How Has Google’s AI Overviews Changed User Behavior? Reviewing Eye-Tracking Studies 20 Years Apart (A Hamsterdam History Lesson)

Last updated:

September 8, 2026

Welcome to another edition of Hamsterdam History!

This is a series of posts where we look at “vintage” SEO articles to celebrate their contributors, gain historical knowledge, and discuss how things have changed (or haven’t) since.

In this edition, we’ll look at eye-tracking studies for Google users 20+ years apart. Essentially, we want to understand if behavior on a traditional SERP circa 2005 was drastically different from behavior with AI Overviews today.

TL;DR: familiar attention patterns persist, but AI Overviews change what users encounter in that space.

Google Users in 2005

I got the idea for this post when I came across this 2005 eye-tracking study: “Did-it, Enquiro, and Eyetools uncover Search’s Golden Triangle.”

They found a “Golden Triangle” that covered the upper-left results area, including top organic listings, top sponsored results, and sometimes shopping, news, or local suggestions. Every participant looked within this area:

The Golden Triangle circa 2005. Source.

They also found organic visibility dropped sharply with rank. The first three listings were each seen by 100% of participants, compared with 60% for rank 5 and 20% for rank 10:

Organic position	Participants who looked at the listing
1–3	100%
4	85%
5	60%
6–7	50%
8–9	30%
10	20%

Users appeared to scan in an F-shaped pattern, as well. Their eyes moved down the left side looking for relevant keywords, brands, or other cues, then moved right when something caught their attention.

2025-2026 Google Users (with AI Overviews)

Let’s compare the Golden Triangle of 2005 with eye-tracking studies from modern day results that include AI Overviews. We have at least 3 studies to look at:

The mirror of attention: which elements of the SERP are viewed but not clicked on
[EYE TRACKING STUDY] (Laika):

How AI is Changing Google Search: Study on AI Overviews (usability.de):

An Eye Tracking Study: Are AI Overviews Changing Search Behavior? (academic study):

Collectively, the authors of these three different studies found that AI Overviews capture attention in an area users already tend to look at, but that attention doesn’t necessarily extend to their sources or produce clicks.

There are four overlapping takeaways from these studies:

1. Position still matters. The academic study found that the upper-left “Golden Triangle” persisted — “the ‘golden triangle’ remains valid, where the top-left section of the search page attracts the most attention.” The usability.de study likewise found that AI Overviews attracted the first gaze when placed at the top.

2. AI answers receive substantial attention, while organic results remain relevant. The academic study found greater visual engagement with AI content, but recognizable scanning of traditional listings continued. In usability.de’s study, about two-thirds clicked further, mostly on organic results.

3. Looking and clicking are separate behaviors. Laika documented gaps between attention and clicks across SERP elements. In usability.de’s study, the right-hand sources received only 7% of first clicks. An AI answer attracting eyes therefore doesn’t automatically mean its cited publishers receive traffic.

4. Visibility alone doesn’t establish brand impact. These findings don’t demonstrate that being cited produces brand recall, trust, or purchase intent. The academic study also removed reference links from its AI summaries, limiting what it can tell us about source engagement.

Takeaways: Comparing 2005 to 2025-2026:

In summary, AI Overviews appear to preserve familiar attention patterns while changing what people can accomplish within them.

Dimension	Golden Triangle study (2005)	Three recent studies (2025-2026)
Where attention concentrates	Upper-left “Golden Triangle,” including top organic and sponsored results.	The academic study found that pattern persisted; usability.de found that top-positioned AI Overviews attracted the first gaze.
How people scan	Users scanned down the left, then right when relevant words or brands caught their attention, an F-like pattern.	The academic study found recognizable scanning patterns for traditional results alongside greater engagement with AI content.
What receives attention	Individual listings, snippets, and ads.	AI summaries, traditional listings, source links, and other interface elements.
Role of organic results	Visibility fell sharply by rank: 100% saw each of the first three listings, versus 20% for rank 10.	Organic results remained useful: usability.de reported that about two-thirds clicked further, mostly on organic results.
Attention versus clicks	The announcement discussed both, but its numerical tables measured looking, not clicking.	usability.de and Laika explicitly examined clicks alongside attention, showing that being noticed doesn’t necessarily lead to a click.
Getting information without leaving Google	The announcement focused on visibility and choosing links; it didn’t quantify searches ending without clicks.	usability.de reported that around 20% of searches ended without a click, varying by task.

In 2005, people tended to scan Google’s results in an F-shaped pattern, concentrating on the upper left and looking across when a title, snippet, or brand caught their attention, while lower-ranked results received much less visibility.

The recent studies suggest that this preference for the top of the page persists, but AI Overviews now occupy some of that valuable space, giving users information to read before deciding whether to visit a website.

Traditional organic results still attract attention and clicks, while attention to an AI answer doesn’t necessarily translate into clicks on its sources. The apparent shift is therefore from scanning mainly to choose a link toward reading an answer and then deciding whether to explore further.

Caveat: We should also note these studies aren’t directly comparable, as they had different groups of users doing the testing in different settings and measured different things.