AI product titles in Google Ads: what they are and how to stay in control

AI product titles are automatically generated versions of your Shopping ad titles that aim to improve relevance and performance. Google now rewrites these labels in real time. If a different version of your product title is predicted to be more relevant to a specific search query, Google can serve that version instead of the one you submitted in your feed. This is not a future feature. It started appearing in Google Ads accounts in late August 2026, alongside a dedicated product titles report that lets you compare AI-generated labels against your originals.

This article explains how AI product titles work, what the new report shows, how Google decides which heading to serve, and what options you have if you want to limit or disable the feature.

What are AI product titles in Google Ads?

AI product titles are rewritten versions of your Shopping ad copy, generated by Google and served in place of your original feed label when predicted to improve ad performance. The system reads your Google Merchant Center data, your website content, and the searcher's query, then generates a heading it believes is more relevant for that specific search.

Google is beta testing a new product titles report within Google Ads specifically for AI-created product titles. The report shows AI-generated labels in your listings when they are more relevant than your original titles.

This is separate from standard feed rules or supplemental feeds. You are not rewriting the label in your data source. Google is generating a new version at the point of serving, drawing on information it already has about your items.

The feature sits inside AI Max for Shopping campaigns, which Google describes as a beta suite of features that uses its system to optimise Shopping campaigns and connect ads with potential customers making complex, conversational searches. According to Google internal data (Global, 2026), advertisers that activate AI Max for Shopping see on average five percent more conversions or conversion value at a similar CPA or ROAS compared to those who do not use the full feature set.

Why Google Shopping ad titles are changing

Shopping ads were built around static product feeds. You wrote a title, submitted it, and that copy appeared in listings. The problem is that search behaviour has changed significantly. Shoppers now ask longer, more specific, more conversational questions, and a static label written for a broad query rarely fits a precise long-tail search well.

Each change creates new surfaces where conversational queries can trigger product ads. If Shopping ads can now appear in AI Mode responses to complex questions, the ad copy that appears there becomes more consequential. A static product title pulled from a data source may be less effective in that environment than a heading generated by the system to meet the specific language of the query.

The AI-generated label is Google's response to this. Rather than relying on the single title you submitted, automation can generate a version that speaks directly to the intent behind the search.

Text customisation allows Google to dynamically update product titles to align with customer search queries and intent. By using content from your Merchant Center data source, the setting automatically generates highly relevant, query-matched product titles when predicted to improve ad performance. This flexibility allows campaigns to go beyond static catalogue data, delivering personalised labels at a scale that is impossible to achieve manually.

How Google decides which title performs best for relevant searches

The most important thing to understand is that Google does not always serve the AI-generated title. Your original label remains in play. The system only replaces it when the generated version is predicted to perform better.

Customised titles will only serve if they are predicted to perform better than the original titles uploaded. If the system does not predict a performance uplift, the original label will serve.

Google generates titles by treating your Merchant Center data as a product database rather than a static list. It extracts key attributes and combines them to match the query.

The data source is treated as a deep product database. It extracts key product attributes such as "waterproof," "stability-focused," or "available for store pickup" to match complex, long-tail search terms with high precision. Understanding which attributes matter most for a given search is central to how the system chooses what to include in each generated heading.

A practical example from the Google Ads Help Centre shows how this can work. An item submitted as "Lounge deep 93" sofa" might serve as "Made in USA 93" Deep Sofa" or "Lounge Deep Sofa with Soft Cushions" depending on what the shopper searched for. The original label also remains eligible and will serve when it is the strongest match.

Every generated title also goes through quality checks. Google states that generated labels must be grounded in your actual product data, must look natural and professional to shoppers, and must not fabricate attributes the item does not have.

Performance data from previous impressions also influences the process. Google allows the system to learn over time which generated labels drive better outcomes for a given search term cluster. This is why the product titles report, which surfaces performance data alongside each generated heading, matters so much for evaluating whether automation is working in your favour.

What the new product titles report shows

The product titles report appeared in Google Ads accounts from late August 2026. It was first spotted by Yash Mandlesha, who shared a screenshot on LinkedIn. The report includes the AI-created product title, the original title, impressions, product clicks, CTR, cost, and average CPC.

The report interface also includes a note that AI-created labels will appear once they have received enough impressions, so accounts with lower traffic may not see data immediately.

The columns that matter most for evaluation are the side-by-side title comparison and the conversion metrics you choose to include. This combination makes it possible to assess whether generated labels are improving outcomes beyond CTR alone. Impressions and clicks tell you reach. Conversion data tells you whether the traffic is worth anything.

The report describes what it shows as a sample of product titles created by Google that showed in your listings. It is not a full accounting of every impression served with a rewritten label. It is a representative sample.

