(Bloomberg) — Alphabet Inc. (GOOG)shares gained after the company said it will offer “AI mode” in search to all US users, showing its commitment to redesigning its core business to keep pace with new rivals in the artificial intelligence age.
“We want to get our best models into your hands,” Chief Executive Officer Sundar Pichai said Tuesday at the company’s developer conference in Mountain View, California. “So we are shipping faster than ever.”
Alphabet’s stock rose much as 5.6% on Wednesday after some analysts expressed confidence that the company can reorient its search product.
Google has been making profound changes to its search engine to ensure that its core product remains relevant as more people turn to chatbots for information. “More intelligence is available, for everyone, everywhere,” Pichai said at the event. “And the world is responding, adopting AI faster than ever before.” He said the company’s Gemini AI app has more than 400 million downloads, and that AI Overviews, the answers at the top of search queries, have more than 1.5 billion users.
AI mode is a tab in search where users can interact directly with AI models, much like OpenAI’s ChatGPT. Google has framed the product as a space for users to ask more complicated questions and drill deeper with follow-ups. Over time, Google will bring features from AI mode to AI Overviews, where it answers some questions on the main search page with AI.
“This really is building the future of search,” said Liz Reid, who leads Google Search, in a media briefing before Tuesday’s event. “It creates this time where you can truly ask anything, and that searching starts to feel really effortless.”
After decades of dominance, Google’s search empire is increasingly under threat. AI chatbot products from startups such as OpenAI and Perplexity, which use generative artificial intelligence to answer questions, are becoming a first stop for online queries. Earlier this month, an Apple Inc. executive said the number of Google searches in the iPhone maker’s Safari browser is beginning to fall. Meanwhile, a federal judge, after concluding that Google has an illegal monopoly in search, is mulling what penalties to impose.
Yet with roughly 90% of the search market, Google still has one big advantage over its rivals: its audience.
“Google Search is bringing gen AI to more people than any other product in the world,” Pichai told reporters in the briefing.
Google announced an assortment of updates to its suite of AI offerings, including the planned release this summer of Gemini 2.5 Flash, a faster version of its flagship model. It also said it’s testing a model called Deep Think that excels at coding and reasoning.
The company will bring AI agent capabilities — essentially creating digital assistants — to its Chrome browser, search and the Gemini app. Initially, the company said, Gemini in Chrome will let people clarify or summarize complex information on any website they’re browsing; in the future, Gemini will be able to work across tabs and even navigate websites for users.
Google also offered a glimpse of features that are coming for search. That includes Project Mariner, a web-based assistant that can present options from across the internet to users, such as the most affordable concert tickets. Google is also testing ways to help users ask more complex questions for sports and finance, creating charts that pull from real-time data, Reid said.
Many of those features will be offered first to people who pay to use Google’s products. Google has increasingly debuted AI features via a monthly subscription, and on Tuesday the company said it would introduce a $250 tier in its subscription plan, AI Ultra, for users who want more access to the latest technology. It represents a strategic shift for a company that has long made most of its money by offering tech for free, with a healthy dose of ads.
The fact that Google is devoting so much attention to search at its premier event is a sign of the times, said Arvind Jain, a former Googler who is CEO of search startup Glean, noting that the company traditionally used the conference to spotlight other products.
“This is the first time there is credible competition,” Jain said.
As Google moves deeper into what Google DeepMind leader Demis Hassabis called “personal, proactive and powerful” AI, it is freshening up tools like Deep Research, an AI assistant that uses web search and browsing to gather information and create custom reports. On Tuesday, the company said it would let people blend public materials with personal documents to produce more targeted reports.
Google showed it’s also making headway in an increasingly popular application of generative AI for artists and creators: AI filmmaking. On that front, Google is releasing a platform called Flow, which the company said will let people use natural language to describe movie shots as well as manage characters, locations and objects in a narrative. Flow will include not just Google’s latest video generation model but Imagen 4, the company’s image generation model, with upgraded capabilities like better spelling and typography in photos.
Hassabis also touted Veo 3, Google’s latest video generation model that can, for the first time, generate videos with contextual audio, like traffic noises in the background of a scene in the city or even dialogue between characters. “Veo 3 being audio and video at the same time is one of the holy grails of AI research,” he said.
Google sees these generative AI models as a way to make media creation more accessible to broader audiences, not just high-end professionals.
But as low-quality AI-generated media soars, and Google acts as a main purveyor of AI tools that make it ever easier to generate synthetic content, Eli Collins, a vice president of product management at Google DeepMind, said the company has a responsibility to ensure that it does not supercharge misinformation. “We’re not trying to build a slop machine,” he said in an interview. “We’re trying to actually build things that are useful for people who really want to bring a creative vision to life.”
Google in 2023 released SynthID, a watermarking technology that aims to help identify AI-generated media. The company also said it would release SynthID Detector, a portal that lets people upload content to determine if some or all of a file has SynthID in it.
More than two years after ChatGPT sent Google into a scramble, the company’s AI models top many leaderboards. Yet OpenAI’s models remain more popular than Google’s, and closing that gap will be the search giant’s next task, said Hassabis.
“More and more, as the models get more capable, you kind of want to get out of the way of the model in a way and let it shine,” he said.
(Updates with stock reaction starting in the first paragraph.)
