars).
• 4-10 callouts typical: "Free Shipping", "24/7 Support", "Money-Back Guarantee".

Structured snippets

• Header-value format: "Types: Trail, Road, Marathon, Casual".
• Header options include: Brands, Models, Types, Featured, Services.

Image extensions

• Product images shown alongside text ad on mobile/desktop SERP.
• Square (1200x1200) and landscape (1200x628) recommended.
• High-quality lifestyle images outperform stock photography.

Other extensions worth using

• Price extensions (for SKUs with clear pricing)
• Promotion extensions (for time-bound offers)
• Lead form extensions (for lead-gen campaigns)
• Business information (logo, business name)

15. The 12 most common ad copy mistakes

16. Anti-patterns: what NOT to do

Quick reference: the “good ad copy” checklist