To read the report effectively, treat each row as a direct comparison between two states. A row might look like this:

1. Original title: "Men's Running Shoe SKU-4821." AI-generated title: "Men's Lightweight Road Running Shoe, Cushioned Sole." Impressions: 4,200. CTR: 3.1%. Conversion rate: 1.4%.
2. Original title: "Ceramic Mug 350ml." AI-generated title: "Handmade Ceramic Coffee Mug, 350ml, Dishwasher Safe." Impressions: 1,800. CTR: 4.7%. Conversion rate: 2.9%.
3. Original title: "Blue Yoga Mat." AI-generated title: "Non-Slip Blue Yoga Mat, 6mm, Eco-Friendly TPE." Impressions: 3,100. CTR: 5.2%. Conversion rate: 3.6%.

In all three cases, the generated label added specific attributes the original title omitted: material, use case, and key product details that match more precise search queries. The CTR and conversion rate differences show that the improvement is not just cosmetic. Shoppers clicking the more specific label are more likely to be in the right buying mindset when they land on the page.

This report is also distinct from the product title A/B experiments report that Google introduced earlier in 2026. That tool lets you design controlled experiments. The product titles report is a record of what automation has already done.

How to find the product titles report

The report lives inside Google Ads under the reporting section for Shopping campaigns. Look for a product titles entry in the report categories when you are inside a relevant campaign or account view. The report only populates once generated labels have served enough impressions, so if you enable AI Max for Shopping today, expect a short delay before data appears.

How to control AI product titles in your campaign

This is where advertisers have the most questions. There are two main levers.

Text guidelines

You can help Google create on-brand ad copy by adding text guidelines in campaign settings. If changes to text guidelines have been made in the selected date period, copy that violates the guidelines may still appear in the report, but it would not show in live listings.

Text guidelines let you set term exclusions and messaging restrictions. If there are words, phrases, or positioning angles you want kept out of generated titles, this is where to set them. The text guidelines feature, globally available since February 2026, allows advertisers using text customisation to set term exclusions and messaging restrictions that constrain what the system generates.

Opting out of text customisation

If you want to disable title rewriting entirely, you can turn off text customisation inside your AI Max settings.

To disable text customisation, go to Campaigns in the Campaigns menu. Select the Settings tab, then select the checkbox next to the campaigns you want to edit. Select AI Max, navigate to the Asset optimisation panel, and deselect the Text customisation checkbox. Select Save. Disabling text customisation may impact other AI Max settings.

There is also a separate opt-out path that applies at the account level rather than the campaign level.

You can reach out to your account manager or support teams to opt out of Product Data Customisation for Shopping ad titles. Requests are typically processed within three to five business days.

Keep in mind that opting out of text customisation means your labels will not serve on newer search surfaces. Text customisation is built to perform in newer search experiences such as AI Overviews and AI Mode, so items appear alongside and directly answer customer queries. Opting out is a legitimate choice for brands with strict creative guidelines, but it comes with a visibility trade-off on those surfaces.

How to optimise product titles and key attributes in your product data

Generated labels draw on what is already in your data source. If the catalogue entry is thin, the system has limited material to work with. A strong product feed gives automation better raw inputs and reduces the risk of the system filling gaps with approximations.

The core best practices for titles and descriptions have not changed, but they matter more now that automation is actively using them to construct generated labels.

Using all 150 characters in your title allows the heading to match against a wider range of customer searches. Including the important details that define your item and putting the most important details first both remain core recommendations, since users typically notice only the first 70 characters depending on screen size.

Only around 70 characters are visible in standard Google Shopping results, so the most important attributes such as brand, product type, and key differentiator must appear in the first 70 characters. Prohibited content includes all-caps text, promotional text, gimmicky characters, and the store name when the brand attribute already provides it.

The system builds on what you have submitted. If your original title front-loads brand, product type, and a key differentiator in the first 70 characters, automation has a solid foundation. If your title starts with an internal SKU reference or a creative name with no product type in it, the system may produce something that drifts from your intended positioning.

Beyond the title itself, attributes like colour, size, material, and product type all feed into the matching logic. The Merchant Center data source is treated as a deep product database, and key product attributes are extracted to match complex, long-tail search terms with high precision. A well-populated catalogue means automation has more accurate signals to draw on.

For Shopify stores managing large catalogues, enriching product attributes at scale is the challenge. Feed rules in Merchant Center can automate some of this. For stores managing profitability segmentation alongside title optimisation, custom labels remain the primary tool for grouping items by margin, seasonality, or sales performance so that AI Max bidding and label variation can work together on the right product groups.

