Google introduced AI Mode in Labs in Mar 2025.
In May 2025, they officially rolled it out to all US users, and will continue to implementing it in other countries.
What is AI Mode?
AI Mode is a new, conversational way to use Google Search. It’s currently powered by Gemini 2.5.
AI Mode includes citations and you can ask it follow-up questions—when we first tried it, it felt like AI Overviews when Google first launched it.
Of course, the responses are much better now.
It’s likely Google’s way of transitioning to AI-powered search, and eventually phasing out AI Overviews.
To access AI Mode, just head to Google Search and you’ll see it sitting on the navigation bar (only in the US and India for now).
What you can do with AI Mode
AI mode lets you ask complex, multi-part questions, and what’s amazing is that it uses a query fan-out technique to respond.
The query fan-out technique issues multiple related searches simultaneously and brings back a consolidated, well thought-out answer.
Some of these searches are direct, others are inferred.
For example, when you search for “Best washing machine for a family of 4”, it doesn’t treat it as one long query.
Instead, AI Mode breaks it into multiple sub-queries like:
- Best washing machines 2025
- Washing machine size for family of 4
- Front load vs top load washing machine
- Customer reviews: washing machines for families
Then, it combines and organizes the information before sharing recommendations with you.
How does AI Mode work?
We’ll use the query “Best washing machine for a family of 4” as an example throughout the process.
1. When a user enters a query, Google analyzes the query to extract:
- Intent (informational, comparative, commercial, etc.)
- Entities (products, attributes, constraints like price)
- Session context (previous queries, clicks, memory)
In this example, the user is looking to buy a washing machine — so the intent is clearly commercial, and may become transactional later.
Entities will be things like “family of 4” (usage constraint) and “washing machine” (product type). “Best” acts as a modifier, suggesting a comparison. AI can also infer implicit attributes such as energy efficiency, quiet operation, or reliability. It may assume factors like load type, capacity, or even brand preferences — especially if those signals appear in session history.
AI Mode also factors in contextual signals, such as pages the user previously clicked on. Like product reviews or a blog post on fitting a machine machine into a small space.
It may also look at broader shopping behavior, such as engagement with similar product categories, past product views, or related searches. Even device type and location can play a role.
There’re more contextual signals Google may use. You can read more here.
2. Then it triggers a query fan-out
This is where the system creates multiple subqueries to help answer the original question more thoroughly.
For instance, someone new to shopping for a washing machine might start with:
“Best washing machine for a family of 4.”
Through the course of their journey, they’ll learn about front-load vs top-load, energy-saving features, or noise levels. Eventually, their searches might evolve into:
- “Best water-saving washing machines”
- “Washing machine size for family of 4”
- “Front-load vs top-load for large households”
- “LG vs Samsung washers for quiet performance”
Query fan-out covers discovery process. AI Mode generates and processes these subqueries behind the scenes. So the user gets the answer without having to do multiple searches.
3. AI Mode retrieves documents and builds a custom corpus
At this stage, Google retrieves a set of documents from Search and other internal sources.
From these, it extracts the most relevant segments that align with the subqueries and user intent.
This becomes a custom corpus, which may include:
- Review snippets
- Product specifications
- Blog recommendations
- Forum posts
- Previously viewed or interacted-with content
4. Google reranks and filters documents
Once the custom corpus is assembled, Google doesn’t just take everything at face value.
It scores and ranks each document or based on a combination of relevance, quality, and contextual fit. It’ll look at factors like:
- Semantic match to the query
- How well the passage aligns with the original query and its subqueries — not just by keywords, but by meaning.
- Factual density
- Whether the passage contains concrete, useful facts rather than generic fluff.
- Recency and trustworthiness
- Preference for up-to-date sources and domains that have historically been reliable or authoritative.
- Diversity of perspectives
- Google may surface multiple viewpoints, especially for queries where opinion or nuance matters (e.g., “best” products, comparisons, or controversial topics).
Only the top ones move to the synthesis stage.
5. Google routes the query to the right model for the job
Finally, Google determines what kind of response is needed to answer “Best washing machine for a family of 4.”
