Google and ChatGPT Used Together: What Happens to Queries and Clicks?

Google and ChatGPT don't have to compete in a "one wins, one loses" pattern. The latest data shows most ChatGPT users still use Google, but at the same time, some households run fewer traditional searches, and AI Mode users tend to click through to external websites less often.

This finding is covered in a Search Engine Journal article on Google, ChatGPT audience, search queries, and clicks. The article pulls together several different datasets to show that audience overlap, query volume, search sessions, and referral clicks shouldn't be read as if they were all the same metric.

That's why two seemingly contradictory headlines can both be true: 95% of ChatGPT users are still counted as Google users, yet some users who gain access to ChatGPT Search still reduce their traditional search queries.

What Does the 95% "ChatGPT Users Still Use Google" Figure Actually Mean?

Similarweb reports that around 95% of ChatGPT users also showed up in Google's audience as of September 2025, and that percentage held steady at around 95% through May 2026.

Over the same period, visits to generative AI platforms grew roughly 70% year-over-year.

Similarweb reads this pattern as a sign that AI is being added to search behavior, not fully replacing it.

But the 95% figure only shows audience overlap.

It doesn't answer:

• how many queries are being run on Google;
• how many queries have shifted to ChatGPT;
• how many clicks reach a website;
• what type of task is actually moving.

Why Can Audience Overlap Be Misleading?

Because the threshold for being counted a "Google user" is very low.

Someone who runs just one Google search a month still counts as part of Google's audience.

A person who has shifted most of their research to ChatGPT but still uses Google for:

• Maps;
• store hours;
• login navigation;
• brand searches;

still falls inside that 95% overlap.

In the context of audience measurement, overlap only answers whether the same person shows up on two platforms during a given period — not how large a share of their activity each platform actually gets.

What Did the Bocconi Study Find?

A Bocconi University working paper used Comscore U.S. desktop clickstream data from October 2024 to July 2025.

Researchers compared households that gained access to ChatGPT Search against similar households that previously showed no ChatGPT or Claude activity.

The average results:

• traditional search queries dropped by roughly 3.14 per week;
• that drop equals about 9.4% of a 33.51-query baseline;
• after 20 weeks, the gap reached 17.0%.

So someone can still be a Google user while simply running fewer queries.

Why Did the Researchers Also Report a 4.9% Figure?

The study ran an additional comparison between logged-in ChatGPT users who got Search access in December 2024 and anonymous users who got access in February 2025.

In that comparison, the drop in traditional search was smaller:

• about 4.9% on average;
• about 8.2% after 20 weeks.

The study's authors interpret the smaller figure as a sign that comparing adopters against non-users may partly be capturing a difference in the kind of person who chooses to use AI, not purely the effect of Search access itself.

Does the Bocconi Study Prove ChatGPT Is Causing the Drop in Google Search?

Not entirely.

Search Engine Journal stresses that the paper is still a working paper, and that some of its category-level findings are descriptive.

The dataset also only covers U.S. desktop browsing through July 2025.

That means the results:

• don't fully cover mobile behavior;
• predate the current scale of AI Mode;
• don't automatically represent every user globally.

Which Type of Website Loses the Most Referrals?

In the Bocconi paper, the largest drop in search referrals shows up in the informational category.

Marketplace and entertainment sites barely changed, or the change couldn't be distinguished from zero in the dataset.

The paper stresses that this split is descriptive and doesn't prove a causal mechanism per category.

Why Are Informational Sites More Vulnerable?

One plausible explanation is that an AI assistant can complete more informational tasks without ever needing to send the user to a website.

For example:

User:
"What's the difference between OAuth and an API key?"

Traditional search:
query → SERP → article → answer

AI:
query → generated answer

But the source doesn't prove a single mechanism is behind the entire referral drop.

How Can Google's Reach Stay Stable While Queries Drop?

In that same Comscore panel, Google's household reach stayed at roughly 49%, while ChatGPT rose from 4.6% to 7.9%.

So:

Google audience
≈ stable

Google queries per household
↓

Those two metrics don't contradict each other.

One measures whether a household uses Google at all. The other measures how many searches it actually runs.

What Did Another Study on Search and AI Find?

SEJ also covers another working paper that analyzed 4,394 Comscore desktop users who started using a large language model between August 2023 and January 2024.

That study found chatbot adoption was associated with an increase in the number of unique websites visited per week, and a temporary rise in traditional search activity that later declined.

The authors call the search estimates noisy and don't claim there's a lasting reduction.

That shows the evidence is still mixed.

Can Users Use Search and AI Within a Single Session?

Yes.

A 2025 field study of 308 U.S. Chrome users found that sessions using both an LLM and a search engine together tended to:

• run longer;
• involve more websites.

While LLM-only sessions were shorter and narrower.

