News sites expect Google to send them far less traffic soon — because AI answers now skip the click entirely.
How much referral traffic do ChatGPT and Google each send to publishers?
This explores how much reader traffic publishers get from ChatGPT compared with Google Search, and how that balance is changing as AI answers replace links.
This explores how much traffic news and content publishers get from ChatGPT versus Google, and whether AI is changing that. The short answer is that this collection doesn't have a head-to-head count of referrals from each platform, so it can't give you figures for ChatGPT or Google. What it does have is evidence about where the traffic is heading and why. That may be more useful than a snapshot.
On the Google side, the clearest signal is what publishers expect. A Reuters Institute survey of news executives found they expect Google search referrals to fall by 40–43% over three years Will AI Overviews reduce search referral traffic to publishers?. The cause they name is AI Overviews, which answer the question on the results page so the reader never clicks through. Measured search traffic to news sites has already started to drop, though nobody has pinned down how far. So Google is still the larger source of referrals, but publishers are planning for it to shrink.
The ChatGPT side is harder to measure, and the corpus offers a few explanations for why chatbots may never make up for that loss. Most ChatGPT use isn't information lookup at all. Non-work conversations have grown to over 70% of usage, and work use centers on writing rather than finding sources How is ChatGPT actually being used by real people?. A field experiment comparing ChatGPT learners with Google searchers points to a design reason too. The chat interface narrows what people explore, steering them to one synthesized answer instead of a page of links to pick from Does ChatGPT harm informal learning compared to Google Search?. That study measured learning, not clicks, but the same mechanism, less browsing and fewer choices among sources, is what produces fewer referrals.
There's also an early example of what happens to a site when people ask a chatbot instead of visiting. Stack Overflow posting fell 16% relative to comparable Q&A sites after ChatGPT launched, reaching about 25% within six months Did ChatGPT cause Stack Overflow posting to decline?. That measures participation rather than referrals, but it shows the same substitution effect publishers fear.
One point you might not expect: whether AI tools send traffic at all may depend more on business incentives than on technology. Azhar argues that AI agents paid referral fees by merchants can't be fully loyal to users, and that fee structures decide which platforms build or block these agents Can an AI agent serve both merchant and user interests fairly?. For publishers, that suggests ChatGPT's referral volume could depend on licensing and revenue deals as much as on reader behavior. If you need actual referral numbers, look to industry analytics reports, which this collection doesn't currently include.
Sources 5 notes
A Reuters Institute survey of news executives found publishers expect Google search referrals to fall 40-43% over three years, driven by AI Overviews that answer queries in-place rather than directing clicks. Measured data shows search traffic to news sites has already begun declining, though the full magnitude remains unquantified.
Analysis of representative ChatGPT conversations shows non-work use growing from 53% to over 70%, while work use centers on writing tasks rather than information retrieval, concentrating among educated professionals.
In an 8-day field experiment, ChatGPT users scored lower on knowledge tests, especially on critical thinking items, due to reduced agency in information selection and two distortions: ChatGPT's bias toward solutions over principled knowledge, and its conversational interface reducing exploration of the broader knowledge space.
A difference-in-differences model comparing Stack Overflow to four Q&A sites with lower ChatGPT exposure found weekly posts dropped 16% post-launch, growing to around 25% within six months. The effect was largest for programming languages with more training data in ChatGPT.
Azhar argues that agents like Meta's Muse, which earn referral fees from merchants like Expedia, face structural conflicts that prevent unbiased recommendations. The incentive to collect fees, not technology, determines which platforms build or block such agents.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Are Large Language Models a Threat to Digital Public Goods? Evidence from Activity on Stack Overflow
- The effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking: insights from a meta-analysis
- How People Use ChatGPT
- Experimental evidence of the effects of large language models versus web search on depth of learning
- Developing Effective Educational Chatbots with ChatGPT prompts: Insights from Preliminary Tests in a Case Study on Social Media Literacy
- Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia
- Your Programming Students' Cognition with ChatGPT: Higher Performance, Lower Retention, and Reduced Ownership
- How AI Is Changing Search Behaviors