What Sundar Pichai Won’t Tell You

Google claims that AI examinations revolutionize search behavior.

But the data tells a different story.

Since in 2024. Launched AI examination, Google CEO Sundar Pichai has repeatedly claimed to transform the search behavior:

People use it to search in brand new ways and ask new types of questions, longer and more complex inquiries … and restore the best that the web can offer.¹

The narrative was consistent with the calls and interviews about earnings:

• “We were able to increase traffic to the ecosystem.”²
• “Growth increases over time as people learn to adapt to this new behavior.”³
• “Based on our testing, we encourage ourselves to notice an increase in the use of searches among people using new AI examinations, as well as increased satisfaction of users with results.” (Source)

Still, Google never supported these claims with real data.

So, I united with a similarWeb to analyze more than 5 billion queries to search in multiple markets.

There is too much good information in this analysis not to share. So today you get one part of the two -piece series. Here’s what we will cover in both parts:

1. 1. Part: How AI examinations affect user behavior on Google.
2. Part 2: How AI examinations affect traffic and engagement on websites.

Increase your skills with weekly professional insights. Subscribe for free!!

The result of allocation?

Google’s claims are partly true but significantly simplified.

My analysis with a similarweb shows:

• People visit Google (+9%) more often, but spend less time on visit.
• The length of the query has barely changed.
• And – most importantly for the SEO professional and traders – data reveal critical insights on how to adapt to this new search paradigm.

About data

The data set includes:

• Over 5 billion search inquiries and 20 million websites.⁴
• Average time on site, session search and visit to Google.com – Total comparing the UK, USA and Germany.
• Comparison of keywords with and without AI examination that analyzes session sessions, an average time spent on Google and a zero-clik share.
• Page and time examinations spent on Google.com for keywords that show AI examinations over keywords without AI examinations.
• The average query length for the UK, the US and Germany.

In this overall requirement, the phrase “searge use” is not very well defined.

Is that more search, more engagement with SERP features or longer sessions?

Unfortunately, I have not been able to specify the exact definition of using the search. This is Google’s own formulation and could be deliberately unclear. (Do you have your thoughts about this? Let me know!)

Whether there was actually a total increase in the use of search due to the AI ​​examination is more complex, depending on several factors. And the analysis below shows clear patterns from which we can learn.

Data especially for the US market, data confirm that the claim “We see an increase in the use of searching among people using new AI reviews” directed correctly.

Here’s how we know it’s right: Google visits increased +9% after the introduction of May 24 (from 26.9% to 29.1%).

The initial fall could be explained by PR disaster from the first two weeks. (Remember those strange results that mentioned smoking at a pregnant woman or glue on pizza?)

While American visits to Google have modestly grew from 2023 to mid -2024, a clearer trend of ascension began around August 2024.

When we look at two comparative sets of keywords – remember, one set in this study shows Aios, and one no – we can see that Page examinations on AIO Keywords web pages have increased by 22% of the USA launch. (Shown in red line below).

I know we all want to talk about the effect of AIOS on organic clicks, but we’ll get there. I will return to the fact that Inquiries that are not aio Drive multiple pages views in the 2nd part.

However, the “site on the website” chart for us searches below reveals two critical insights:

1. Websites over time receive more views than AIO’s keyword searches but
2. Inquiries that are not AIO triggers about twice as many page examinations (shown in a black line).

This suggests that AI examinations can increase engagement for certain types of inquiries, while at the same time have a limited impact on total traffic patterns.

Then let’s look at the US market, the UK and Germany.

Although Google argued that “we see an increase in the use of searches among people using new AI examinations,” in general, similarly similarWeB data show a tinted story.

Here’s how we know that the claim is only partially correct, depending on the market:

The growth of American visits to Google is proportional to Google than in Germany (chart below), which our control market is because AiOS did not poison there until March 2025.

However, in the UK, where AI examinations appeared in August 2024, the visit moves down after the introduction (shown via the green line above).

In fact, there was a greater growth of engagement from 2023 to 2024 (before the introduction of AIO).

Ultimately, I think Google’s claim is incorrect for other markets outside the US:More SERP interactions are not translated into a longer session on Google.

In the lower chart, we can see that the time place for Google.com in the USA and in the UK is or flat or declined.

And something in Google De is quite significantly reduced in place. Maybe it’s related toGoogle loss of market share in the EU.

We see the same trend when we compare the descent of AIO with non-Aio inquiries in the lower chart.

The time on Google for AIO inquiries falls by -1%.

Although this is not a big fall, it is certainly not “an increase in the use of searching”.

Significantly, in the lower chart, you will see that the pages-per-pages of Google.com have rejected the entire panel of 2024 after they performed AIOS, but then they began recovering and growing in 2025.

