Prabhakar Raghavan occupies an unusual position in Google’s artificial-intelligence transition. He no longer runs the organisation that contains Search, Ads, Maps and Commerce. Nick Fox took that operating responsibility in October 2024. Raghavan became Chief Technologist, working with Sundar Pichai and Google’s leaders on technical direction.
The distinction matters in 2026. Google has transformed Search through AI Overviews and AI Mode, but those launches belong to the product teams now under Fox and to Google DeepMind’s model organisation. Raghavan’s influence is advisory and architectural. He can shape how the company thinks about retrieval, ranking and the interaction between models and information, without being credited for every feature or held as the sole operator of the business.
That role nevertheless carries weight. Search remains Alphabet’s largest economic engine, and generative AI changes its basic unit from a ranked set of links to a composed answer, conversation or action. Raghavan spent much of his career on the science and business of finding information. His leadership challenge is to help Google preserve the discipline of retrieval while models make the interface more expansive—and less predictable.
Search growth gives Google room to rebuild
Alphabet’s fourth-quarter 2025 results showed that AI had not displaced the core business. Google Search and other advertising revenue rose 17 per cent to $63.1 billion. The company said AI Mode queries were about three times longer than traditional searches, queries per user had doubled since launch and nearly one in six AI Mode queries included non-text inputs. Search usage reached record levels.
By May 2026, Google said AI Mode had more than one billion monthly users and query volume was more than doubling each quarter. AI Overviews reached more than 2.5 billion monthly users by June. These are company-reported usage measures rather than evidence of incremental profit, but they demonstrate extraordinary distribution. Google can introduce a new information interface to a billion people without asking them to install a separate application.
Distribution buys experimentation time. Google can test where users prefer a concise answer, a conversation, images, video or conventional results. It can route difficult questions to more capable models and use cheaper systems for simple retrieval. The company’s advertising platform can connect commercial intent with merchants. Few competitors possess the same combination of index, user behaviour, models and monetisation.
Scale also increases the cost of error. A model-generated answer placed above links carries Google’s authority even when it is wrong. Longer queries may include health, finance or legal questions where context and uncertainty matter. A conventional search engine presents sources for users to compare; a synthesised response decides what to include and how confidently to say it. The interface can compress disagreement into a smooth paragraph.
Raghavan’s background in information retrieval is relevant because generation does not remove the need to find evidence. Strong search systems measure relevance, freshness, authority and user satisfaction. AI search adds another layer that must select sources, reconcile them and express uncertainty. The model should not become a substitute for a healthy index.
The advertising model must adapt without distorting answers
Google’s search economics depend on commercial queries. Advertisers bid for access to users with intent, and those bids fund the free product and much of Alphabet’s wider investment. An AI interface can deepen intent by helping a user compare options or complete a task. It can also reduce the number of pages viewed and blur the boundary between recommendation and advertisement.
The company must make sponsored content unmistakable inside conversational results. If a model recommends a product, users should know whether the choice reflects relevance, payment or both. Advertisers need measures tied to outcomes rather than impressions embedded in a long answer. Regulators will examine whether Google gives its own shopping, travel or local services preferential treatment.
AI queries also consume more computing resources than conventional search. Revenue per query can rise if the conversation produces stronger commercial intent, but cost may rise first. Google can use its own tensor-processing units and models to improve efficiency, yet the margin equation depends on routing. Spending frontier-model inference on a navigational query would be wasteful; using a weak model for a complex purchase could reduce trust and conversion.
The relevant business measure is not simply AI Mode usage. It is the contribution margin of completed tasks, including inference cost, user retention, advertising value and the downstream health of the web. Raghavan can help ensure technical teams optimise for that system rather than one model benchmark or engagement statistic.
Publishers are infrastructure, not an externality
Search requires material to index. News organisations, specialist sites, retailers, forums and independent creators invest in producing the information that models retrieve and summarise. If AI answers reduce referral traffic without creating another return, some publishers will produce less or block access. The index then becomes narrower and less current.
