Why Google Gemini Is Emerging as a Hot New AI Tool for Startups

For years, OpenAI held the default position in most startups’ tech stacks. It was the tool founders reached for when they needed a language model, a voice engine, or a general-purpose AI backbone. But for some startups, Google’s Gemini AI has emerged as a newly preferred productivity tool, and their reasons for adopting the tech go well beyond the technology itself. Google is in the midst of an aggressive push to convince startups that its AI solutions are superior. Leading that charge is Darren Mowry, head of Google Cloud’s global startup team. Mowry confirms that yes, Gemini use is rising among startups, and it’s resulting in new business for Google Cloud, which is the only way for businesses to use the Gemini API. Instead of just automatically selecting Amazon Web Services as their cloud provider, Mowry says, new startups are now choosing Google Cloud in part so they can get access to Gemini. Google has always been central to the AI business; in 2017, the company released a seminal AI research paper called “Attention is All You Need,” which introduced the “transformer” architecture that makes modern AI models possible. But up until last year, the company lagged behind its competitors when it came to business adoption. When Gemini was launched in 2023, it was originally named Bard, and it quickly developed a reputation for hallucinating facts. Remember when it recommended that people put glue on pizza? That changed in April 2025, when Google released Gemini 2.5 Flash, a model that handily beat OpenAI across a number of benchmarks, and according to Mowry, ignited a wave of interest in Google’s AI offerings that has only grown with the release of subsequent models. One of the factors that differentiates Google from its competitors is that it offers a fully vertically-integrated solution, in which the company can handle each part of the tech stack. Not only can startups choose from a wide selection of Google-made and external models on Google Cloud, Mowry says, but the company can also provide those startups with technical assistance to help make sure they’re getting the most out of both the models and the Google-made chips they run on. According to Mowry, this vertical integration “shrinks down the time” it takes for founders to build.” Some founders are finding that Gemini is a useful way for non-technical employees to enjoy the benefits that software developers have gotten from agentic coding tools like Claude Code. Aakash Shah, founder and CEO of allergy care startup Wyndly, says that while his engineers have gravitated toward Anthropic, his operations team wanted to use Gemini in the applications they’re already comfortable with, like Google Docs and Gmail. A common use case? Asking Gemini “who did I email on such-and-such day?” Shah says that everyone at his company now has Gemini, enabling them to chat with Gemini across the entire Google Workspace suite, including Gmail, Docs, Sheets, Meet, and Notebook LM, Google’s app that turns documents into audio podcasts. “I’m trying to get everyone to be AI-first,” Shah says, “and part of that is helping them use it where they already are instead of forcing it on them.” Sheltered International, a customs broker and freight forwarder that primarily deals with international imports, is currently using Gemini to help speed up the process of filling out customs paperwork. Founder Andrew Ciccarone says that when a shipment comes in, his company is responsible for verifying its commercial invoice data and ensuring that it’s marked with the correct Harmonized Tariff (HTS) code. According to Ciccarone, Sheltered International has started using a fine-tuned Gemini model to extract relevant information from the commercial invoice data (which often comes in the form of a PDF), validate it, and reformat it into an Excel spreadsheet. The Gemini models are considered to have state-of-the-art computer vision, enabling them to examine images and documents in incredibly granular detail. “What the AI can do is just give us a huge leap forward before the customs broker comes in to ensure everything is classified correctly,” says Ciccarone, adding that for a small operation handling the complexity of international trade documentation, Gemini has streamlined a process that used to be painfully manual. When this process was fully done by humans, it required hours of manual scanning through lengthy, unformatted documents. Still, Ciccarone admits that the company isn’t saving that much time yet, because employees still need to verify that the AI’s output is accurate. But as the fine-tuned Gemini model improves, he expects to see a significant increase in productivity. Companies are also integrating Gemini’s machine vision abilities into their actual products. Take Validity, a startup that sells itself as an all-in-one solution for the entire email marketing process. Validity chief technology officer Matt Gore says that the company’s newest product, a platform called Validity Engage, has been largely built around Gemini’s capabilities. Engage gives marketers access to four purpose-built AI agents that can analyze, optimize, and reformat emails according to internal campaign style guides. Using Gemini 3 Pro, Validity can now detect and fix granular visual details in emails (like whether a certain font matches a brand’s approved style guide) that no model could reliably catch a year ago. For instance, Gore says, emails can often appear illegible to computers and phones that are set to dark mode; with Validity Engage, marketers can guarantee that their email will be visible to everyone who received it. Geo says that Validity decided to “hitch our wagon to Gemini” following extensive testing. When developing the new feature, Gore used an orchestration tool called Mastra.ai to compare how various models approached roughly 150 common email issues, taking note of each model’s cost and speed. In the testing, Gore says, Gemini 3 Pro stood out as being “just leaps and bounds ahead of others in terms of computer vision.” He says that the Gemini 3 models are particularly good at identifying the “bounding box” of the email—basically the frame containing the email’s content. Beyond computer vision and text, Gemini’s speech capabilities have been a major selling point for founders. David Yang, founder of solopreneur-focused AI receptionist company Newo, is using Gemini to provide his virtual receptionists with voices. Yang founded Newo to help solo founders capture more inbound leads by giving everyone access to an always-on receptionist who can answer a phone call at any hour of the day. But for Newo to work, Yang needed voice models that have extremely low latency and high emotional intelligence. Originally, the company’s AI receptionists were powered by OpenAI’s text-to-speech and speech-to-text models, but the lag between asking a question and hearing an answer was too long. Now, Yang says that Newo uses Gemini 2.5 Flash Native Audio, a recently-released model that can understand and generate audio in real time. Not only is the new model incredibly fast, Yang says, it can also understand emotional intent, an important data point that’s usually lost with more traditional speech-to-text transcription models. As part of his push to bring startups into the Google ecosystem, Mowry says his team is currently hiring engineers and former founders to staff up a “founder advocacy” group. These employees “sole purpose in life” will be “to wake up and meet founders that have really big problems,” he says, and “help them move from ideation into actually getting things built.” The goal is to “catch these cohorts of startups early, give them a little bit of credit assistance, engineering assistance, and help them get off the ground.” This soup-to-nuts approach is helping Google win startup business, and positioning the company as the new default AI partner for the next generation of businesses. BY BEN SHERRY @BENLUCASSHERRY