Google hosted its annual I/O developer convention final week. Predictably, each product showcase emphasised AI-enabled performance; this video supplies a useful, 10-minute overview of every little thing that was highlighted. Google revealed a number of new options and core merchandise on the occasion, together with:

• Two new subscription plans for Gemini: Google AI Professional ($19.99 / mo) and Google AI Extremely ($249.99 / mo);
• A “Deep Assume” mode for the Gemini 2.5 mannequin;
• “Agent Mode” for the Gemini app and expanded performance for Venture Mariner, each of which are actually out there for Google AI Extremely subscribers. It’s not clear how these options are differentiated (Venture Mariner was launched into beta late final 12 months as a browser-based agent), as they each present agentic capabilities;
• Veo 3, an up to date video generative mannequin that may additionally produce audio, and Imagen 4, an up to date picture technology mannequin.
• Stream, an AI film-making software based mostly on Veo 3.

However the firm made two different revelations at I/O that present readability round its industrial technique with AI. The primary is the introduction of “AI Mode” to Search, which was rolled out to US-based customers on Might twentieth. AI Mode makes use of Venture Mariner to convey “extra superior reasoning, considering and multimodal capabilities” to Google Search queries: basically, it’s a search-optimized chatbot accessible from the homepage.

Google introduced at Google Advertising Reside, the advertising-focused facet occasion held the day after the I/O mainstage occasion, that advertisements are already stay in AI Mode. Additional, at I/O, the corporate unveiled a digital dressing room characteristic that makes use of AI Mode to assist facilitate product discovery and buying instantly from the question.

I view AI Mode because the pure and apparent development of what I’ve referred to as Google’s gambit: its crucial to transition Search from a content material distribution mechanism to a content material sink, capturing person engagement by AI-enabled interplay touchpoints and monetizing that elevated engagement with promoting. From that piece:

As a distribution mechanism, Google optimizes to generate one click on per search: a person considers a search question profitable if it results in a click on to the suitable vacation spot from the primary web page of outcomes. However the incentive with AI Overviews could be very clearly completely different: it’s to unravel the person’s wants with out resulting in a click on, doubtlessly whereas instigating additional queries and, due to this fact, elevated time spent with AI Overviews. And advertisements are displayed every time AI Overviews are rendered, doubtlessly main to 1 click on per question because the person refines their search. In impact, with AI Overviews, Google can seize the conversions and advert clicks that happen subsequent to a profitable search question within the present distribution mechanism configuration.

It appears clear that Google seeks to execute a whole Ship of Theseus transformation of Search from a hyperlink rating product right into a fully-contained client discovery engine that facilitates commerce by its chatbot and with different, novel options just like the aforementioned digital becoming room. AI Mode is completely in keeping with Google’s motivations associated to the open internet: as I talk about in Google was the open internet’s imperfect benefactor, with its Community enterprise threatened by the result of its “advert tech” antitrust trial, Google has little incentive to protect Search’s position as a content material discovery software since it could not have the ability to absolutely or effectively monetize the engagement that takes place on internet locations. I made the purpose extra pointedly in a latest interview with Ben Thompson on Stratechery:

And also you may say, “Watch out what you want for”, as a result of in the event that they lose whichever a part of the Google community enterprise, it could possibly be the provision facet, could possibly be the demand facet, they’ve zero incentive to assist the open internet … What are they already doing? They’re already inserting AI overviews into all the above the fold phase of Google search and if they are saying, “Look, we simply don’t care concerning the open internet anymore, we are able to’t monetize it”, then they will desire all the businesses that they’ve knowledge licensing offers with, like Reddit, to desire and prioritize these ends in the AI overviews, drive visitors to them, instantiate that suggestions loop and earn money on the advertisements that they serve on AI overviews.

This technique seems to be working: Google revealed at I/O that, “in [its] greatest markets, just like the US and India, AI Overviews are driving over 10% development within the kinds of queries that present them. What’s notably thrilling is that this development will increase over time.” Additional, the corporate famous in its Q1 2025 earnings report that AI Overviews are monetizing “on the similar fee” as Search. If AI Overviews is rising question quantity whereas attaining the identical degree of monetization as Search, why would Google need to protect conventional Search? AI Mode appears explicitly designed to cannibalize outbound search end result clicks even additional.

Taken collectively, these updates trace at Google’s intent to completely reposition Search: not as a place to begin for third-party content material navigation, however as a self-contained commerce and discovery platform constructed round AI engagement.

The second revelation from I/O that brings Google’s industrial AI technique into sharper focus was extra delicate: Sundar Pichai famous that Gemini with personalization, a product characteristic launched in March, could be utilized to Gmail with Personalised Good Replies, which permit generative textual content to be tailor-made to the linguistic patterns of the person. However the characteristic was described in much more normal and far-reaching phrases than that:

Along with your permission, Gemini fashions can use related context throughout your Google apps in a manner that’s personal, clear, and absolutely underneath your management … Gemini can do virtually all of the work for me: wanting up my notes in Drive, scanning previous emails for reservations, Gemini matches my typical greetings from previous emails, captures my tone, model, and favourite phrase selections, after which it routinely generates a reply.

What Pichai describes right here isn’t merely a Gmail characteristic: it’s a broadly-applicable personalization engine that Google could also be singularly positioned to construct and deploy, given the a number of touchpoints — and huge pool of information — it enjoys with most customers throughout Gmail, Search, Drive, Maps, and elsewhere. Word that what’s being described isn’t a mannequin skilled on user-level knowledge, which is impractical. This sounds much more profound: a mannequin utilized to the output layer of Gmail, however related in lots of extra contexts, that determines how and the place personalization will be completed utilizing that person’s cross-product knowledge footprint as enter.

Given the varied breadth of merchandise that Google operates, the corporate has an awfully privileged place in bringing personalization to bear for its customers. E mail building is an apparent and pretty tame software, however the use circumstances are practically countless. If Google’s instruments can really personalize output using the various volumes of information collected throughout its multitude of floor areas, then its AI instruments stand the most effective probability at offering an all-in-one agentic expertise. And thru its pricing mannequin, Google appears to be nudging customers on this course: the $249.99 month-to-month Google AI Extremely subscription is priced to preclude the potential for a client subscribing to a number of AI instruments.

A premium client pricing tier isn’t distinctive: ChatGPT Professional is $200 per thirty days, and Anthropic’s Claude Max is $100 per thirty days. And whereas different fashions can connect with a person’s Google Workspace account through API, which may provide learn entry to a person’s Gmail account, Google has the good thing about the totality of the information generated by its customers in logged-in state, throughout its properties in addition to Chrome. Google lords over a useful trove of diversified historic knowledge for every of its customers that’s incomparable in quantity and scope.

And clearly, Google’s capability to personalize AI output is instantly associated to its potential to personalize the advertisements uncovered to its prospects, with elevated engagement delivering extra knowledge to make use of for precisely that objective. Google launched AI Max for Search Campaigns, an automatic suite of concentrating on and keyword-matching instruments, in early Might, and the corporate launched Good Bidding, which adjusts bids dynamically, on the Google Advertising Reside occasion. These units of instruments share the identical basic objective — personalization — and they’re mutually reinforcing. Google’s potential to ship elevated engagement with AI Overviews and AI Mode supplies not solely extra alternatives to show advertisements, however as engagement throughout different floor areas deepens by personalization, the corporate gathers extra and higher knowledge for concentrating on advertisements.