Google AI Mode is a new, AI-only way to search that Google started testing in Search Labs on 5 March 2025. Instead of a page of ten blue links, you get one long written answer built by Gemini, with links to websites inside and beside it. You can then ask follow-up questions, like a chat.
If you run a website, the big question is simple: will this send you less traffic? Right now nobody outside Google knows. It is a small, opt-in test. But it shows clearly where Google Search is heading, so it is worth understanding today. In this article I cover what Google has confirmed, how AI Mode works, who can use it, and what I would do about it as a site owner.
What Google AI Mode Is
Google announced AI Mode on its official blog in a post by Robby Stein, VP of Product for Google Search. Google calls it “an early experiment” and a new Search mode with more advanced reasoning, thinking and multimodal abilities. Multimodal means it can work with more than just text, such as images or voice.
AI Mode runs on a custom version of Gemini 2.0, Google’s AI model. You type a question, and it writes an answer that pulls together information from many sources. Then you can keep asking follow-ups, and it remembers what you asked before.
Google says it is most useful for questions that need more exploring, comparing and reasoning. These are the questions that would normally take you several searches.
An Example of the Kind of Query It Is Built For
Think of a question like: “What is the difference between a heat pump and a gas boiler for a three-bedroom house in a cold climate, and which is cheaper to run?”
In normal Search, you might run three or four searches and open several tabs. AI Mode tries to answer the whole thing in one go. Then you could follow up with “What about installation costs?” without typing the full context again.
Where It Lives Right Now
AI Mode sits inside Search Labs. Labs is Google’s opt-in area for testing new Search features before (or instead of) a full launch. Not everything in Labs becomes a real product. Some tests change a lot, and some are dropped.
From what Google has shown, people with access see an AI Mode option in Search. Answers appear in a chat-style layout, with source links next to the text.
Who Can Use AI Mode in Search Labs
Access is limited for now. Google says it is starting by inviting Google One AI Premium subscribers to try it. That is Google’s paid plan that includes its more advanced Gemini features.
Other people can sign up through Search Labs, but there is no promise of when they will get in. Labs features are usually US-only at first, and this one is no different.
So for most site owners, the honest answer is that very few of your visitors can use AI Mode today. It is an early test with a small group of paying users. I would not panic about traffic from it this month.
How AI Mode Works: The “Query Fan-Out” Idea
The most useful thing in Google’s announcement, for SEO people, is how AI Mode finds its information. Google names a technique it calls “query fan-out.”
In plain words, AI Mode does not run just your one search. It breaks your question into smaller parts, runs many related searches at the same time across those parts and across different data sources, and then combines the results into one answer.
Google also says the model can make a plan, run searches, and then adjust the plan based on what it finds. It is a multi-step process, closer to how a person researches than to a single keyword lookup.
The Data Sources Google Names
According to Google, AI Mode can draw on:
- The web, using high-quality web content found through normal Search.
- The Knowledge Graph, Google’s database of facts about people, places, businesses and things.
- Real-world and fresh information, the kind of live data Google already shows in Search.
- Shopping data for billions of products.
That list matters. It tells you AI Mode is not only reading web pages. Your business data, product data and entity information all feed into what it can say about you.
Why Fan-Out Matters for Your Content
Fan-out means a single question from a user can turn into many hidden searches. Each hidden search is a chance for your page to be pulled in.
In my view, this rewards sites that cover a topic in depth, across several related pages. If someone asks a big question and Google splits it into five sub-questions, a site that clearly answers three of them has more chances than a site with one thin page. This is why I still think planning content as pillar pages and topic clusters is still one of the best uses of your time.
It also means specific, detailed questions get more weight. The sub-queries AI Mode creates are likely to look a lot like long-tail keywords: longer, more exact phrases. Pages that answer those clearly and directly are easier to use as a source.
AI Mode vs AI Overviews: What Is the Difference?
This is the question I get most, so let me keep it short. Both use Gemini. Both write AI answers with links. But they work in very different places.
| AI Overviews | AI Mode | |
|---|---|---|
| Where it shows | At the top of normal search results, for some queries | A separate mode, opened on purpose |
| Who can use it | Widely available; Google says more than a billion people use it | Opt-in Labs test, starting with Google One AI Premium subscribers |
| Page layout | AI summary plus the usual results below it | A full AI answer with links, chat style |
| Follow-up questions | No, each search starts fresh | Yes, you can keep the conversation going |
| Best for | Quick summaries on top of normal results | Longer questions that need comparing and reasoning |
The key point for site owners: with AI Overviews, the normal blue links are still right there below the summary. With AI Mode, the AI answer is the main event. Links are inside and beside the answer, not in a list below it.
The Other Changes Google Announced the Same Day
AI Mode got the headlines, but the same announcement changed AI Overviews too. These changes reach far more people today, so I would pay as much attention to them.
- Gemini 2.0 in AI Overviews. In the US, AI Overviews now use Gemini 2.0 to help with harder questions. Google says it is starting with coding, advanced maths and multimodal queries.
- More AI Overviews for those queries. Google says AI Overviews will show more often for these kinds of searches, with faster, better answers.
- Teens can now see them. AI Overviews are now open to teenage users.
- No sign-in needed. People no longer have to be signed in to a Google account to get AI Overviews.
If your site covers coding, maths or technical how-to topics, expect to see AI Overviews on more of your keywords. I would start checking those searches by hand now, so you know what the results page looks like before you see a change in your numbers.
What Google Says About Links and Websites
Google is clearly aware that site owners are nervous. In the announcement, it says that helping people discover content from the web “remains central” to its approach. It says AI Mode is designed to give access to web content, with links for people to read more.
