If AI just answers your question instead of sending you to websites, who keeps getting traffic — and who gets cut out?

This explores whether AI agents and AI answers will shrink the flow of visitors that search engines send to websites, and what takes its place. The short answer from the corpus: the decline is already measurable, and the bigger change may be *who* is doing the browsing, not only how much traffic arrives. A randomized field experiment found that Google's AI Overviews cut clicks to third-party sites by 39.8%, and the clicks that remained weren't higher quality and users weren't more satisfied Do AI Overviews actually reduce clicks to publisher sites?. News executives are planning for something similar: they expect search referrals to fall 40–43% over three years Will AI Overviews reduce search referral traffic to publishers?. Those two numbers come from very different methods but land in the same range, which makes the estimate more credible.

AI Overviews are only the first step. They still put a human in front of a results page. Agents go further: they take a goal like "find me a hotel under $200 near the venue" and do the comparing themselves. One argument is that this collapses discovery and purchase into a single zero-click transaction. That hurts search ads, which depend on capturing intent at the moment of need, but leaves influencer and social marketing mostly untouched, because agents can fulfil a want but can't create one Do AI shopping agents destroy search ad revenue while leaving influencer marketing intact?. Travel offers an early signal: Tripadvisor's hotel revenue fell 23% in one quarter, and management blamed structural shifts in search. One analyst reads this as agents keeping the user's demand signal for themselves and ranking options by the user's stated criteria, which makes attention-based ads on aggregator sites irrelevant Are AI agents making attention-based ads irrelevant for aggregators?.

The less obvious point is that the competition for attention doesn't go away. It moves from people to agents. Services will compete to be *selected by agents* rather than clicked by users, and a new layer of agent-facing ranking and recommendation will grow up to serve that Will agents compete for attention just like users do?. This new audience has its own quirks. Shopping agents favor products based on where they appear in a list, even in text-only interfaces and even when told to ignore position, and the bias shifts with each model update Do AI shopping agents show persistent position bias in text interfaces?. AI search engines also cite brands mostly by how famous they already are: 73% citation rates for global brands, 11% for niche ones, even on unbranded queries Do AI search engines favor well-known brands over others?. So the traffic that does survive may concentrate further among incumbents.

That raises the question of who pays for the agent. If an agent earns referral fees from merchants, it can't be fully loyal to both the merchant and the user Can an AI agent serve both merchant and user interests fairly?. A former Visa data chief warns that ad-funded shopping agents carry the same pull toward paid placement that search ads did Do advertising-funded shopping agents favor paid product placements?. In other words, the referral economy may not die. It could reappear inside the agent, where users can't see it.

The corpus has a clear gap. It has strong evidence on AI answers reducing clicks, and good argument about agents, but little measured data on traffic changes caused specifically by autonomous agents, as opposed to AI Overviews. The 40% figures describe the first wave. How large the agent wave turns out to be is still mostly forecast.