Google Ads generates responsive search ads by mixing and matching the headlines and descriptions you provide, using machine learning to test different combinations and serve the one most likely to perform well for each individual search. You supply the raw ingredients (up to 15 headlines and 4 descriptions), and Google’s system assembles them in real time based on the searcher’s query, device, browsing context, and predicted performance.
What You Provide vs. What Google Builds
When you create a responsive search ad, you write multiple headlines and descriptions rather than a single fixed ad. Google allows up to 15 headlines and 4 descriptions per ad. From that pool, the system selects a subset each time your ad is eligible to appear. A served ad can show up to three headlines and two descriptions, though the exact count varies. Google may drop a headline or description entirely if its models predict the ad will perform better without it.
The text you enter in the Headline 1 and Description 1 fields will almost always appear when your ad is served. The second headline, third headline, and second description show up based on what Google predicts will work best, or when there’s enough screen space to display them. This means your strongest, most essential message should go in those first slots.
How Machine Learning Picks Each Combination
Every time someone searches a query that triggers your ad, Google’s system evaluates which combination of your assets to show. It weighs several signals: the specific words in the search query, the user’s device type and screen size, and historical performance data on how different asset pairings have performed in the past. The goal is to match the messaging that resonates most with that particular searcher in that moment.
Different combinations of content perform differently depending on the user and query. A headline emphasizing price might win for someone searching “affordable running shoes,” while a headline about quality or reviews might win for “best running shoes.” Over time, the system learns which pairings drive clicks and conversions for your specific audience, and it shifts serving toward those combinations. In some cases, the system may decide not to show certain pieces of content, like a second description line, when omitting it leads to better results.
How Google Creates Assets Automatically
Beyond the headlines and descriptions you write yourself, Google can also generate additional text assets on your behalf through a feature called text customization. This uses a mix of extractive techniques (pulling snippets from your landing page titles, meta descriptions, and on-page content) and generative AI to produce new headlines or descriptions grounded in your actual product or service.
The inputs for this automated generation include your domain, your landing page content, the keywords in your ad group, and the ads you’ve already written. Google’s generative models can synthesize thousands of query signals alongside your landing page content to create assets for searches you might not have anticipated. For example, if your landing page sells shampoo, the system won’t generate a headline about makeup. Quality models check each generated asset for accuracy and relevance before it can serve.
These auto-generated assets are reviewed and refreshed at least every 48 hours. If you update your landing page, new assets reflecting those changes will appear within that window. Importantly, a Google-generated asset will only serve if it’s predicted to outperform the headlines and descriptions you uploaded yourself. Your own copy remains the baseline.
Pinning Assets to Specific Positions
If you need a particular headline or description to always appear in a certain spot (a legal disclaimer, a brand name, a specific offer), you can “pin” it to a position. Pinning Headline 1 means that asset will always show in the first headline slot whenever the ad is served. You can pin multiple assets to the same position, and Google will rotate among them.
The tradeoff is flexibility. The more assets you pin, the fewer combinations the system can test, which limits its ability to optimize. An ad with every position pinned behaves much like a traditional static ad. Google’s Ad Strength indicator will typically score pinned ads lower because the system has less room to experiment. If you only need to guarantee one message appears, pin just that one and leave the rest open.
How Google Measures Asset Performance
Google provides two layers of feedback on your responsive search ads. Ad Strength is a pre-serving score that evaluates the diversity and quantity of your assets. It ranges from “Poor” to “Excellent” and encourages you to add a wider variety of headlines and descriptions up to the maximum. Think of it as a setup grade: it reflects your potential, not your results.
Asset-level performance metrics, on the other hand, show how individual headlines and descriptions actually performed after serving. As of June 2025, Google provides full performance statistics for each asset, including clicks, impressions, and conversions. However, Google cautions that ratio metrics like click-through rate or cost per conversion at the individual asset level should be treated as directional indicators only. Because assets always serve in combination with other assets, isolating the impact of a single headline is inherently imprecise. Evaluating performance at the campaign or ad group level gives you a more accurate picture.
How to Give the System Better Material
The quality of your responsive search ads depends heavily on what you feed the machine. Write headlines that are genuinely distinct from each other. If five of your fifteen headlines are slight rewrites of the same selling point, you’re wasting slots and limiting the system’s ability to find winning combinations. Aim for variety across your value propositions: price, features, brand reputation, urgency, social proof, and calls to action.
Make sure every headline and description can work alongside any other. Since Google assembles them dynamically, a headline that only makes sense when paired with a specific description will occasionally appear out of context. Each asset should be a standalone, coherent piece of messaging. Keep your landing page content accurate and up to date, since Google’s text customization feature pulls directly from it. If your page is outdated or vague, any auto-generated assets will reflect that.