That dataset comes from a period before ChatGPT Search and today's version of AI Mode existed, so it doesn't fully represent the latest generation of these products.

What Happens When Google Search Turns Into AI Mode?

This is a different question from "are users leaving Google for ChatGPT?"

Google can keep a user while swapping the experience from a traditional SERP to an AI-generated interface.

At I/O 2026, Google said AI Overviews has more than 2.5 billion monthly active users, and AI Mode passed 1 billion users in its first year.

Google also said query volumes are hitting record levels.

But SEJ stresses that claim isn't accompanied by a baseline or user-level data sufficient for independent verification.

How Was the AI Mode Experiment Run?

A working paper published on arXiv on August 18, 2026, recruited 1,100 U.S. Chrome users who use Google as their primary search engine.

After a three-day baseline, participants were randomly split into three groups for one week:

1. standard Google;
2. Google with AI Overviews hidden;
3. every search routed to AI Mode.

At baseline, only 0.6% of searches used AI Mode.

In the treatment group, 94.7% of searches were routed to AI Mode.

The researchers themselves describe this condition as forced, full adoption, and caution that the effect of voluntary adoption could differ.

What Was AI Mode's Impact on External Clicks?

In that experiment, AI Mode produced an 18.8 percentage-point drop in click-through to external websites.

The 95% confidence interval sits at roughly -22.2 to -15.3 percentage points.

Search sessions also fell by about 0.92 per day from a baseline average of 4.0.

Meanwhile, average session duration rose by about 0.43 minutes.

What Does It Mean to Have Longer Sessions but Fewer Clicks?

A user can spend more time on Google while visiting external sites less often.

This shows a change in value flow:

Before:
Google session
→ website click

AI Mode:
longer Google session
→ more answers on Google itself
→ fewer outbound clicks

For a publisher, retaining Google's audience doesn't always mean referral traffic stays stable too.

Which Websites Saw Fewer Clicks in the AI Mode Experiment?

In that treatment, the share of participants clicking through to:

• news sites dropped 12.5 percentage points;
• Reddit dropped 21.2 points;
• Wikipedia dropped 9.9 points.

The study also found use of Bing, DuckDuckGo, or Yahoo rose 11.2 points.

But trust, satisfaction, and usefulness ratings all declined.

Do Users Have Trouble Reaching Specific Websites in AI Mode?

Of 309 AI Mode users who answered an open-ended question, 47 — about 15.3% — reported difficulty reaching a specific website.

SEJ quotes one participant who said that when they typed in a site's name, AI Mode would sometimes describe the website instead of taking them directly to it.

This finding comes from open-ended responses in a limited experiment, not a representative survey of every Google user.

What Are the Limitations of the AI Mode Experiment?

SEJ notes several limitations:

• the period was only seven days;
• the sample skews younger and more educated than the general U.S. population;
• the sample skews more left-leaning;
• AI Mode use was forced on the treatment group;
• the AI Mode product at the time of the experiment didn't yet carry ads the way a later version does.

Because of that, the results shouldn't be read as a permanent forecast.

Why Doesn't 95% Audience Overlap Answer the Traffic Question?

Because a website can experience:

Google audience staying high
+
Google queries dropping
+
external clicks dropping

all at the same time.

Take a news publisher as an example:

• Google still has a large audience;
• users are still searching on Google;
• AI Mode answers the question directly;
• the publisher gets fewer clicks.

This explains why impressions or platform reach alone aren't enough to assess business impact.

What's the Attribution Problem When ChatGPT Drives Branded Search?

For example, a user asks ChatGPT:

"What's a good payroll software for a UK startup?"

ChatGPT names Brand X.

The user then opens Google and searches:

"Brand X pricing"

Search Console will see this as organic branded search.

But the actual discovery started at ChatGPT.

Traditional analytics struggle to connect those two steps.

What Is "Branded Search Lift" in AI Measurement?

Similarweb calls branded search lift one of the metrics that can move once AI is driving demand.

It's a form of indirect conversion that often goes unattributed to the AI session.

A team can monitor:

• brand query volume;
• direct traffic;
• AI referrals;
• self-reported attribution;
• sales-call source;
• new branded search patterns.

What Five Metrics Need to Be Kept Separate?

All five of these metrics can move in different directions at the same time.

What Can Current Datasets Not Show Yet?

According to SEJ, there's no dataset yet that tracks an individual user over time across:

• Google Search;
• AI Mode;
• ChatGPT;
• Gemini;

at the task level.

As a result, we can't yet see the full journey, like:

User:
ChatGPT research
→ Google branded search
→ website
→ conversion

complete within a single independent dataset.

Why Do the API and Desktop App Become a Blind Spot?

The Similarweb audience overlap discussed in the article doesn't cover all of:

• API usage;
• desktop app activity;
• in-app AI features.

