This article was AI-generated as part of an experimental historical-content project. The date reflects the period being analyzed rather than the date the article was originally written.
The number in Google’s Gemini announcement that I keep coming back to is 40%, and it has nothing to do with benchmarks.
What Google announced
On December 6, Google and Google DeepMind introduced Gemini 1.0, a model family in three sizes: Ultra, Pro, and Nano. Google says Gemini Ultra exceeds the previous state of the art on 30 of 32 widely used academic benchmarks and scored 90.0% on MMLU, which it describes as the first time a model has outperformed human experts on that test.
The product news was more immediate. Bard now runs on a tuned version of Gemini Pro in English, in more than 170 countries and territories, with a “Bard Advanced” tier using Ultra promised for early next year. Gemini Nano is coming to the Pixel 8 Pro. And Google said Gemini will reach Search, Ads, Chrome, and Duet AI “in the coming months.”
Then the line that caught my attention: Google is already experimenting with Gemini inside its Search Generative Experience, where it has made SGE 40% faster in English in the U.S., alongside quality improvements.
Why speed is the real story for reputation
A generated overview that takes several seconds to appear is competing with a user’s instinct to scroll down to the links. Make it appear faster, and the summary gets read first more often. That quiet mechanism is probably how generative search moves from something people opt into toward something they simply encounter: friction disappearing step by step, with no single dramatic switch.
So my prediction, offered as a prediction: over the next year, more people will form a first impression of a company or executive from a machine-written summary, before they read anything the company wrote or anything a journalist reported.
There are real reasons for caution. SGE is still a Search Labs experiment, and Google has not said whether or when it becomes the default. The benchmark results are Google’s own, and Ultra is not yet publicly available. None of this happens overnight.
What multimodality adds
Google describes Gemini as natively multimodal, trained from the start across text, code, audio, images, and video. Today Bard’s Gemini upgrade is text-only, with other modalities “coming soon.” Over time, though, the material that feeds a summary of you may include conference talks, interview clips, and earnings-call recordings, not only articles and web pages. An executive’s spoken record could become part of the input in a way it largely has not been for search.
It is also telling that Google’s own announcement lists “factuality, grounding, attribution and corroboration” among the known challenges it is still addressing. That echoes the double-check feature Google added to Bard in September: the company knows these summaries need to be anchored to evidence, and that the anchoring is unfinished.
What to do while it is still early
Check how Bard, now running Gemini Pro, describes your company and leadership, and note where errors likely originate. Keep the basic facts consistent in the places search systems and models lean on, including your own site, credible coverage, and encyclopedia entries. If you operate across markets, remember that generative search now reaches more than 120 additional countries and the source pool differs by language. And pay attention to the accuracy of public video and transcripts, which have mostly been treated as an archive rather than a source.
Search will not be replaced in a quarter. The question of who summarizes you first, though, is gradually shifting from editors and journalists toward models. Those models just got noticeably faster.