The current environment
Ranked results remain important, but they increasingly share the decision surface with machine-composed answers and conversational interfaces. Google continues integrating advertising into AI-powered Search experiences and testing new conversational and generated ad formats. ChatGPT Ads has moved beyond its initial US test into multiple markets, with advertiser buying and measurement capabilities now available through a mix of direct, partner and self-service access. Microsoft also monetizes conversational search experiences. Community threads, video and other discovery environments continue shaping what people search for before explicit demand reaches a traditional search box.
The exact formats and availability will keep changing faster than the underlying shift: finding, comparing and choosing are spreading across more surfaces, and paid placement is following those decisions.
What is changing
The work of finding, comparing and choosing is spreading across surfaces that compose answers instead of listing links. That moves selection earlier: retrieval systems read pages, weigh evidence and decide what to present before a visitor ever arrives. The website becomes an input to answers as well as a destination for clicks.
That is why relevance, evidence and measurement stopped being three separate vendors' jobs.
Paid placement is following the answer, and two models are emerging. Some AI placements extend existing advertising systems into new Search experiences; others are being built as purpose-specific conversational ad platforms with their own buying and measurement layers. The common requirement is still relevance to the decision and a trustworthy signal about what happened next.
What remains durable
Commercial intent still concentrates where action is possible: someone ready to book, call or buy still expresses it, and businesses still pay to be present at that moment. Evidence still decides trust, for people and machines alike. And the loop between spend and outcome still runs through measurement the business controls.
The instruments change; the system does not.
Websites, agents and actions
Software agents acting for customers read structure, labels and states, the same things assistive technology reads. Sites that are legible and completable get represented accurately and can receive delegated actions; sites that live behind scripts and hover states do not. Agent readiness is a quality layer on the site you already have.
Two forecasts
Paid placement will continue moving deeper into AI-assisted decision environments through a mix of extensions to existing campaign systems and purpose-built conversational ad products. Both models are already visible. Google extends existing campaign types into AI Overviews and is testing AI Mode formats inside its existing system; OpenAI has built a separate ad platform with its own buying, bidding and measurement layer. What remains uncertain is which interfaces, pricing models and optimization systems become dominant.
Agent-mediated visits and actions will remain a small share of most businesses' traffic over the next two years, but will be disproportionately high-intent where they occur. Agents visit with a task. The forecast is about share and character of traffic, not about which agent platforms win.
Current uncertainties
- Which buying model wins on assistant surfaces: extensions of existing campaign systems, purpose-built conversational ad platforms, or both side by side.
- Whether answer-engine referral data will become measurable enough to manage, or stay largely dark.
- How quickly agent-mediated actions become a meaningful share of high-consideration transactions.
- Whether publishers' economics survive as reading moves from pages to answers, and what that does to the evidence supply answer engines depend on.
Implications by business model
High-consideration lead generation: the research window before a direct query is where preferences form; answer-stage visibility and returned lead quality matter more than raw click share.
Local services: presence surfaces (Maps, reviews, local pages) feed both classic results and composed answers; call and sold-work measurement remains the binding constraint on spend.
Publishers and content businesses: the economics of being read are diverging from the economics of being cited; original, citable material is the asset that survives both.