Google Ads Used to Reward the Best Bid. Now It Rewards the Best Data. Most Brands Haven't Noticed Yet.

Something I've been observing with a few brands I consult for. The performance team is optimising ad copy, tweaking bids, and running creative tests. They are doing everything that's worked for the last decade. 

Meanwhile the AI is generating product responses from a Google Merchant Centre feed that hasn't been properly audited since the last platform migration.

The gap between those two activities is where a significant amount of money is about to disappear.

Google's conversational ad formats aren't just a new placement. They change what the ad unit actually is. In the old model, an ad sat next to a search result and competed for a click. In the new model, the ad is generated from your product data in response to a specific conversational query. A visitor tells an AI they need a running shoe that handles wet trails, doesn't feel heavy over marathon distance, and comes in wide fit. Google's AI reads your assets and constructs a response. The ad isn't adjacent to the answer. It is the answer or it isn't there at all. You choose between visibility or obscurity.

That's a structural change, not a feature update.

What it means practically is that the inputs determining whether you appear have shifted. Keyword bids still matter, but they're not the primary filter anymore. The primary filter is whether your product data is complete and specific enough for the AI to construct a confident recommendation.

In the old model, you bid on a keyword and wrote a headline. The ad's job was to get the click. In the new model, the AI is reading your product assets and generating a response. The quality of that response depends entirely on what you've given it to work with. Not generic ad copy. Not a brand tagline. Actual product content: who this product is for, what problem it solves specifically, how it differs from the alternatives, what the proof points are, what the trade-offs are, what it costs and in what contexts that cost makes sense. Additionally, who it isn't for. 

That last one matters more than most brands expect. An AI recommending products to a user with a specific, nuanced query will deprioritise a brand whose content makes everything sound like it's for everyone.

If that information isn't in your feed, schema, structured data, or product copy it isn't in the AI's response.

I've been writing about this across the series under the term ‘Agentic Trust Layer’. The infrastructure that determines how confidently an AI agent can recommend your brand and your products. It sits across three things: 

• Entity signals that tell the AI who you are and that you're the legitimate source
• Structured data and schema that make your product attributes machine-readable
• Product feed completeness that gives the AI specific enough information to cite accurately.

In the context of organic AI shopping discovery, a weak Agentic Trust Layer means you're either not recommended or recommended with less specificity than a competitor. In the context of conversational ad formats, it means the same thing, just with budget behind the absence.

Brands that have invested in building the Trust Layer will get significantly more from this format. The gap between those two groups will widen quickly.

The agentic checkout piece tends to get less attention, but it matters more over the medium term. Google is building toward a model where a user compares products across multiple brands through a Gemini conversation, adds to a unified cart, and completes payment without visiting any brand's website. If that becomes a meaningful purchase path, and I'd expect it to, then the last mile of conversion is no longer on your site. Your PDP, your checkout flow, your trust signals are invisible. The decision has already been made in the conversation, based entirely on what the AI could verify about you from the outside.

That's when the Agentic Trust Layer stops being a nice-to-have and becomes the only thing that matters. The brands that appear in that flow are the ones whose entity signals are coherent, whose product data is complete, and whose feed attributes are specific enough that the AI can recommend them without needing to second guess anything. The brands that don't appear haven't necessarily done anything wrong with their ads. They just haven't built the infrastructure the AI needs to trust them.

Most ecommerce teams aren't thinking about that yet, which is understandable. It's not a problem at meaningful scale today. But the work required to appear in agentic checkout is exactly the same work required to appear in AI shopping agent recommendations now. It compounds. The brands building it today are building a structural advantage that becomes harder to close the longer it's left.

The practical starting point hasn't changed. Audit your GMC feed. Check your GTIN coverage. Make sure your product attributes like sizing, materials, weights, product styling, etc are specific enough that an AI can reconstruct an accurate recommendation from them without guessing. Then look at your Organisation schema, your sameAs links, your entity coherence across platforms. That's the Trust Layer. It's not glamorous work and it rarely sits clearly in anyone's job description, which is exactly why most brands haven't done it.

The brands caught out by AI Overviews said they didn't see it coming. The brands that will be caught out by conversational ad formats are saying the same thing now. The data foundations required are identical. The window to build them quietly, before it matters, is still open but not indefinitely.