AI Mode is Google's conversational search mode: you ask a question in plain language, Google splits it into sub-questions behind the scenes, searches for those, and writes one answer with sources and, for product questions, with products in it. For online stores that is not a cosmetic change to the results page. It changes what a 'search' is, which pages Google picks for it and what role your product data plays. I am writing this from what we see at Traze in the engines we track and in conversations with store owners, and I say where I do not know yet.
What AI Mode is, and what it is not
AI Mode is a separate tab in Google Search, next to 'All', where the answer takes centre stage and the classic list of links fades into the background. It works differently from an AI Overview: an overview is a summary above the regular results, AI Mode is a conversation in which you can keep asking. Under the hood, Google uses a technique it calls 'query fan-out': one question is fanned out into several searches whose results together form the answer.
What AI Mode is not: a replacement for the search index. Everything it shows comes from pages Google already knows and from product data Google already has. There is no separate 'AI index' to sign up for. Whoever is not indexable is invisible here too; whoever is indexable now has to answer a second question: is my page the best source for a sub-question?
What I see changing
In the questions we track for online stores I see three shifts, and they are connected.
The first is that the question gets richer. Where someone types 'garden chair' into the classic search bar and clicks through filters, that same person types a sentence with requirements into AI Mode: material, budget, delivery time, use. Google answers that sentence with a selection and an explanation. The pages used for that are not the pages with the highest volume on 'garden chair', but the pages that describe the combination of requirements best.
The second is that the answer takes over the comparison work. Where a visitor used to open five stores, AI Mode makes the first selection and then shows the stores where that selection is for sale. The click comes later in the process and with more intent. Fewer visits, better visits, and a greater dependence on whether you are in the selection at all.
The third is that product data feeds the answer. Price, stock, variants, ratings and availability in the answer come from the same source as Google Shopping: Merchant Center. A store without a feed, or with a feed full of missing attributes, is simply less complete in product answers than a competitor with a well-kept feed.
Why product data is the new battleground
For years the product feed was something for the advertising team. That separation does not hold. When Google recommends a product in a conversation, it takes the facts about that product from structured data, not from marketing copy. The title, the GTIN, the material, the size, the colour, the price, the delivery time: every attribute is a property the answer can select on. An attribute that is missing is a filter you drop out of.
That is good news too. Improving feed data is concrete, measurable and within your own control, unlike the question of which page Google cites. The order I recommend: first fix disapprovals and missing required attributes, then the optional attributes that make the difference in your category, then rewrite titles and descriptions so they answer the question instead of only naming the product. You can check your feed for free with our Shopping Feed Tester.
The category page gets a second life
There is a widespread idea that AI search makes category pages redundant. I see the opposite. A well-made category page is exactly what a sub-question needs: a bounded topic, a selection, and — if you do it right — an explanation of what to look for. A page 'garden chairs for a small balcony' with a short, factual introduction about dimensions, foldability and materials, plus the products, is a better source for AI Mode than a product page and a better source than a blog post.
The condition is that the page actually says something. A category page with only a grid and a hidden text block at the bottom stuffed with keywords answers nothing. The explanation has to be at the top, short, and answer the questions a customer really has, in the HTML and not only after a script runs. That is the same demand GEO makes of every page, only now for the pages most directly tied to revenue.
What I would advise online stores to do now
Five things, in this order. One: get your feed in order and treat it as product content, not as an advertising attachment. Two: pick your ten most important categories and write a real introduction for each that answers customers' choice questions, at the top and server-side rendered. Three: create question-specific sub-categories or guides for the combinations customers actually name — not a hundred variants, but the ten that matter. Four: check your technology: rendering, speed and whether filters are creating thousands of duplicate URLs. Five: start measuring whether you appear in AI answers, per question and per engine, with a fixed set of questions; the free AI visibility check gives a first impression.
What I do not know yet
A few things I cannot back up, and I do not want to pretend otherwise. I do not know how large AI Mode's share of total search behaviour will become, or how fast; the rollout differs per country and Google does not share numbers about it that I can verify. I do not know how stable the source selection is: the selection varies from run to run, and whether that converges over time to a fixed set of 'trusted' stores is still open. And I do not know how Google will eventually work ads into the conversation and what that does to the space for organic sources. What we can do is measure — and that is exactly why we track mentions per engine instead of waiting for Search Console to tell us.
Conclusion: prepare for questions, not keywords
AI Mode forces online stores back to the customer's question: what does someone want, with which requirements, and which page answers that best? Feeds, category pages and technology are the three places where that work happens, and measurement is how you know whether it works. If you want this done structurally — research, writing, publishing into your store and measuring per engine — see how our E-commerce SEO service works.