This will come as no surprise: Google is relying more than ever on AI to maintain its strong position in the online search market. And to achieve this, the company has announced several new features related to AI Overviews, these famous answers generated (sometimes haphazardly) by Gemini, but also by experimenting with an “AI mode”.
The era of generative AI has not opened without pain at Google. The web giant was overtaken by OpenAI and since then, it has been working hard to try to stay in the race. Google is of course still the most used search engine in the world, but it only takes a little something to jam the machine.
An AI mode that searches for you
For example, the shoehorning of generative AI into search results. The famous AI Overviews responses were much talked about at the beginning, advising Internet users to eat small stones or to put glue in the pizza sauce. Google has tightened a few screws behind the scenes, and relaunched AI Overviews almost everywhere in the world, with the notable exception of Europe.
Google has just added another layer to this cumbersome and not always very relevant function, by announcing the launch of Gemini 2.0 for AI Overviews in the United States. The advanced capabilities of the AI model should make it possible to generate more complex answers to questions containing code, mathematical operations or multimodal requests (text, audio, video). These answers will also be offered to teenagers, and it will no longer be necessary to log in to a Google account to consult them.
By dint of insisting, AI Overviews will eventually become established: more than a billion people use them, Google rejoiced. It’s not as if we had a choice.
Another new feature: an AI mode, which offers more advanced answers with improved reasoning and analysis capabilities, all powered by Gemini 2.0. This involves exploring topics in depth to obtain detailed answers, accompanied by the appropriate links. Google here combines its AI with its real-time information systems (Knowledge Graph, shopping data, etc.). And it is possible to ask questions about the results to go further.
This experiment is currently being tested with a limited number of users, with initial access reserved for Google One AI Premium subscribers. The goal is to improve the quality of responses and integrate richer formats (images, videos) based on feedback from testers.