In this case, the user is likely looking to evaluate options, weigh pros and cons, and ultimately make a purchase decision. So Google routes the task to a model that can:
- Compare different washing machines side by side
- Explain which product specs, reviews, and fit matter for a family of four
- Highlight trade-offs (e.g. price vs. capacity, top-load vs. front-load)
After all these are done, it comes up with an answer like this
What does AI Mode mean for ecommerce brands
AI Mode marks a shift in how shoppers discover and evaluate products online. Instead of relying solely on traditional search rankings, Google’s LLMs are now selecting content based on how well it answers a constellation of subqueries — things like comparisons, use cases, specifications, and real user insights.
For ecommerce brands, this means your product pages, data, and content need to do more than just okay — they need to be complete, structured, and semantically rich. Brands that don’t adapt may find themselves less visible in AI-generated results, even if they ranked well in traditional search.
We'll talk about how you optimize for AI Mode in a later part of this article.
What are AI Overviews?
AI Overviews are Google’s AI-generated summaries that appear above your ten blue links when it’s confident the response adds value and is factually accurate.
They’re powered by Gemini and include clickable citations that link to source pages.
Overviews are separate from AI Mode. But it’s likely that AI Overviews are just a lightweight preview of Google’s broader shift toward fully AI-powered search.
AI Overviews are driving more zero-click searches and eating up your traffic
As AI Overviews provide direct answers at the top of search results, users are increasingly getting what they need without clicking through to websites.
That results in many zero-click searches.
Websites relying on SEO are already seeing their traffic fall significantly, with media sites taking the hardest hit.
Clicks that do happen tend to be of higher quality, as Google suggests. Although we can’t be sure since it’s currently not labeled separately.
What’s the difference between AI Mode and AI Overviews?
1. They serve different purposes:
AI Mode lets you explore a topic conversationally, handling complex follow-up questions and branching prompts.
AI Overviews deliver a quick cited answer together with its sources, but there’s no way to ask a follow-up question.
2. AI Mode personalizes responses while AI Overviews don’t
AI Mode adapts its replies with rich context—ongoing dialogue, prior queries, saved preferences, and optional signals such as location, calendar events, or pages you’ve already viewed (when you’ve granted permission).
AI Overviews are mostly static—summaries are generated based on the current query alone, with little or no personalization.
3. AI Mode gives more in-depth responses
AI Mode performs deeper reasoning using multiple subqueries (query fan-out), builds a custom set of relevant passages, and synthesizes a more comprehensive answer.
AI Overviews use a lighter model that summarizes passages from a narrower set of sources, optimized for speed and confidence rather than depth.
How does AI Mode compare with ChatGPT?
AI Mode and ChatGPT both offer a chat-based experience. They let users ask follow-up questions, refine answers, and explore topics interactively.
But they differ in a few ways.
1. AI Mode is search first, ChatGPT is general purpose
AI Mode is built into Google Search. It’s designed for search-related tasks like research, comparison, shopping, and planning — where up-to-date web content is central.
ChatGPT is a general-purpose AI assistant built for flexible, open-ended tasks like writing, coding, summarizing, and brainstorming — not limited to search.
2. AI Mode always has web access, ChatGPT only does with browsing enabled
AI Mode is always connected to the web. It uses Google Search to retrieve current information and generate answers from live sources.
ChatGPT doesn’t access the internet by default — unless you’re using GPT-4 with web browsing enabled. Otherwise, it replies based on its training data, which may be out of date.
3. AI Mode uses real-world context, ChatGPT doesn’t by default
AI Mode can personalize responses using signals like your location, search history, calendar, and even emails — if you’ve opted in.
ChatGPT doesn’t access real-world data unless explicitly connected. It can personalize answers based on your instructions or memory (if turned on), but it doesn’t integrate with other apps or services.
How can ecommerce brands optimize for AI Mode?
Like what John Mueller wrote in his article, the core principles apply:
- Provide a great on-site experience
- Create unique, valuable content for people
- Make sure LLMs can access your content
Today, we’re not just optimizing for human users anymore. Agents are now shopping too. That means we need to think about agent experience, not just user experience.
Here’s how ecommerce brands can improve their chances of being selected by AI Mode — and influence when and how they show up.
Create detail-rich product pages
Earlier, we talked about the query fan-out technique AI Mode uses.
These subqueries might include comparisons, use cases, alternatives, “best for” scenarios, or technical specifications.
To show up for these, your product descriptions can’t just rely on fluffy or feature-only language. They need to reflect the actual language, needs, and decision points of your target audience.