This chart shows a clear fall on the pages-visits immediately after the start of AIO from May 2024, which suggests that users need smaller pages of results when AI examinations have responded directly to their inquiries.

Subsequent recovery 2025 indicates customer customization or Google adjustment that AIOS works within the search experience.

But what about this sudden rising in 2025?

This also happens in our controlled market, Germany. So this is not because of AIOS.

How do we know that? Pull up to this time on the chart above, which shows all three markets. And you will see that the time of Germany on the spot shows the fall after the starting of Aio.

Although I am not sure what starts this trend, I wonder how less time on Google affects its essence.

MBI has published very interesting Deeply dived to the alphabet with a chart that indicates that AI examinations are not injected and hard.

Instead, it seems that the increase in the cost of click is the unspecified earning of the alphabet.

To be fair, that trend began in 2018, so it is not clear how much Aio accelerated him.

We see an increase in using searches among people using new AI examinations.

Based on data, this claim has layers of truth and omissions.

Google visits have increased post-AIO launch (+9%), and AIO’s keywords of the display have grown impressively (+22%).

However, the full picture reveals important shades that we need to consider:

• Visit to the UK remained straight after the introduction of AIO, despite the American gains.
• The measuring data in the site are straight or declining in markets.
• Pages -s visit initially fell after launching AIOS.

Data suggest that users visit Google more often, but spend less time on visit, probably because AI examinations provide faster answers without deeper research.

This pattern is aligned with the behavior of the “Realve and Rele and Leave” users, not an increased engagement with Google itself.

Although it is technically true that the “use of search” increases with some measuring data, the statement darkens that AIO is basically changing the forms of search interaction at the cost of web traffic.

When we look at the information carefully, this claim is not worth it.

Here’s how we know it’s incorrect:

The growth of the length of the query is small-certainly not a change in steps in “brand new ways”.

We are talking about a very gradual increase from 3.27 to 3.37 average words upon inquiry in the US for two years.

Of course, it’s only 3% on Google, it has a huge impact.

But this is not a change in steps.

Difference in query length between May 2024 and February 2025. Only +0.6%.

In the UK inquiries, the length of inquiries decreased by -0.3% after the AI ​​examinations launched with 3.18 words in August 2024. On 3.17 words in February 2025.

In Germany inquiries, the length of inquiries was slightly increased (+0.4%) ago AI review triggers.

Judgment: This claim is overrated and incorrect

While Google reports “People use it to search … longer and more complex inquiries”, a closer look shows differently.

Data show only minimal changes in the length of inquiries in the US, and the UK has been reduced after AIO -and appears.

Data simply do not support the narrative that AI examinations drive users to construct “longer and more complex inquiries” in any meaningful way.

When we carefully examine the data, a clear pattern appears:

Google’s claims about how examinations AI are fundamentally changed in the way we search largely overrated.

Yes, users visit Google more often, but spend less time on visit and do not make a much longer or more complex inquiries.

This suggests that AI examinations create a “quick response” behavior, not deeper engagement with search.

The modest increase in visits is balanced by reducing the time and pages-the visit.

And the minimum change in query lengths in all markets – whether the AI ​​examination has started – indicates that any evolution in behavioral behavior occurs independently of AI features.

These findings are important because they challenge the narrative that AI examinations represent a revolutionary improvement for search.

Instead, they change the patterns of user interactions in ways that Google has not fully acknowledged.

Keep in mind that I work with third -party data, which can always be distorted or partial. I think it’s wrong, but we always have to keep in mind the limits of data.

1. Optimize for a new “Visit More, Stay Less”: Customers turn to Google more often, but spend less time looking for answers. Your content strategy should focus on being precisely presented in AI examinations and providing a deeper value that promotes clicks when the review is not sufficient.
2. Focus on the quality of engagement: The sample suggests that users are more select with clicking, which makes the quality of experience more important than ever when they reach your site.
3. Factor in regional differences: Significant differences between US -Britain’s behavior after launching a review of Ai suggest that regional testing is necessary – what acts in one market may not transfer directly to others.

In the second part, we will explore even more critical questions: What happens to a wider web ecosystem when users get their answers directly on Google and not click on the websites?

The answer will reveal whether Google’s claims of “growing traffic on the ecosystem” will be kept in control.

---

1 Google I/O 2024: I/O for a new generation

2 CNBC Exclusive: CNBC Transcript: Executive Director of Alphabet Sundar Pichai Talk to CNBC Deirdre Bosa on “Final Bell: Overtime Work” Today

3 2024. Q3 Call for Earnings

4 360 degrees view in a digital landscape

---

FEADED Picture: Paulo Bobita/Search Science Journal