Hi everyone, this is Neo.

If you’re still agonizing over what your ad copy should say in the AI Mode era, let me save you the trouble: nobody writes ads in Google AI Mode. Your product data feed IS the ad.

That’s not me being dramatic — it’s the direct conclusion of the new ad system Google announced at Marketing Live, paired with a deep-dive SEJ piece on August 26. Today I’ll walk through the new AI Mode ad formats, the logic shift underneath them, and what independent ecommerce sellers should do right now.

What AI Mode ads look like: two new formats

At Marketing Live, Google announced it’s testing two Gemini-powered ad formats in AI Mode:

1. Conversational Discovery Ads

A shopper describes a problem conversationally (“I want my house to smell like a fancy spa, or a rainy forest”). Gemini writes a short explainer connecting your product to their exact question. The ad carries a Sponsored label and reads less like a banner, more like part of the conversation.

2. Highlighted Answers

When AI Mode returns a recommendation set (say, “best language apps for an upcoming trip”), your product can appear inside that set, alongside the organic suggestions — again Sponsored-labeled, with a Gemini-written note on why it fits.

Both formats share a signature element: an independent AI explainer. Gemini evaluates and synthesizes information about your product, then displays that context alongside your creative, so shoppers understand why this was recommended.

Two related formats run on traditional Search instead (not in AI Mode):

• AI-powered Shopping Ads: built for considered purchases (fridges, TVs, espresso machines). When someone researches one, Gemini writes a custom explainer for your product against that exact query.
• Business Agent for Leads: a chat agent embedded directly in the ad, replacing the static lead form. More relevant for services and considered-purchase businesses than straight ecommerce.

One stat from Google worth holding onto: 75% of people report making faster, more confident decisions using AI Mode in Search.

The core shift: the feed is the creative

This is the most disruptive part of the whole system, so read it twice:

Ads in AI Mode aren’t written — they’re generated. Gemini writes your ad copy on the fly from your product feed. That means the factor deciding whether you show up is no longer your creative team. It’s the quality of your product data.

Here’s the difference in practice — a title that says “Blue Shirt” versus one that says “Men’s Blue Oxford Slim Fit Shirt, 100% Cotton.” A description that says “high quality for every need” versus one that names the material, the use case, the fit, and what it’s not right for. Vague data produces vague ads. Rich, specific data gives Gemini something to write with.

There’s also a query-shape change behind this: searches in AI Mode run about three times longer than traditional searches. Nobody types “running shoes” into a conversation. They type “a neutral running shoe with extra cushion for a heavy runner under 150 dollars that does not squeak.” Long, spoken-language queries like that are exactly what generative ads are built for.

To support the system, Google added a set of optional conversational attributes to Merchant Center, built specifically for AI surfaces — the product data spec explicitly notes they’re primarily intended for conversational experiences like AI Mode. Fill them in if you can; these fields are the raw material for AI Mode ads.

One operational note: the new formats require your campaigns to run on Performance Max or AI Max.

What independent sellers should do: a 5-point checklist

1. Treat product data as a creative asset. From today: titles should be specific (brand + category + fit + material), descriptions parameterized (materials, use cases, applicable scenarios, what it’s NOT for), images and attributes complete. This isn’t “feed optimization” — this is writing your ads.
2. Fill the conversational attributes in Merchant Center. They’re AI-Mode-specific material, and most competitors haven’t touched them yet. First in, first positioned.
3. Sort out feed ownership. In traditional paid media the feed belonged to the ads team. Now the feed decides ad survival — product, operations, and engineering all need a hand in it. Whoever owns product data owns AI Mode ads.
4. Don’t port classic search logic into AI Mode. Traditional search runs on a keyword-ad-landing-page funnel; AI Mode runs on data-generation-explanation. The same asset library serves both, but the strategy is separate.
5. Switch your A/B testing target from ad copy to product data. Test different product title/description versions and see which ones generate stronger AI output. Data assets are iterable — treat them that way.

How this connects to UCP: two parallel tracks

One thing worth clarifying: AI Mode ads and UCP (Universal Commerce Protocol, the agentic shopping protocol) are two different tracks.

• AI Mode ads: a human sees the ad, makes the decision, and checks out — the only change is that “creative” became “generated.”
• UCP / agentic commerce: an AI agent completes research and purchase on the shopper’s behalf, possibly without the human ever viewing a page.

Both tracks are advancing at once. AI Mode ads are the transitional form — research happens in conversation, but the transaction stays human. UCP is the dress rehearsal for the end state. Either way, both tracks feed on the same product data. So making your data thick and complete is the one investment that can’t be wasted.

Neo’s take

Once Google industrializes ad creative, the battleground shifts from “creative” to “data assets.” And for small independent sites, that’s actually good news:

In the old game, beating big brands in search ads required a top-tier creative team and a huge budget for testing variations — that’s money talking. AI Mode ads flatten the creative layer: Gemini writes roughly equal copy for everyone. What separates you now is whether your product data is specific, complete, and persuasive. It’s a data-asset race, not a budget race.

But three realities before you go all in:

1. AI Mode ads are still in testing. Don’t expect volume yet. Getting in now is positioning, not harvesting. The real payoff comes when AI Mode’s user base scales.
2. Feed data governance is a “do it now” job. Whether it’s ads, organic recommendations, or UCP agent shopping, everything reads the same product data. Every day you delay is a day you hand your competitors.
3. Don’t be spooked by “AI writes copy,” and don’t buy the “creative is dead” line either. Ad copy isn’t disappearing — it’s moving from human production to machine generation with human feeding. What you feed determines what gets generated. Product knowledge, customer insight, selling-point distillation — still human work, and now it directly determines ad quality.

One line to remember: in the AI Mode ad era, your product data is your ad agency. The more professional your feed, the harder your ads punch.