Alphabet's 2025 Form 10-K reports Google Search & other revenue of approximately $175.033 billion in 2023, $198.084 billion in 2024, and $224.532 billion in 2025. The 2025 increase was approximately 13.4%. These are not the figures of a search business that has ceased to be commercially viable [1].

In the second quarter of 2026, Alphabet reported consolidated revenue of $119.8 billion, up 24% year over year. Google Services revenue was approximately $94.5 billion, while Google Search & other grew 17%. Consolidated operating margin was 34% [2].

Aggregate revenue growth does not, however, prove that the traditional web-link business model is unaffected. Public segment results do not disclose the full unit economics of AI search queries. Advertising inventory, conversion paths, and traffic composition may all be changing.

My assessment is that a claim that Google has already been disrupted must explain its continuing search-revenue growth. Conversely, a claim that Google cannot be disrupted must explain why new interfaces could never capture high-value commercial intent.

Alphabet reported approximately $294.691 billion in Google advertising revenue for 2025: $224.532 billion from Google Search & other, $40.367 billion from YouTube ads, and $29.792 billion from Google Network [1]. These are revenue categories—not profits or revenue per query.

A query asking why the sky is blue and a query expressing intent to purchase a service do not necessarily have the same commercial value. Fewer outbound clicks after informational queries do not automatically imply a proportional decrease in advertising revenue from high-intent transactions.

I expect the next competitive contest to revolve around how commercial intent is expressed, interpreted, compared, acted upon by agents, and converted into transactions. Google may integrate AI into its existing advertising system; challengers may shift a transaction from a search box into a conversational or agent interface. Both developments can happen at once.

Microsoft's FY2026 annual report presents search advertising revenue of approximately $15.176 billion, compared with $13.878 billion in FY2025; it describes search advertising revenue growth excluding traffic-acquisition costs as 12% [5]. Those figures cannot be treated as a directly comparable market-share denominator against Alphabet's calendar-year Google Search & other segment.

Microsoft benefits from operating-system, browser, productivity-software, and cloud distribution. Google combines its index, browser, mobile ecosystem, advertising technology, and AI search experience. Independent AI assistants can become alternative interfaces for interaction, task execution, and recurring workflows.

I would not reduce this contest to a model leaderboard. Distribution, inference cost, retention, advertiser relationships, trust, and transaction completion matter together. Current public financial statements do not establish that any challenger has replaced Google's search economy.

Pew Research Center studied browsing behavior from 900 U.S. adults in March 2025, covering 68,879 Google searches. In the observed sample, 8% of searches displaying an AI summary were followed by a click on a traditional search result, compared with 15% when no AI summary appeared [3].

The same study reported that 26% of sessions with an AI summary ended without another recorded browse action, compared with 16% without one. About 1% of visits presenting an AI summary involved a click on a citation inside the summary [3].

A publisher may contribute information without receiving a site visit. But query types have different inherent click propensities, and AI summaries were not randomly assigned. The observed association should not be presented as a fully identified causal effect of AI. Nor does a decrease in publisher clicks by itself demonstrate a decrease in Google's advertising revenue.

In an Ahrefs Web Analytics sample of 115,254 websites, conventional search accounted for about 26.82% of site visits in September 2026, while AI-assistant referrals accounted for about 0.43% [4]. These figures measure website referrals—not total AI usage, brand influence, or return on investment.

The sample is not a census of the internet, and channel classifications may overlap. Being named in an AI answer without a click may influence a brand without generating an observable referral. It would be wrong to treat 0.43% as a ceiling on the commercial relevance of GEO.

My conclusion is that conventional search retains a significant measured distribution advantage today, while companies should separately examine high-intent visits, subsequent branded search, and verified inquiries. A citation count is not revenue.

A company can verify that its website responds correctly, permits intended crawling, uses clear canonical URLs, provides substantive page content and traceable sources, maintains version history, and measures conversion paths. It cannot guarantee a first-place Google ranking, a fixed AI citation, a buyer click, or increased sales.

Google Search Central states that appearing in AI Overviews or AI Mode does not require a special optimization method or proprietary machine-readable file beyond the ordinary applicable search requirements. Eligibility does not guarantee crawling, indexing, or visibility [6].

I separate three levels of certainty: work completed and accepted; intermediate signals that can be observed; and commercial outcomes that remain uncertain. Mature channels should meet efficiency standards, emerging channels deserve bounded experiments, and both require evidence. One should not infer a completed transaction from a completed technical checklist.

If AI agents increasingly search, screen, compare, negotiate, and take actions for us, the next network may require more than identity and payments. It may require inspectable evidence and trust mechanisms.

I use “Web 4.0” as a forward-looking description of this possible intelligent network, not as an established technical standard. The objective should be to preserve people's rights to know, choose, carry their data, revoke authorization, correct errors, exit systems, and retain final judgment over consequential decisions.

Evidence is not surveillance. Responsible systems should minimize data collection, restrict purposes and retention, support revocation, and permit independent review and dispute resolution. Traceability without meaningful rights can become another form of control.

My argument is not that machines must become gentle. Institutions must be capable of constraining machine indifference.

I do not want the next internet to compete only for ever-more-precise attention capture. Infinite feeds, immediate rewards, and personalized suggestions can become extraordinarily effective even as people find it harder to leave them.

If AI simply intensifies these mechanisms, knowing users better does not necessarily make users freer. Efficiency and autonomy are not synonyms.

I want intelligent systems to reduce passively consumed time and return it to curiosity, creativity, genuine learning, independent judgment, and kindness toward another person. Agency does not require doing everything without assistance; it means remaining able to understand why one chooses.

A person's willingness to listen, pursue an inquiry without immediate commercial value, or take responsibility that cannot be quantified belongs to human life—not to platform engagement metrics. Machines may save time. That time should ultimately return to people.

For Google's competitive position, track search advertising revenue, relevant profitability, advertiser spending, demonstrable transactions on competing platforms, and shifts in behavior toward new interfaces. If Google's search unit economics remain robust, the claim that it has already been disrupted should be rejected.

For GEO, track attributable high-quality visits, subsequent branded interest, and qualified opportunities, while acknowledging brand effects that are not directly observed. A citation is not a transaction; an absence of referral clicks does not necessarily mean zero value.

For Web 4.0, ask whether people can inspect an agent's actions, withdraw permission, export their data, correct mistakes, refuse profiling, and exit products. Trust infrastructure that increases surveillance, dependence, or manipulation fails the sovereignty test.

My research position is to constrain business claims with data, constrain technical power with institutions, and use technology to expand human capability—not let theory stand in for reality.

I am in no hurry to announce any company's final defeat. Google may reinvent itself; challengers may establish new entrances to the internet; the distribution of future profits requires additional evidence.

More important than who wins a platform contest is whether intelligent systems make facts easier to verify, errors easier to correct, permission genuinely revocable, and time more available for living.

Curiosity carries us toward the unknown. Creativity allows us to move beyond an existing answer. Kindness keeps us from seeing another person as merely a profile or variable.

Visibility is not credibility. Citation is not verification. Evidence is not trust. Trust requires responsibility.

The Web 4.0 I hope for is not a world in which machines finally answer everything. It is one in which people remain able to ask, question, create, and care for one another as machines become more powerful.