ChatGPT Access Linked to 9% Drop in Traditional Search Queries

A Bocconi University study finds ChatGPT referred users to external websites in just 5.2% of sessions compared to Google's 31.1% — while broader ChatGPT access reduced weekly traditional searches by nearly 10%.

The findings arrive at a pivotal moment for the search industry. As AI-powered tools increasingly used as mainstream alternatives to conventional search engines become normalised, publishers, content creators, and marketers face a shifting digital landscape where traffic patterns — and revenue streams — may never look quite the same again. Familiar channels are losing audiences to a fundamentally different kind of experience — one that answers questions directly rather than routing users through a list of links.

---

What the Data Reveals About Referral Traffic

Researchers from Bocconi University analyzed Comscore clickstream data from U.S. desktop users to understand how ChatGPT's growing reach has affected traditional search behavior. The results paint a clear picture of divergence between how Google and ChatGPT send people across the web.

ChatGPT directed users to outside websites in just 5.2% of sessions. Google search generated outbound clicks in 31.1% of queries. That gap — nearly six times — underscores how differently AI chat interfaces engage users compared to a traditional results page. Where Google's model is built on routing users outward, ChatGPT is built on keeping them engaged within the conversation.

Where the Traffic Actually Goes

The types of websites each platform sends traffic to also differ significantly. Google referrals flow heavily toward high-traffic destinations like YouTube, Reddit, and Wikipedia. ChatGPT referrals lean toward smaller and more specialised destinations including reference sites, SaaS platforms, academic resources, and developer tools.

Ad-supported websites received 27.6 percentage points less referral traffic from ChatGPT compared to Google. The authors note that ChatGPT tends to favour non-profit, subscription-based, and freemium platforms — a pattern with serious implications for publishers whose business models depend on advertising revenue tied to inbound clicks.

Despite the lower per-session referral rate, ChatGPT's monthly referral rate grew during the measurement period — rising from approximately 2.5% to nearly 6.5%. Even so, it remains well below Google's per-query referral benchmark.

The researchers were careful to frame their conclusions precisely, stating: "Our claim is deliberately narrower than a welfare claim: we measure a change in observable traffic allocation, not consumer surplus, publisher revenue, or long-run content production."

This distinction matters. The study identifies a structural shift in traffic distribution, not a complete accounting of economic consequences — those remain to be measured as the data matures.

---

How Expanded ChatGPT Access Reshaped Search Volume

OpenAI rolled out ChatGPT Search access in stages. Paid subscribers gained access in October 2024. Free users followed in December 2024. Anonymous users were included by February 2025.

Each expansion wave corresponded with measurable reductions in traditional search activity.

• Broader access reduced weekly traditional search queries by 9.4% overall
• That figure climbed to 17% after 20 weeks of observation
• Among users already active on ChatGPT, compared across access tiers, the decline averaged 4.9% — rising to 8.2% after 20 weeks

The Categories Hit Hardest

The sharpest declines appeared in informational search categories. Academic research referrals dropped 32.8% and reference searches fell 26.5%. Transactional and recreational searches showed little change — suggesting users still turn to traditional search when they want to buy something or browse entertainment content.

This pattern aligns with earlier Pew Research Center data cited by Search Engine Journal showing that users click links far less frequently when an AI-generated summary is present. The Bocconi paper expands that finding by quantifying how access expansion — not just AI summaries — drives the behavioral shift.

It is worth noting that this behavioral change is accelerating in parallel with the broader transformation of SEO strategies and how search itself is evolving — a shift that was already underway before AI chat tools entered the mainstream.

---

Why Publishers and Marketers Should Pay Close Attention

The implications of this research extend well beyond academic interest. For anyone whose business depends on organic search traffic, the findings represent a structural warning, not a temporary fluctuation.

The Sustainability Question for Content Creators

Reference sites, news outlets, academic publishers, and developer platforms are absorbing the most significant traffic declines — even though those same categories appear among the more common destinations ChatGPT does reference. The irony is sharp: the sites ChatGPT draws on most heavily for informational content are also the sites losing the most referral traffic from traditional search.

The authors raise a pointed question about long-term sustainability: will content creators continue publishing at the same scale when AI systems use their work to answer queries but generate fewer direct visits in return? That question remains unanswered. The researchers note that a full picture will require mobile data, international clickstream records, and publisher revenue data to assess true economic impact.

Practical Steps for Businesses and Marketers

For businesses and marketers navigating this environment, the Bocconi findings offer a concrete framework for rethinking traffic strategy. Understanding how to increase website traffic through diversified and sustainable channels is no longer a secondary consideration — it is becoming a primary strategic priority.

• Audit your traffic sources now. If your site depends heavily on informational or academic search referrals, you are in the highest-risk category identified by this study. Diversifying traffic channels is no longer optional.
• Evaluate your content for AI visibility. Since ChatGPT favours reference, developer, and SaaS-type destinations, optimising content for AI citation — not just Google ranking — is becoming a distinct and necessary discipline. This means prioritising accuracy, authority, and source credibility over keyword density alone.
• Reconsider ad-revenue dependency. With ad-supported sites receiving dramatically less ChatGPT referral traffic, business models built around high-volume display advertising face structural pressure that is likely to intensify as AI search access continues expanding.

Reading the Signal Correctly

The data is still incomplete and the researchers acknowledge its limits. But the directional signal is strong: AI search is not simply adding to how people find information — it is actively replacing traditional search for a growing share of queries, and doing so in ways that redistribute web traffic toward a fundamentally different set of destinations.

For further context on how AI is reshaping information discovery at a systemic level, the Stanford HAI research on AI and the web ecosystem offers relevant ongoing analysis from an independent academic perspective.

The businesses and publishers that move earliest to understand and adapt to these changes will be best positioned as the redistribution of search traffic continues to accelerate.