Google AI Mode vs. Traditional Search & Other LLMs [Study]

How Google’s AI Mode Compares to Traditional Search and Other LLMs [AI Mode Study]

Author: Eugene Levin

9 min read

July 21, 2025

Contributors: Anna Yudina and Luke Harsel

Table of contents

Google launched its AI Mode in the US in May, and SEOs are now facing a new frontier.

In this study, we set out to understand how AI Mode compares to traditional Google search, AI Overviews, and two leading LLM-powered search engines: ChatGPT and Perplexity.

Key Findings at a Glance

Methodology

What We Analyzed

To ensure balanced representation, we selected keywords across four core intent types and two distinct search volume tiers. Then, we queried them each into the four AI search platforms (desktop format) and recorded the results.

What We Looked For

AI Citations Only Partially Overlap with Google Search

The first question we asked was about the overlap:

How closely did AI Mode and other LLMs mirror Google’s traditional top 10 rankings?

To measure this, we took each query and counted the number of overlapping domains and URLs between each platform and Google’s top 10 organic rankings.

While we saw a solid overlap between Google’s top 10 organic rankings and LLM citations, the overlap was far from absolute—and it varied by platform.

How AI Mode Compares to AI Overviews

In their 2024 announcements, Google framed AI Overviews as a way to help users get quick, AI-powered answers—paving the way for the fully interactive AI Mode launched just one year later in May 2025.

While Google positioned AI Overviews as a bridge to its next-generation search experience, we saw a few specific ways AI Mode strayed from the AI Overview playbook.

A Response + Sidebar with Links Is the Default Format

If you’re not familiar with AI Mode, here’s what it looks like:

The most common format we saw included an AI-generated response and a sidebar with links. In some cases, there was a local pack of results. And in others, we saw more additional links below the response.

Here’s how often we saw each format:

Sidebar links show Google experimenting with new sources, while links placed below the response stick closely to its traditional organic rankings.

While the links below the response only appeared 7% of the time, they had a very strong overlap with traditional Google results.

This split implies a dual system: AI Mode drew broadly for exploratory queries, but leaned on rankings for precision when needed.

Specifically, we saw AI Mode add additional links most often on navigational searches.

Navigational Queries Were Most Similar to Google Search Results

We noted that the additional links below the response had a high overlap with traditional search results. But mostly on navigational queries.

This makes sense. If you’re using Google to navigate somewhere specific, you’d probably like a list of links (and not just an AI-generated response).

This could indicate one difference in how Google sees AI Mode and traditional search differently: AI Mode helps people gather information, ask questions, and compare options while they’re still making up their mind.

Traditional search is there for when people know where they want to go.

Your Brand in Google’s AI Mode? Now You Can Track It.

Track your brand's presence in AI search and compare it to your rivals—all inside the AI Visibility Toolkit.

AI Mode Shows More Unique Domains Than Other LLMs

One early critique of AI Mode is Google’s lack of transparency in how it sends clicks and traffic to sites, much like with AI Overviews.

Google recently confirmed that some AI Mode tracking will be available in Search Console, but those clicks will be lumped into overall totals, not separated from traditional search data.

In our study, we saw that AI Mode showed more unique domains than the other LLMs, on average, in its results.

AI Overviews linked to an average of three unique domains per query, while AI mode linked to 7.

Perplexity and ChatGPT fell in the middle.

AI Mode Most Resembled ChatGPT Search in Response Length

When it comes to response length, we saw Google’s AI Overviews stand out among the LLMs as the briefest option.

AI Mode was closest to ChatGPT with response length, averaging around 300 words per response.

How AI Mode and Other LLMs Handle Query Intent

Across all platforms, commercial and transactional queries triggered the longest and most detailed responses—often double the length of informational responses.

We can see how much shorter AI Overviews are, on average. But more interesting was the trend of commercial and transaction responses being longer on every platform.

For SEOs, this means aligning content depth with query intent is essential:

A Note on AI Search Intent: While we used four classic search intents for analysis, AI-driven queries are increasingly fluid. With conversational prompts, follow-up questions, and personalized results, AI-driven search is expanding the boundaries of traditional search intent. These four categories serve as a useful reference point for Google search—but marketers should prepare for a more dynamic, multi-intent AI search landscape.

Citation Patterns in AI Mode vs. Other LLMs

The first graph in this study showed the big-picture overlap between the LLMs and Google search.

To find more patterns, we looked at a few more correlations across the platforms—comparing what these AI systems cited with how content ranked and performed in classic organic search.

User-Generated Content Platforms Are Everywhere

We looked at the most frequently appearing domains across the four LLMs (AI Mode, AI Overviews, ChatGPT, and Perplexity) and saw several UGC and social sites at the top of this list.

Strong Rankings at the Domain Level Correlate with AI Visibility

One of the clearest signals in our study: domains that ranked well in Google’s organic top 10 results were far more likely to be cited in AI-generated answers.

The relationship was consistent across all LLMs we examined. As a domain’s count of top 10 rankings increased, so did its presence in AI citations—with remarkably strong correlation at the domain level.

What this tells us is that across every channel, if your domain gets a lot of top 10 Google rankings, it's very likely that you’ll get a lot of AI citations as well.

The takeaway: LLMs may trust the same sources, but they won’t always pull from the same page.