Google Shopping title optimization: what the keyword structure should look like

The phrase "titles for Google Shopping" is commonly used interchangeably with "Google Shopping title optimization," but they refer to slightly different things. The former describes the collection of labels in your feed. The latter describes the deliberate process of improving them to improve product visibility and click performance.

To optimize titles effectively, you need to understand how Google uses them. Google Shopping does not use keywords in the same way a Search campaign does. Instead, the system reads your product feed and determines which search queries your items are relevant for. The product title is one of the strongest signals in that process.

To optimize your product titles for Google Shopping, start by identifying the most important keyword for each item type. Use that keyword in the first 70 characters. Then use the remaining space to add supporting attributes: colour, size, material, product category, and any other detail that a shopper would include in a specific search.

Google allows up to 150 characters in the title attribute. Most advertisers do not use all of them. Leaving characters unused means leaving matching potential on the table. A title like "Blue running shoes" is far weaker than "Nike Blue Men's Running Shoes, Lightweight Cushioned Sole, Size 10, Road Running." Every added detail is a potential match for a different search term.

Optimized titles also give the generation system better raw material. When Google's model constructs a generated label for a specific query, it draws on the attributes in your title and description. The more structured and complete your titles and descriptions are, the more accurately automation can match products to relevant searches.

Attributes in your title function differently from attributes submitted in separate feed fields. Both matter, but the title is what shoppers see directly in the listing. Attributes in your title that match the shopper's language are what drive the initial click. Supporting feed fields inform the system's understanding of the item but do not appear in the visible heading.

The reporting question: using performance data to evaluate generated labels

The product titles report raises a question that matters more than the feature itself: how do you evaluate whether generated labels are actually improving results?

CTR in isolation is not the answer. A heading written for click volume can produce high CTR and poor conversion rates. The report includes conversion metrics you choose to include, which is the right place to look. Compare the conversion rate and revenue attached to generated labels against your originals. If generated variants are consistently underperforming on conversion despite higher CTR, that is a signal to review your text guidelines or tighten your catalogue data to give the system better constraints to work within.

If generated variants are outperforming your originals on conversion, that is a sign your original labels were not fully matched to the search terms that convert.

The report is a sample, not a complete log. Use it as a directional signal rather than a definitive performance readout. Over time, as more impressions accumulate, the data will become more reliable for making decisions about whether to keep text customisation on, refine it with guidelines, or turn it off.

Performance data should also inform your feed work. If you find that generated labels consistently outperform your originals on a particular product category, that is a signal that your original titles for that category were under-optimised. Fix the source titles rather than relying on automation to compensate indefinitely. Automation works best when it has good titles to start from, not weak ones to correct.

Best practices for Shopping ads: what to do now

The immediate actions for any Shopping advertiser are straightforward.

Check whether AI Max for Shopping is active in your account. If it is, the product titles report is available and may already contain data. Review what Google has generated and compare it against your original labels. Identify any generated label that misrepresents your item, conflicts with your brand positioning, or includes promotional text you would not approve.

Set text guidelines if you have brand voice requirements, restricted terminology, or competitive positioning you want to protect. This is the most practical control layer available without disabling the feature entirely. You need to add your guidelines in campaign settings, not at the account level, so check each relevant campaign individually.

Audit your product feed titles. Since generated labels build on what is already in your data source, an enriched, attribute-dense feed gives automation better material to work with. Front-load product type and key attributes in the first 70 characters. Fill in colour, size, material, and other structured attributes. Avoid promotional text in titles since this risks disapproval regardless of whether a human or the system wrote the label. Consider whether keywords for your titles match the actual search terms your customers use, not just your internal product naming conventions.

Decide on your opt-out position. For most advertisers, text customisation with tight text guidelines is the right balance. For brands with strict creative requirements or those running in regulated product categories, the account-level opt-out via your account manager is available.

Monitor the report monthly once data starts to accumulate. The side-by-side comparison of generated labels versus original titles, combined with performance data, is one of the most transparent views Google has given advertisers into how automation is changing their creative. Use it to identify which product categories benefit most from generated labels and which need better source titles before automation can help them.

Final thoughts

AI product titles represent a significant shift in how Google Shopping ads work. For years, the title in your feed was the title in your listing. That is no longer always true. Google's system now generates alternatives in real time, based on your catalogue data and the searcher's query.

The controls exist. Text guidelines, text customisation toggles, and account-level opt-outs give advertisers meaningful options. The question is not whether to engage with this feature, but how to engage with it on terms that protect your brand while taking advantage of the improvements it can deliver on complex, long-tail search terms.

The product titles report is the starting point. Use it to understand what Google is generating, evaluate it against performance data, and adjust your text guidelines or catalogue data based on what you find. The goal is not to hand full control to automation, but to give automation good enough inputs that the labels it generates are ones you would have written yourself.

‍