Google is still attempting to get you to utilize its AI Mode when conducting online searches by introducing new and improved AI features. Deep Search, a tool for creating long-form reports, and Google’s Gemini 2.5 Pro AI model are now accessible in AI Mode through the conversational search tool.
Gemini 2.5 Pro will now be an option for Google AI Pro and AI Ultra subscribers in the United States who are also participating in the AI Mode experiment in Search Labs when posing challenging queries.
The most sophisticated AI technologies from Google are based on the same robust paradigm. Additionally, they will be able to use Deep Search, a function in the standard Gemini app that can run hundreds of searches at once and compile the results into a report.
The more significant shifts, however, are in the way Search is developing. Gemini 2.5 Pro does more than simply retrieve answers. It gives a rationale. It provides step-by-step arithmetic explanations. It even generates code and explains its actions. Additionally, it may almost carry out a research marathon on your behalf when combined with Deep Search.
AI Phoning
Search’s new call function is a whole different experience. It uses AI to link your phone call to a store’s information search. You can search for “pet groomers near me” and request information that isn’t immediately available, as Google demonstrates in a demo.
Alternatively, you can select “Have Google call for you,” which will cause Google to contact nearby stores, inquire about availability or prices, and then send you the information via text or email. It is powered by Google’s Duplex technology, which is why it sounds like Google Duplex.
Naturally, there are some caveats to all of this, like the requirement to pay for a subscription. The powerful AI Mode tools are only available to paying clients, however free users still receive a limited number of AI call tries.
Even when OpenAI and other AI developers try to create comparable solutions, Google may have a sustained advantage due to the scale of its Search database. There is a learning curve, though.
When AI does too much too quickly, it might cause issues. Your research helper shouldn’t choose a Reddit thread with 38 upvotes above reliable sources. Furthermore, you don’t want your AI to phone the incorrect person to arrange a colonoscopy.
