Google RSA Copy in the AI Max Era: How to Write Ads That Guide the System Instead of Getting Overridden

AI Max can generate headline and description variants from your landing pages and use them in place of your manually written copy. If you do not understand how to write RSA assets and use pinning strategically, you are effectively letting the system write your ads for you, and the results may not reflect your brand voice, compliance requirements, or conversion strategy.

The September AI Max migration brought with it a capability that many advertisers did not fully prepare for: AI-generated text assets. In AI Max campaigns, Google’s system can produce new headline and description variants by scanning your landing page and existing assets, then serve those AI-generated variants alongside or instead of your manually written copy if its models predict they will outperform.

This is not purely a threat. AI-generated assets sometimes identify effective angles that advertisers did not test. But for brands with specific messaging requirements, compliance language, or conversion-optimised copy developed through testing, an unmanaged AI Max text asset setup can erode the quality of ad copy that you have spent months or years refining.

The solution is not to disable AI Max text assets entirely (though you can). It is to write RSA assets and use the pinning system in a way that keeps your best copy in the rotation while still giving the system enough flexibility to learn and optimise. Here is how to do that in 2026.

What AI Max Does to Your Copy

AI Max’s text asset generation works by scanning your final URL landing page content and existing RSA assets. It uses natural language understanding to identify the main value propositions, features, and calls to action on the page, then generates headline and description variants that reflect those elements in different combinations and framings.

The generated variants are added to your asset pool and compete with your manually written headlines and descriptions for inclusion in served ads. Google’s delivery system then rotates through the full pool, learning which combinations perform best for different query types and user contexts. Over time, higher-performing combinations appear more frequently.

The practical risk is that if your landing page contains generic marketing language, AI-generated assets will reflect that generic language. If your landing page mixes product features with brand claims without clear hierarchy, AI-generated assets may elevate the wrong elements. The quality of AI-generated copy depends directly on the quality of your landing page copy, not on the quality of your manually written RSA assets.

This means AI Max text asset generation is both a copy quality amplifier and a copy quality risk, depending on the strength of your landing page. Improving your landing page copy is therefore not just an SEO and conversion rate optimisation task in 2026; it directly affects your Google Ads copy quality through AI Max’s asset generation mechanism.

Asset Quality Scoring

Google rates each RSA headline and description with a performance rating: Best, Good, Low, or Learning. In AI Max campaigns, asset performance ratings take on additional weight because the system uses them as one of its signals for which assets to include in generated ad combinations.

Assets rated “Low” are still served, but at lower frequency. Assets rated “Best” are served frequently and may be used as the foundation for AI-generated variant creation. This means your Best-rated assets are particularly important in AI Max campaigns: they function as quality anchors for the system’s learning, and they may be reproduced in AI-generated variants with slight modifications.

Assets that are consistently rated “Low” should be reviewed and either rewritten or removed. Unlike earlier campaign types where keeping a broader asset pool was generally neutral, in AI Max campaigns low-performing assets can slightly drag overall ad quality scores and reduce the frequency of your best combinations.

Pinning Strategy for AI Max

Pinning in RSA campaigns locks a specific asset to a specific position in the ad. Position 1 is the first headline, Position 2 is the second headline, Position 3 is the third headline, and similarly for descriptions. Pinning ensures that asset always appears in that position when the ad is served.

The tradeoff with pinning is that it reduces combination variety, which can limit the learning the system gets from testing different arrangements. Over-pinning all positions removes most of the responsive element from responsive search ads. Under-pinning leaves your most important copy elements subject to variable positioning.

The recommended pinning approach for AI Max campaigns in 2026 is selective pinning of your highest-value assets while leaving significant combination space for the system to optimise. Specifically: pin your most critical headline, typically the one that directly states what you do or names the product, to Position 1. This ensures users always see the most important identifying information first. Pin your primary call to action headline to Position 3. Leave Position 2 unpinned with three to four good headline options, allowing the system to test which supporting headline context works best for different query types.

For descriptions, pin your most important Description 1 asset if it contains compliance language, a specific guarantee, or a conversion-critical claim. Leave Description 2 unpinned with two to three variants. This gives you control over the most critical message elements while leaving enough flexibility for the system to learn.

Writing RSA Headlines for AI Max

Effective RSA headlines in the AI Max environment are short, specific, and semantically distinct from each other. Short means under 25 characters where possible, leaving room for dynamic keyword insertion and combination flexibility. Specific means each headline communicates one clear, distinct message rather than a vague brand claim. Semantically distinct means your 15 headlines cover genuinely different angles, not the same message reworded.

