The AI-empowered future of Google Search diverges from common assumptions. Although the tech giant is working on a chatbot called Bard, its focus extends beyond just that. Transforming the homepage to mimic a ChatGPT-like messaging interface isn’t the path Google is pursuing. Rather, Google is integrating AI more prominently into its highly valuable existing search results, thereby innovating how users engage with online information.

The gradual transformation of Google’s search engine is unfolding three months after Microsoft’s Bing search engine began utilizing technology akin to what powers the AI-driven chatbot, ChatGPT. This technology has stirred some of the most significant excitement in Silicon Valley since the launch of Apple’s first iPhone 16 years ago.

Owned by Alphabet Inc., Google has already been trialing its own dialog-oriented chatbot, Bard. This product, driven by the generative AI technology that also propels ChatGPT, has only been accessible to individuals who made it off a waiting list. However, Google disclosed on Wednesday that Bard would become available to everyone in over 180 countries, expanding beyond English to other languages.

Bard’s expansion into multiple languages will kick off with Japanese and Korean, and approximately 40 more languages will be added subsequently. Now, Google is prepared to venture into the AI domain with its search engine, which has been the go-to tool for internet searches for the past two decades. It serves as the cornerstone of a digital advertising empire that produced more than $220 billion in revenue the previous year.

Google’s AI Transition

AI technology will further infiltrate Google’s Gmail through a “Help Me Write” feature that will swiftly craft extended email responses, as well as a “Magic Editor” tool for photos that will automatically enhance images. The AI integration will commence cautiously with the search engine, which is considered Google’s most prized asset.

This measured approach reflects the tightrope Google must walk as it strives to stay at the forefront of innovation while maintaining its status for providing dependable search results. This reputation could be undermined by artificial intelligence’s tendency to concoct seemingly authoritative information.

This propensity to generate deceptively plausible answers to queries — a phenomenon gently referred to as “hallucinations” — has already surfaced during the preliminary testing of Bard, which, like ChatGPT, is built on the still-developing generative AI technology.

Google will embark on its forthcoming AI ventures via a newly established search lab. Here, U.S.-based individuals can join a waitlist to evaluate how generative AI is integrated into search outcomes. These trials will also encompass the conventional links to external websites, where users can delve deeper into the subjects they’re investigating. Google may require several weeks before it starts dispatching invitations to those selected from the waitlist to experiment with the AI-enhanced search engine.

The AI-generated results will be conspicuously marked as an experimental technology. Google assures that these AI-crafted summaries will exude a more factual tone rather than a conversational one — a stark deviation from Bard and ChatGPT, which are designed to emulate more human-like personas. Google is implementing protective measures to prevent the AI integrated into the search engine from answering sensitive queries concerning health — for example, “Should I administer Advil to a 4-year-old?” — and financial matters. In such cases, Google will continue to direct individuals to trusted websites.

An Inside Look at Google’s New AI-Integrated Search Engine

Liz Reid, Google’s VP of Search, demonstrated the new AI-integrated search engine for The Verge, an American technology news website operated by Vox Media. Liz opens her laptop and starts inputting a query into the Google search bar. She types, “Why is sourdough bread still so popular?” and presses enter. The standard Google search results appear almost instantly. Above them, a pulsating, orange rectangular section illuminates, displaying the words “Generative AI is experimental.” After a few moments, the glowing transitions into an AI-crafted summary: several paragraphs describing the appealing taste of sourdough, its prebiotic benefits, and more. To the right, three links point to websites with information that Reid indicates “substantiates” the contents of the summary.

Google refers to this as the “AI snapshot.” The entirety of the information is generated by Google’s large language models, with all content derived from the public web. Reid then moves her cursor to the upper right corner of the box and clicks on an icon that Google’s designers have nicknamed “the bear claw.” It resembles a hamburger menu with a vertical line on its left. Upon clicking the bear claw, a fresh view is presented: the AI snapshot is now dissected sentence by sentence, with links to the sources of information specific to each sentence situated underneath. Reid reiterates that this feature exemplifies corroboration, a critical aspect that distinguishes Google’s AI implementation. “We desire the LLM to divulge its sources when it provides information, indicating where one could learn more about the topic.”

A few moments later, Reid navigates back and initiates another search. Now, she inquires about the best Bluetooth speakers suitable for beach outings. Once again, conventional search results materialize almost immediately, followed shortly by AI-derived outcomes. This time, a concise summary is positioned at the top, explaining the essential factors to consider for such a speaker: battery longevity, water resistance, and sound quality. Links to three purchasing guides are aligned to the right, and beneath them are shopping links to about six excellent options, each accompanied by an AI-produced summary. When asked to refine the search with the phrase “under $100,” Reid complies. The snapshot refreshes, showcasing new summaries and selections.

Are there any downsides to Google’s AI-integrated search engine?

There are a few stipulations: in order to utilize these AI snapshots, users must consent to a novel feature named Search Generative Experience (or SGE), which is a part of a newly introduced function called Search Labs. Not every search will trigger an AI response — the AI only makes an appearance when Google’s algorithms perceive it as more valuable than traditional results, and sensitive topics like health and finances are currently designed to circumvent AI involvement altogether. While AI might not be eliminating the traditional 10 blue links, it’s undeniably relegating them to a lower position on the page.

Google executives repeatedly emphasize that SGE is experimental. However, they also assert their belief that it represents a fundamental, long-term transformation in the manner in which people search. AI provides an additional stratum of input, assisting in formulating more comprehensive and insightful queries. Moreover, it introduces another level of output, devised to both answer your questions and lead you to new inquiries.

While the introduction of an opt-in box at the top of search results may seem like a minor adjustment from Google, especially when compared to Microsoft’s AI-driven Bing redesign or the entirely novel ChatGPT, SGE constitutes the initial step towards a thorough reimagining of how billions of individuals access information online — and how Google generates revenue. As far as internet pixels are concerned, these are as momentous as they come.