Yet that activity is exactly the kind that could be replacing traditional search tasks without ever showing up in browser-based audience measurement.

What's the Implication for SEO Reporting?

An SEO team needs to stop reading traffic through a single metric.

A more complete dashboard needs to separate:

• impressions;
• clicks;
• brand queries;
• AI referrals;
• AI citations;
• direct traffic;
• conversion.

If impressions stay stable but clicks drop, don't immediately assume it's a ranking issue.

Search behavior can change inside the platform itself.

What's the Implication for Publishers?

An informational publisher is potentially more vulnerable to zero-click behavior.

Bocconi's data shows academic, reference, developer, and news referrals dropping more than marketplace or entertainment referrals.

A publisher's strategy needs to pay attention not just to visibility, but to:

• click-worthiness;
• subscription conversion;
• brand demand;
• unique data;
• content that isn't easy to compress into one answer.

What's the Implication for E-commerce?

Bocconi found marketplace referrals relatively unchanged in their dataset.

That doesn't mean e-commerce is safe from AI substitution.

Agentic commerce, AI shopping, product comparison, and direct checkout can still reshape the journey.

But for certain transactional tasks, a user still needs a merchant or marketplace to actually complete the action.

What's the Implication for B2B?

A B2B buyer can go through:

AI research
→ shortlist
→ Google brand search
→ product page
→ demo

If you only look at last-click analytics, the AI's contribution disappears.

Because of that, a B2B business can add:

• a "How did you hear about us?" field;
• sales-source notes;
• an AI-assistant option;
• brand-search monitoring.

What's the Connection to Generative AI Strategy?

Generative AI doesn't just create a new channel. It reorders the channels that already exist.

Before:

Google
→ Website
→ Conversion

Now:

ChatGPT
→ Google branded search
→ Website
→ Conversion

or:

Google AI Mode
→ Answer
→ no click

or:

AI
→ recommendation
→ direct visit

If your team has previously covered AI visibility or referral-traffic changes, the internal article Getting Into ChatGPT Is Easy — AI Agent Checkout Is the Real Challenge can serve as an internal link once the previous article's URL is verified.

How Do You Build a Better Measurement Framework?

1. Separate audience from usage.
2. Separate queries from clicks.
3. Monitor branded search.
4. Track AI referrals.
5. Add self-reported attribution.
6. Segment informational versus transactional.
7. Don't apply one study to every vertical.
8. Compare business outcomes, not just traffic.

What Shouldn't Be Concluded From This Data?

• 95% overlap doesn't mean 95% of search behavior stays the same.
• A 9.4% query decline doesn't apply to every user and country.
• The 18.8-point click decline came from a seven-day forced AI Mode experiment.
• A stable Google audience doesn't mean stable referral traffic.
• AI hasn't been proven to fully replace Google.
• Google hasn't been proven immune to query substitution.

FAQ About Google and ChatGPT

Do ChatGPT users still use Google?

Yes. Similarweb reports about 95% of ChatGPT users also fall inside Google's audience through May 2026.

Does ChatGPT reduce traditional search?

A Bocconi working paper found households with ChatGPT Search access ran about 9.4% fewer traditional searches compared to a specific baseline, but this result comes from U.S. desktop data and the paper isn't yet peer-reviewed.

Does AI Mode reduce clicks?

In an experiment on 1,100 U.S. Chrome users, routing search to AI Mode reduced external click-through by about 18.8 percentage points during the seven-day treatment.

Is Google losing its audience?

In the Comscore data Bocconi analyzed, Google's household reach stayed at roughly 49% even as queries per household dropped.

What type of website is most affected?

In the Bocconi data, academic, reference, developer, and news sites saw a bigger referral drop than marketplace and entertainment sites.

Why does branded search need monitoring?

Because a user can discover a brand through ChatGPT and then search for it on Google. Traditional analytics can attribute that visit to organic Google traffic without ever showing AI as the actual discovery source.

Conclusion

Google and ChatGPT aren't moving through a simple substitution pattern. Similarweb shows about 95% of ChatGPT users still use Google, but a Bocconi working paper finds some households genuinely running fewer traditional searches after gaining ChatGPT Search access.

At the same time, the AI Mode experiment shows Google can retain users and search sessions while sending fewer clicks to external websites. In that seven-day treatment, external click-through dropped 18.8 percentage points.

The most important lesson for SEO and generative AI is to keep audience size, query volume, search sessions, referral traffic, and AI visibility separate. All five can move in different directions.

For a business, the more useful question isn't "is AI replacing Google?" — it's "which part of the customer journey now starts, shifts, or completes inside AI?"

If your business wants to build AI visibility monitoring, search attribution, branded-demand tracking, or analytics that connect Google, ChatGPT, and the generative AI journey, you can discuss your business's technology needs with our technical team.