Transcript
How Google AI Mode Really Chooses Which Sites to Show (and What You Can Do About It)
If you’ve played with Google’s new AI mode lately, you’ve probably noticed something: the way links are surfaced feels different from classic Google Search, but not completely different. Under the hood, the fundamentals are still about relevance, authority, and user intent—but they’re being interpreted through a conversational, large language model (LLM) lens.
A common question right now is: how does Google AI mode decide which websites to recommend in a chat? And more importantly, what can brands do to be in that short list of links?
Let’s break down what’s really going on—and what to focus on next.
From Blue Links to Chat Answers (With Ads on the Way)
We’re living through a transition:
- Old model: Type a query into Google Search, get a page of links.
- New model: Ask Google AI a question, get a conversational answer with a small set of recommended sites.
Google has even confirmed that ads will appear inside these AI chats, just as many marketers suspected. So, we’re moving from “ten blue links and some ads” to AI-generated answers with embedded recommendations and relevant ads.
Despite the new interface, the underlying goal hasn’t changed:
Help the user get the best answer and a good experience as quickly as possible.
What has changed is how those answers are assembled.
Why Early AI Overviews Felt So Wrong
Cast your mind back to the early days of Bard, early Gemini, and AI Overviews.
- AI Overviews would sometimes recommend sites that didn’t even appear on page one for the same query in classic Google Search.
- The experience often felt confusing and inconsistent.
- For many searches, the AI layer looked disconnected from what SEOs and marketers saw in organic rankings.
That’s happening less and less now.
With Gemini Pro embedded into AI mode, the sites recommended in AI answers are increasingly aligned with what would appear organically for similar searches. They’re not identical—because the queries and conversational flow are different—but there’s now far more overlap and consistency.
Navigational vs Discovery: Why Brand Searches Behave Differently
Most Google searches are still brand searches. People type things like:
- “jb hifi”
- “nike”
- “xyz bank login”
In classic search, that’s basically using Google as a navigation tool. Type the brand, click the homepage, done.
In AI mode, that doesn’t work as cleanly.
Ask an AI assistant for a brand and it will often:
- Explain who the brand is
- Give you background information
- Summarise what they do
…when all you wanted was a big, obvious link to the site.
That mismatch is why chat-based search is great for research and product discovery (“Which 3D printer suits my needs if I care about X, Y, and Z?”), but still clunky for simple navigational tasks (“Just take me to JB Hi-Fi, please”).
Don’t panic if your brand doesn’t show up perfectly in AI mode for simple navigational prompts yet—that’s still being figured out. Focus instead on where AI mode shines: exploratory, research-heavy buying journeys.
Search Rankings vs AI Recommendations: Same Ingredients, Different Recipe
Classic Google Search and AI mode share some ingredients, but the recipe is different.
Traditional search rankings are still driven by things like:
- Keyword relevance
- Content quality and depth
- Authority and expertise signals
- Overall user experience (speed, usability, etc.)
LLM-powered AI mode is doing something more ambitious. It tries to:
- Understand the exact question being asked in context
- Pull in information that directly answers that question
- Weave data from multiple sources into a coherent answer
- Then suggest a few links that seem particularly helpful
Instead of just ranking pages on one static query, the AI is effectively “reading” the web and your content, deciding:
“Which sites best answer this evolving conversation?”
That’s why structure and clarity of information matter more than ever.
Structured Data and FAQs: The New Power Tools for AI Visibility
Many brands avoided implementing certain types of structured content in the past because “the juice wasn’t worth the squeeze.”
- Adding FAQs to every product page used to be time-consuming and hard to justify.
- Experimenting with additional structured data types felt like overkill if they didn’t drive clear Google rich results.
That calculus has changed.
Now, structured content and FAQs are incredibly valuable because:
- They signpost relationships between concepts clearly.
- They make it much easier for large language models to understand what your page covers.
- They help AI mode connect specific user questions with specific parts of your site.
Many teams are now using internal tools or apps to generate FAQs at scale, and implementing richer structured data beyond the bare minimum that “helps SEO.” This isn’t just for Google’s crawler anymore—it’s for any AI assistant trying to understand your site.