Google has argued that AI features can lead users to a wider range of sources and has introduced publisher controls and reporting. The critical evidence will be distributional. Aggregate clicks can conceal losses among original reporting and gains for derivative pages. Publishers need clear data separating citations, referrals and model usage, as well as practical controls that do not require withdrawing from conventional search entirely.
This is not only a media-policy issue. It is a supply-chain risk for search quality. Google can train on historical material, but fresh facts and first-hand expertise require continuing investment outside Google. Payments, licensing, product partnerships or traffic may all form part of a sustainable arrangement. The correct mix will differ by content type.
Asia makes the dependency visible. Mobile-first users search across local languages and often rely on small publishers, merchants and community sources that lack sophisticated optimisation. A model may favour abundant English material or large international sites even when local knowledge is more relevant. Better multilingual retrieval and transparent source selection are essential if AI search is to expand access rather than centralise attention further.
Regulators are changing the architecture
Google’s technical transition is occurring under structural legal pressure. In September 2025, the United States Department of Justice secured remedies after its search-monopoly case. The measures prohibited certain exclusive distribution contracts and required forms of search data and syndication access for competitors. Compliance continued into 2026. In Europe, the European Commission adopted binding Digital Markets Act measures in July 2026 requiring Google to share anonymised search data with eligible rival search engines, including AI chatbots with search functions, on fair, reasonable and non-discriminatory terms.
Data sharing can lower barriers for competitors that cannot reproduce Google’s query and click history. It also introduces privacy, security and quality questions. Data must be sufficiently useful to support competition without exposing individual behaviour or enabling manipulation. The technical design of access—fields, latency, sampling and price—will determine whether a remedy works in practice.
Raghavan’s advisory role places him near these architecture questions, even though legal teams and operating executives own compliance. A technically narrow response that satisfies the text while preserving every incumbent advantage would invite further intervention. An overly broad release could expose sensitive data or degrade the incentives to maintain infrastructure. Google needs durable interfaces that regulators and competitors can test.
The end of some exclusive arrangements may also change distribution economics. Google has historically paid large sums to appear as a default search provider. Users increasingly enter AI through assistants, browsers and device operating systems, creating new control points. Search quality alone may not determine which service receives a query. Google must compete in products where it does not own the interface.
Competition will also be measured by interoperability. If users create histories, preferences and agent workflows inside AI Mode, moving to a rival may become difficult even without a default contract. Google should make export and deletion practical and avoid tying unrelated services unnecessarily. Regulators are likely to treat accumulated context as the next generation of search lock-in.
Chief Technologist is an influence role with an accountability problem
Raghavan’s career includes foundational work in search and senior roles across IBM, Yahoo and Google. At Google he led Ads and Commerce before taking charge of Knowledge and Information. That breadth helps him see the relationship between ranking, distribution and revenue. It can also make his present contribution difficult for outsiders to measure.
A Chief Technologist should do more than review projects. The role can establish principles that span organisational boundaries: when answers require citations, how retrieval quality is evaluated, which workloads deserve expensive models and how new interfaces preserve competition. Raghavan can also challenge teams when short-term growth creates long-term damage to source ecosystems or trust.
Accountability requires visible outcomes even if individual advice remains confidential. Google should demonstrate fewer factual failures in high-risk queries, stronger performance across languages, clearer advertising separation, useful publisher controls and compliance interfaces that function beyond formal documentation. These are institutional results, not features owned by one executive.
Prabhakar Raghavan is no longer the operating head of Search, and treating him as such would misstate Google’s organisation. His importance lies in what the change of role acknowledges: AI search is more than another product cycle. It requires technical judgement across models, retrieval, economics and regulation. If Google can turn billion-user AI interfaces into reliable gateways that continue to reward the information ecosystem, his influence will be evident in the system’s coherence. If answers grow while sources, margins or trust weaken, no amount of usage growth will resolve the architecture he is meant to help guide.
