In responsive ads, the system combines headlines, descriptions and images into dozens of versions. Evaluating a single ad no longer makes sense. What gets evaluated are the assets, the combinations that show most and the message patterns that work.
The essentials in a definition
Responsive ads automatically combine headlines, descriptions and images supplied by the advertiser to create the combination with the best predicted performance. Assessing the performance of combinations and assets, not individual ads, is the right approach. Relevance and asset quality influence Quality Score, and policies apply to every possible combination.
From the ad to the set of assets
Before, you wrote one ad and compared it with another. Now you supply a set of elements and the system decides the combination best suited to each search or person. What you control is the quality and variety of the set.
This changes the work: instead of optimising final texts, you optimise the asset library, the constraints (what must or must not appear) and the diversity of messages.
How to read asset reports
Platforms rate assets by relative performance, qualitatively. Those labels are a guide, not a verdict: a low-rated asset may be serving smaller audiences, and a top-rated one may cannibalise others.
Cross the rating with cost, conversions and quality. Ask for combinations rather than individual assets whenever the report offers them.
Managing the asset set
- Cover different message angles
- Make sure any combination makes sense
- Read ratings alongside cost and conversions
- Replace only the weakest every 4 to 6 weeks
Good composition practice
Cover different angles: price, quality, proof, urgency, location, main benefit. Avoid near-identical headlines, which give the system little to choose from. Make sure any combination makes sense: read a few real examples to spot phrases that contradict each other.
Use pinning sparingly. Pinning too much reduces the flexibility that justifies the format.
In a responsive ad, the product is the library, not the final text.
Improvement cycle
Every four to six weeks, replace the weakest assets with new hypotheses. Keep the best and document what you learned: which angles work, which words avoid useless clicks, which offers attract a better audience.
Avoid replacing everything at once. Continuity gives the system stability and lets you measure the effect of each change.
Decision table
| Element | Good practice | Common mistake |
|---|---|---|
| Headlines | Vary angles: benefit, proof, offer, objection | Repeat the same headline with words swapped |
| Descriptions | Complement headlines without repeating | Generic, vague text |
| Pinning assets | Use only when required by law or brand | Pin everything and lose variation |
| Reading results | Observe assets and combinations over time | Judging after a few days |
Common mistakes to avoid
- Combinations that do not make sense. Any headline can appear with any description; check that all combinations are acceptable.
- Redundant assets. Many very similar assets give the AI no real variety.
- Expecting a single “winner”. The system serves different combinations to different people.
How to measure the result
Track click-through rate, conversion and asset performance rating, and the number of new assets tested per month. Remove low-delivery assets after enough data.
Limits and exceptions
Responsive ads do not give full control over the final combination, and combination reports are limited. In regulated sectors or with mandatory legal notices, pinning certain assets may be necessary, which reduces optimisation. If the exact message is critical, consider lower-automation ads or a rigorous review of all possible combinations. The point is to acknowledge that choosing responsive ads trades some control for more adaptation.
Illustrative example
Illustrative example: a clinic supplies 12 headlines, split between benefit, proof, location and offer, and 4 descriptions. It reviews asset reports after a month, replaces those with low delivery and keeps those that are useful. Headlines with location perform better on mobile, and the team uses that on the landing pages. Think with Google studies help generate hypotheses for new angles.
Frequently asked questions
How many headlines and descriptions should I write?
As many as you can genuinely differentiate. More is not better if they are near-identical.
Should I pin assets?
Only when required (for example legal text). Pinning too much reduces the format’s benefit.
How do I know whether an asset works?
By combining the platform rating with cost and real conversions.
How many headlines and descriptions should I provide?
The maximum allowed, as long as they carry truly different angles and are all acceptable in any combination.
Should I pin headlines?
Only when mandatory (legal notices or brand); pinning too much reduces optimisation capacity.
In responsive ads, you assess the asset set and its combinations; provide distinct angles, all acceptable in any combination, and avoid pinning without need.
Sources and recommended reading
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.