For example, instead of saying “Made with premium memory foam,” try “Ideal for side sleepers who need extra shoulder support.”
This introduces an entity (“side sleepers”) and a real-world use case — helping AI Mode match your content to subqueries like “best pillow for side sleepers” or “neck support for sleeping on your side.”
Consider adding structured sections on your product detail pages such as “Ideal for”, “How to use”, or “What makes it different” on your product detail pages to cover more query angles.
Hypotenuse AI’s bulk product description writer can help generate this at scale.
Include longer-form content like blog articles
Product pages are great for direct purchase intent. But blog content helps you capture a broader range of queries more naturally.
These might include questions like:
- “What’s the best material for summer shirts?”
- “How to choose the right desk size for a home office”
- “What makes organic skincare different?”.
Creating long-form content lets you go deeper into these subqueries, build topical authority, and guide users earlier in their journey.
It’s also a chance to showcase expertise and trustworthiness.
Ensure your product data is complete and accurate
Your product data is the foundation for everything — from product descriptions to spec tables to how AI interprets your catalog.Without complete and accurate product data, your content team can’t write descriptions that truly resonate. If they don’t know a shirt is made of 100% cotton, they won’t be able to highlight that it’s breathable and perfect for summer.
Inaccurate or inconsistent specs — across your website, marketplaces, or feeds — can also hurt your chances of showing up. AI Mode may recommend your product in the wrong context, or skip it altogether due to low confidence.
Especially with AI Mode drawing from the Google Shopping Graph, your Google Merchant Center listings need to be optimized. Make sure all required attributes are filled — and wherever possible, complete optional ones like material, pattern, size, or intended use. The more structured and consistent your product data, the more discoverable and trustworthy it becomes to AI.
Use Hypotenuse AI’s product data enrichment feature to fill in missing attributes, correct inaccuracies, and standardize data across every channel.
Prioritize on-page experience
As AI generates more zero-click searches, fewer users may land on your site — but the ones who do are usually better qualified and more ready to buy.
Make their visit count.
Ensure smooth navigation, clear categorization, and intuitive search. Products should be in the right categories and tagged properly (Hypotenuse AI’s product categorization tool can help here).
Your website should feel clean, clutter-free, and responsive across devices. The easier it is for both users and AI to interpret your content, the higher your chances of being surfaced.
Get the technical foundations right
AI agents now play an active role in evaluating and recommending products. If your site isn’t technically accessible, your content won’t even be considered.
Start by ensuring that all important pages are:
- Indexable
- Fast-loading
- Mobile-friendly
- Marked up with clean metadata and structured data
These technical fundamentals are what allow AI Mode to crawl, retrieve, and understand your content.
Support multimodal search with high-quality visuals
Search isn’t limited to text anymore. Shoppers now search using images, especially on mobile.
To stay visible, your product imagery needs to be high-quality: sharp resolution, clear details, and a variety of angles and backgrounds that reflect real use cases.
With Hypotenuse AI’s image editing tool, you can easily generate and edit visuals in bulk — adding lifestyle backgrounds, upscal low-res images, add shadows, and other enhancements that make your catalog stand out in multimodal search.
Strengthen brand mentions and associations
AI relies on entity recognition and external signals to associate your brand with certain topics or product categories.
To influence those associations, aim to get mentioned on reputable websites and blogs that do in-depth product reviews, comparisons and round-ups.
Encourage reviews on forums like Reddit or industry-specific communities where your target audience hang out.
You can shape the narrative by consistently showing up in the right context — for example, being mentioned as a top pick for “eco-friendly kitchen appliances” or “workwear for hot climates.”
Over time, these signals help position your brand for specific topics — boosting your chances of being surfaced for relevant queries.
Final thoughts
While other AI search engines have already been pulling people away from Google, AI Mode is fundamentally changing how ecommerce brands get discovered.
Optimizing is no longer just about ranking for keywords — it’s about creating structured, rich, and context-aware content that both humans and AI agents can easily understand. So you consistently show up for the right queries, in the right context.
Things are evolving quickly. And for ecommerce brands managing thousands or even millions of products, traditional manual processes simply can’t keep up.
Hypotenuse AI helps brands stay ahead by enriching product data and generating AI-optimized content and visuals that align with your brand guidelines — 10x faster.
If you’re keen to hear more, reach out to us here.