If you haven’t already, this is a prime time to:
- Add well-written FAQs to key product and service pages.
- Implement appropriate schema types (FAQ, Product, Organisation, Article, etc.).
- Ensure your content hierarchy (headings, sections, labels) is crystal clear.
Brand Is Still a Signal: Authority and Freshness in an AI World
One thing hasn’t changed: brand matters—a lot.
Your brand still acts as:
- An authority signal
- Are people searching for your brand by name?
- Are there credible sites referencing and citing your brand?
- A freshness signal
- Are there recent mentions of your brand online?
- Is Google seeing ongoing activity, news, or conversation around you?
This is why you’ll still hear people pushing backlinks—but often, the loudest voices are the ones selling them.
Instead of dumping budget into random link building:
- Put that money into better structured data and content clarity.
- Invest in PR and thought leadership that creates real citations and mentions.
- Use “newsjacking”—tying your brand to current news stories in a relevant way.
Newsjacking (a term popularised by David Meerman Scott) is about inserting your brand into timely conversations where you can add genuine expertise. Those mentions help build the kind of real-world authority that both humans and AI models pay attention to.
Make Customer-Critical Info Easy to Find (for Humans and AI)
For e-commerce especially, a lot of brands still bury the information customers care about:
- Shipping costs and timeframes
- Delays or special conditions
- Returns and refund policies
Historically, you might not have worried about those details for SEO. But now:
- Shoppers are asking AI assistants detailed questions like
“Which store can get this to me fastest?” or
“Where can I buy this with a flexible returns policy?” - AI mode needs to know where that information lives on your site and how it relates to a specific product or service.
If that data is hidden in a vague PDF or buried five clicks deep, it’s far less likely to be used in answers.
Action items:
- Put shipping, returns, and key service info in clear, crawlable text on important pages.
- Link that information logically to the products and categories it applies to.
- Make it easy for both humans and AI to map your content to the questions buyers are asking.
You’re not just optimising for “Google Search” anymore—you’re optimising for the shopper’s AI assistant.
A Better Way to Learn This Stuff: Research + Quiz + Sanity Check
Keeping up with AI search changes can feel overwhelming, but there’s a practical workflow that works well:
- Run a deep research report using an AI research tool (for example, something like NotebookLM or similar).
- Read through the report to get a high-level understanding.
- Have the tool quiz you on the content so you can check what you’ve absorbed.
- Pay close attention to:
- Places where you’re not confident
- Claims that sound suspicious or contradict your own expertise
This last step is crucial: hallucinations are still very much a thing.
Whenever you see an AI-generated “fact” in an area you know well, treat it as a test case. Does it hold up? If not, assume there are other weak spots in the output and verify anything important before acting on it.
So, How Do You Get Picked by Google AI Mode?
Bringing it all together, if you want your site to be one of the few links surfaced in AI mode:
- Focus on the customer first. Make your content genuinely helpful and clearly written.
- Clarify structure. Use headings, FAQs, and schema so AI can “see” how your information fits together.
- Surface key decision info. Shipping, returns, guarantees, timelines—don’t hide them.
- Invest in brand authority. Real PR, newsjacking, and thought leadership beat random paid links.
- Stay sceptical but curious. Use AI tools to learn fast, but always sanity-check against your own expertise.
The interface may have shifted from blue links to chat bubbles, but the core principle remains the same:
Be the most useful, trustworthy answer to the user’s question—and make it easy for both humans and machines to see that.
If your current SEO strategy is still aimed solely at “page one rankings,” it’s time to zoom out. Review your key pages, FAQs, structured data, and brand presence through the lens of AI-powered search and shopping assistants.
Share this article with your team, audit one core product or service page using the points above, and start adapting now—before AI mode becomes the default way your customers search. If you have any questions or would like to reach out, feel free to contact Jim Stewart at jim@stewartmedia.biz
The post How Google AI mode selects websites to rank (It’s not standard SEO) appeared first on StewArt Media.