ChatGPT access drives 9% drop in Google search use
As ChatGPT displaces Google queries and rarely replaces their referrals, B2B brands face a shrinking discovery window on both channels.
Key takeaways
- Broader ChatGPT access correlates with a 9% drop in traditional Google search volume.
- ChatGPT sends far fewer users to external websites than Google does, making the referral loss structural, not temporary.
- Brands optimised for Google rankings lose twice: fewer queries reach them, and ChatGPT citations favour a narrow set of authoritative sources.
- Getting cited in LLM answers requires upstream authority-building, not keyword optimisation.
- Institutions whose content appears in training corpora (standards bodies, UN agencies) hold a structural citation advantage over commercial publishers.
A 9% fall in traditional search volume is not a rounding error. Search Engine Journal reports on a new research paper finding that broader access to ChatGPT correlates with a statistically significant drop in Google queries, a figure large enough to represent hundreds of millions of searches per day redirected away from the index that has governed brand visibility for two decades.
The mechanism is straightforward: when users gain access to ChatGPT, a share of them stops typing queries into Google. They get an answer synthesised from training data and retrieval, with no list of blue links to click. The downstream effect is not merely that Google loses traffic. It is that any brand whose discoverability depended on ranking well in those displaced queries loses a discovery pathway entirely.
The referral gap that the headline number obscures
The same research paper examined not just search displacement but referral behaviour: how often does ChatGPT actually send users to a website, compared with Google? The answer matters enormously. A drop in Google searches hurts brands that rank well but receive clicks. ChatGPT citations, by contrast, are sparse and non-democratic. The model cites when it must, not when it can, and the sources it chooses skew heavily toward a small cluster of authoritative domains. For most brands, the net effect of this transition is a double loss: fewer Google impressions and near-zero ChatGPT referrals.
For financial services firms, multilateral institutions, and major industrial groups, the dynamic is particularly pointed. These organisations have spent years optimising for search: commissioning white papers, publishing technical standards, building link profiles. That investment was predicated on a world where users go to search engines to find what they need. A user who now asks ChatGPT "which multilateral development banks are funding climate adaptation in South Asia?" may receive a confident, synthesised answer that never surfaces a specific institution's content at all, regardless of how well that content ranks on Google.
The 9% figure is also almost certainly a floor, not a ceiling. The study captures a specific window of access expansion. Usage of AI-native query interfaces has continued to grow since the period examined. OpenAI reported 400 million weekly active users in February 2025. Google itself has begun redirecting queries toward AI Overviews, which compress the same referral problem into its own product. Brands are being squeezed from both sides.
What citation scarcity means for B2B authority
The referral comparison embedded in the research is the sharper finding for B2B brands. Google's model is abundant with clicks: ten organic results, ads, featured snippets, knowledge panels. ChatGPT's is austere. When it does cite, those citations function more like academic references than search results, conferred on sources the model has already determined to be credible. Getting cited in a ChatGPT answer is not a traffic strategy; it is an authority signal. And authority in this context is built before the query, through the corpus of content a model was trained on and the sources it has learned to trust.