Google now ships more than 30 AI products and experiments, and even experienced users struggle to work out which one is meant for which job. The tools themselves are generally capable; the common mistake is using the wrong one for the task. This guide groups the main products by purpose so that the lineup reads as an organized kit rather than a tangle of overlapping names.
The center: Gemini and Deep Research
Gemini, the chat assistant, is the hub of the ecosystem. Most people treat it as a general chatbot, but its most distinctive capability is Deep Research, a mode that investigates a topic across many sources and returns a structured, cited report. Paid tiers of Gemini give access to Google’s most capable models and a context window of up to one million tokens, enough to load entire books, long reports or large datasets in a single session, which makes it a credible research assistant rather than just a conversational tool. Whether it outperforms rivals such as ChatGPT or Claude on a given research task depends on the task; independent benchmarks vary.
The all-purpose layer: NotebookLM, Gems and Opal
NotebookLM is best known for its generated podcasts, infographics, slide decks and video overviews, but its defining feature is that it answers only from the sources the user uploads. Because it is grounded in those documents rather than in general training data, it is far less prone to inventing facts, which makes it well suited to studying, due diligence and working through long document sets.
Gems are customized versions of Gemini configured for a recurring task, such as drafting proposals, planning a quarter or writing in a specific house style. Once set up, a Gem keeps its instructions across sessions. Opal, from Google Labs, is a lightweight automation builder that chains AI steps into small apps or workflows, in the spirit of Zapier or Make but with far less setup; it does a few things well rather than everything.
Both NotebookLM and Opal started as Google Labs experiments. If there is one habit worth adopting, it is checking labs.google periodically, since that is where Google tests ideas that often become its next mainstream products.
For developers: AI Studio and its companions
Google AI Studio causes the most confusion. It began as a testing console for developers working with the Gemini API, but non-developers now use it to experiment with models, prompts and multimodal inputs. It is a place for prototypes and experiments, not for building finished software. Around it sit Firebase Studio for app development, Jules, an asynchronous coding agent that works on GitHub repositories, Stitch for generating interface designs, Gemini Code Assist in IDEs and the Gemini CLI for the terminal. For demanding coding work, many developers still prefer Cursor or Anthropic’s Claude-based tools, although Google is closing that gap quickly.
Creative and media tools
Google’s image and video models drew the most attention in 2025. Nano Banana, the nickname for its Gemini image model, made precise editing and consistent characters across images practical for everyday users. Veo 3 added native audio and dialogue to generated video, raising the bar for AI video. Flow is Google’s filmmaking workspace built on these models, and Whisk is a lighter experiment for remixing images. For musicians, the Music AI Sandbox is an early-stage experiment worth watching.
Models on devices and AI inside everyday products
A separate group of models is designed to run locally or on small hardware: Gemini Nano on phones, the open-weight Gemma family, and Project Astra, Google’s research effort toward a real-time multimodal assistant. Gemini is also embedded across Google Workspace, with AI features in Gmail, Docs, Sheets and Slides, and similar models power features in Maps, Photos, YouTube and Lens. On Android, Gemini has become the system-level assistant.
How to approach it
No one needs to learn all of these. The useful step is to understand which tool was built for which problem. Keep an eye on Google Labs, spend time learning AI Studio if you build anything, make Deep Research and NotebookLM the defaults for research and study, and try Nano Banana and Veo for creative work. Seen that way, Google’s AI lineup stops looking like an overgrown forest and starts looking like a toolkit with a clear place for each tool.