The distinct categories that a well-constructed headline set should cover are: what you do (product or service description), why choose you (unique differentiator or proof point), price or value signal, social proof or credibility element, urgency or scarcity signal, and the primary call to action. With 15 headline slots and these six categories, you can write two to three variants per category that cover the same angle with different specific framings.

AI Max’s keyword expansion means your ads may serve on queries that are adjacent to your core keywords. This makes it more important that your headlines work for a broader semantic context, not just for the exact keywords you are bidding on. Headlines that are too narrowly keyword-specific may read as off-context when served on expanded queries. Writing headlines that communicate your core value proposition without requiring the reader to already know your brand or product type performs better in AI Max’s expanded serving contexts.

Writing Descriptions

Descriptions in AI Max campaigns carry more weight than in standard Search campaigns because they can be adapted by the AI text asset generation system into new variants. The most important description writing principle is factual specificity: specific claims that reference real data, timeframes, or verifiable attributes are significantly harder for AI to dilute into generic paraphrase than vague brand claims.

A description that says “Most clients see a 30% reduction in wasted spend in the first 90 days” is specific, verifiable, and difficult to paraphrase into meaninglessness. A description that says “We help businesses grow with smarter advertising” is generic enough that AI-generated variants of it will be indistinguishable from a thousand other advertisers. Write descriptions that could only come from your specific offering with your specific results.

Character count discipline is also more important in AI Max. Descriptions that come close to the 90-character limit are less likely to be combined cleanly with AI-generated headline variants. Writing descriptions that communicate their full message in 70 to 80 characters gives the system more flexibility in ad construction while preserving your core message.

What Not to Pin

Do not pin assets that are not significantly better than alternatives. If you are pinning a headline to Position 1 that is only marginally different from the other headlines in your pool, you are sacrificing combination flexibility for no meaningful gain. Pin sparingly and for clear, defensible reasons: compliance requirements, critical brand messaging, calls to action that must always appear in a specific position.

Do not pin all three headline positions with single assets. An RSA with all three positions pinned is functionally an expanded text ad with no learning capability. The responsive element is eliminated entirely, and you lose the combination testing that makes RSA format valuable. If you feel you need to pin all three positions, you are probably better off with a different campaign structure than an RSA in an AI Max campaign.

Do not pin low-performing assets. If an asset is rated “Low” in performance reporting and you have it pinned to a position, you are forcing a poor-performing element into every ad served by that RSA. Review pinned assets against their performance ratings quarterly and remove pins from assets that have declined in quality rating.

Monitoring Asset Performance

In AI Max campaigns, asset performance monitoring should happen at least monthly. The Asset Report in Google Ads (at the campaign or ad group level) shows each headline and description’s performance rating, impression count, and combination frequency. Review this report with three specific questions: which assets are rated “Low” and should be rewritten or removed, which assets are rated “Best” and could be expanded into variants, and whether AI-generated assets have been added to your pool and how they are performing.

To identify AI-generated assets, look for assets in your Asset Report that you do not recognise from your own copy input. These will have been created by the text asset generation system. Check their performance ratings and decide whether to keep them (if performing well), modify them (if the angle is good but the execution is weak), or remove them (if they misrepresent your brand or offer).

At Incisive Growth, our Google Ads management process includes monthly asset performance reviews for all AI Max campaigns, with structured decisions on asset pruning, pinning adjustments, and copy refreshes based on what the performance data shows.

What This Means Going Forward

The trend in Google Ads product development is toward more AI control over ad copy, not less. The text asset generation in AI Max is one step in a direction that will likely produce more automated copy creation capabilities in future product updates. The appropriate response is not to resist this by over-pinning or disabling AI features, but to write RSA copy and configure pinning in a way that makes the AI-managed parts of ad creation work well rather than poorly.

The advertisers who will navigate this well are those who invest in high-quality, specific, factually grounded ad copy that gives AI systems good material to work with, use selective pinning to protect their most critical message elements, monitor asset performance regularly to catch and correct AI-generated copy that misrepresents their brand, and maintain strong landing page copy quality since this directly influences what AI Max generates.

If your Google Ads accounts went through the September AI Max migration and you have not reviewed your RSA assets and pinning strategy since, that review is overdue. The team at Incisive Growth offers Google Ads audits that cover exactly this: what your current asset pool looks like, which assets are underperforming, what the AI text generation system has added, and what pinning changes would improve performance.

Frequently Asked Questions