Google also says AI Mode is built on its core quality and ranking systems. And when it does not have high confidence that an AI answer is helpful and high quality, it shows a set of normal web search results instead.
That last point is worth remembering. AI Mode is not meant to answer everything with AI. For some queries, the fallback is regular search results, which means regular SEO still decides who shows up.
Known Limits Google Admits
Google is open that this is early. It says AI Mode will not always get things right. It also says answers may sometimes seem to take on a persona or reflect a particular opinion, which Google says is not its intent.
Google says it plans to add more visual answers with images and video, richer formatting, and new ways to get to helpful web content. It also says it will make quick changes based on tester feedback. So whatever AI Mode looks like this week, expect it to look different soon.
What Google AI Mode Means for SEO and Site Traffic
Here is my honest take. The test itself is too small to move your traffic today. What matters is the direction. Google has now shown a version of Search where the AI answer comes first and links come second.
That is the same direction we have seen elsewhere. When ChatGPT search opened to everyone without a login, it pushed the idea of chat-style search into the mainstream. Google testing AI Mode tells me it sees the same shift and wants to own it.
Questions Nobody Can Answer Yet
Because AI Mode is so new, a lot is still unknown. I would be careful with anyone who claims to know these answers:
- How many clicks AI Mode sends. There is no public data on click rates from AI Mode answers.
- How it will show in Search Console. Google has not said whether, or how, AI Mode clicks and impressions will be reported.
- Whether it will ever leave Labs. Google calls it an experiment. It could grow, change or be dropped.
- How sources are picked. Google says it uses its core ranking and quality systems, but it has not shared details on how pages are chosen as links in an answer.
What Does Not Change
AI Mode is part of Google Search. It uses the same index and the same ranking systems. So the basics still apply:
- Your pages need to be crawlable and indexed. If Google cannot find a page, AI Mode cannot use it.
- Helpful, accurate, people-first content is still what Google says it wants.
- Clear structure helps. Headings, short answers and well-labelled sections make it easy to pull a fact out of your page.
In my experience, sites that panic about every new AI feature tend to make things worse. They rewrite good pages, chase tricks or pump out AI text. If you are thinking of using AI to write more content to “feed” AI Mode, read what Google’s rules say about AI-generated content first. Scale without quality is not a strategy.
What I Would Do Now as a Site Owner
You do not need a big plan for a Labs test. But a few small steps now will put you in a good place if AI Mode grows.
1. Answer the Full Question, Not Just the Keyword
AI Mode is built for long, multi-part questions. Look at your key pages and ask: does this page answer the obvious follow-up questions too? If a page on “how much does a new roof cost” never mentions roof size, materials or time, it is missing the sub-questions fan-out is likely to look for.
2. Put Clear Answers Near the Top of Each Section
I like a simple pattern. Each H2 or H3 is a question or topic, and the first one or two sentences under it give the direct answer. The detail comes after. This helps people who skim, and it makes it easy for any system, AI or not, to find the answer.
3. Keep Your Facts Consistent Everywhere
AI Mode draws on the Knowledge Graph and shopping data, not just your web pages. Make sure your business name, address, hours, prices and product details match on your site, your Google Business Profile and your Merchant Center feed. Mixed signals make it harder for Google to trust any one version.
4. Use Structured Data Where It Fits
Structured data (code that labels the facts on a page for search engines) helps Google understand what a page is about. Product, organization and local business markup are the ones I check first. Google has not said AI Mode uses it directly, but it feeds the systems AI Mode draws from.
5. Show Who You Are and Why You Know the Topic
Add real author names, short bios and clear “about” information. When Google is choosing which sources to trust for a long answer, signs of real experience are likely to help, and they help human readers either way.
6. Do Not Block Google by Mistake
Some sites have added rules to block AI crawlers. Be careful here. Google’s crawler documentation says the Google-Extended token does not affect whether your pages appear in Google Search. AI Mode is a Search feature, so blocking Googlebot itself is what would remove you, and that would remove you from normal results too.
7. Mark the Date in Your Reports
Add a note for 5 March 2025 in your analytics and reporting. If AI Mode expands later and your traffic shifts, you will want to know when the test started. I do this for every major Google change, because months later nobody remembers the dates.
How to Test AI Mode With Your Own Topics
If you get access through Labs, use it as a research tool for your own site. This is the process I would follow:
- List five to ten real questions your customers ask, written the way they would say them, not as short keywords.
- Run each one in AI Mode and note which websites are linked in the answer.
- Ask two or three follow-up questions on each, and note whether the sources change.
- Run the same questions in normal Search and compare the sources with your usual top results.
- Look for gaps. If the answer covers a sub-topic your site does not, that is a content idea.
- Check the answer for errors about your brand, products or prices, and fix any unclear or outdated facts on your own pages.
- Repeat every month or so, because Google says it will change the experience quickly.
If you do not have access, you can still do steps 1, 4 and 5. Thinking in full questions and sub-questions is useful for normal SEO too.
Conclusion
Google AI Mode is an early, opt-in test in Search Labs. It uses a custom Gemini 2.0 model and a “query fan-out” method to answer long questions with one AI answer and a set of links. Right now only a small group of Google One AI Premium subscribers can use it, so it will not change your traffic overnight.
What it does change is the picture of where Search is going. My advice is to stay calm, keep doing solid SEO, and make your pages the clearest answer to the full question, not just the keyword. For the detail, read Google’s own announcement of AI Mode and the AI Overviews update. Then pick your three most important pages and check whether they answer the follow-up questions a customer would ask next.
