Gemini is the name Google gives both to its model family and to the assistants built on top of it. Four of those assistants appear in the Generative AI Leader exam guide. The Gemini app is a standalone assistant on the web and on mobile, and Gemini Advanced is its paid tier. Gemini for Google Workspace puts the same assistant inside Docs and Gmail, and Gemini Enterprise is the organisational product that reaches into a company's own systems. The four are sold to different buyers, so a question describes a situation and expects one name.
The sections below take the four in the order a buyer meets them, from the app one person opens through to the organisational product. The last section sets out the two questions that tell them apart, one about who the user is and one about what the assistant can see.
The Gemini app
The Gemini app is Google's standalone assistant, on the web and on mobile. It drafts, summarises and analyses whatever a person supplies, and it both understands and generates images.
The app works from what a person gives it. Nothing inside the organisation's own systems reaches the app unless somebody pastes the material into the conversation. The business value is individual productivity. Adoption needs no project, since anybody with an account can start using it.
Gemini Advanced
Gemini Advanced is the paid consumer and business tier of the Gemini app. It gives access to the most capable Gemini models, a longer context window and Gems.
A Gem is a saved configuration. It holds the instructions, the tone and the reference material for one recurring task under a single name. The task then starts from the same setup every time. A team might keep a Gem for turning research notes into a summary in the house format. Nothing about the model changes. Anybody in a team can create a Gem without engineering help.
Gemini for Google Workspace
Gemini for Google Workspace puts the same assistant inside Docs, Gmail, Sheets, Slides and Meet. It drafts inside a document, summarises a long email thread, generates a slide and takes notes during a meeting.
The difference between Gemini for Workspace and the app is context. Gemini for Workspace can see the document a person has open and the files that person already has access to, so nobody has to describe the work before asking for help. The business value is that adoption requires no change of habit, which is usually the largest obstacle to a rollout.
Gemini Enterprise
Gemini Enterprise is the organisational product. It searches and reasons across a company's own systems, meaning its documents, its data and its applications. Gemini Enterprise respects the permissions each person already holds.
The exam guide names three capabilities. Multimodal search runs across company content, so a slide deck or a recorded meeting answers a question as readily as a document does, and much of what an organisation knows sits in decks and recordings a text search could never read. Custom agent capabilities let an organisation build agents that act on its own systems. The Gemini Notebook API brings notebook style research and grounded synthesis over a chosen set of sources into an organisation's own applications.
Permissions separate Gemini Enterprise from the app, since an assistant reading company systems has to show each person exactly what that person is already allowed to see.
Choosing between the four Gemini products
Two questions separate the four products, one about who the user is and one about what the assistant is allowed to see.
Where one person is doing general work and nothing internal is involved, the answer is the Gemini app, with Gemini Advanced added where the strongest models and Gems are worth paying for. When the whole organisation needs help inside the tools it already uses, the answer is Gemini for Google Workspace. Gemini Enterprise is the answer where a question involves the organisation's own knowledge, searched across its systems with each person's permissions respected and agents acting on what they find.
The most reliable signal in a question is a phrase about the company's own data or about the permissions people already hold. Either phrase points at Gemini Enterprise.