Google Launches Tabletop, Shared AI Spaces That Build Group Tools Live

Google introduced Tabletop on November 11, 2026, a Google Labs experiment that generates custom AI tools inside shared spaces where groups plan, play and coordinate in real time. The initial rollout runs through a waitlist, is restricted to US users aged 18 and over, and admits new tables every week on an invite-only basis.

The mechanics are direct. Users supply links, files and ideas to a shared space, then address a prompt to @Tabletop. The system converts that prompt into an interactive tool that lives inside the same space where the group has already gathered its information. Starter templates offer a second entry point for users who prefer not to build from a blank prompt.

The stakes for Google are straightforward. Generative AI products have concentrated on individual assistance — writing, coding, search — while group coordination has remained scattered across chats, spreadsheets and polls. Tabletop tests whether an AI system can generate and continuously revise the tools a group uses, rather than only the content a group consumes. That is a distinct product thesis, and Google has placed it in Labs, its established home for experimental releases with uncertain commercial futures.

What can Tabletop actually build?

The announced examples span itineraries, expense splitters, dinner polls and tournament brackets. Each tool is generated in place, inside the space the group already shares, rather than requiring a separate app or export step.

Collaboration begins when the creator distributes an invite link through a group chat. Multiple participants can then work in the space concurrently. Google likens the experience to the shared editing model of a Google Doc, where several people operate on one document at the same time. Voting, planning and tracking all happen live inside that single environment.

The road-trip example bundles several functions into one space. It combines route mapping with votes on roadside stops, shared-budget splitting and a spinning wheel that selects a dinner location. The demonstration illustrates the core argument: one space can hold the map, the money and the decision, and every participant sees the same current state of the plan.

How do the tools change as plans do?

Generation is only the starting point, according to Google's description of the product. Participants can ask @Tabletop to change a rule, reshape a tab or build something new as plans evolve. The system modifies the group's existing tool — a game, a chart or another planning aid — keeping it aligned with the activity without a fresh start.

The official Tabletop product page demonstrates this with a request that turns a Pong tab into Pickle Pong. The modification replaces the ball with a bouncing pickle and adds a live court scoreboard. A second demonstration makes a seating chart interactive, so guest badges can be dragged between tables.

Both examples make the same structural point. Revision is part of the collaborative workflow, not a separate technical step delegated to whoever originally built the tool. In conventional group software, changing a poll's options or a bracket's structure usually means rebuilding it. Here the change is a conversational request any participant can issue.

What use cases does Google name?

Google organizes the examples around trips, clubs, parties, homes and projects. The categories differ, but the functions repeat: collecting preferences, recording decisions and giving participants a current view of the plan.

• Trips: route mapping, votes on roadside stops, shared-budget splitting and a dinner-selection wheel, all in one space.
• Clubs and leagues: live rosters, tournament brackets, shirt polls and mini-games.
• Event spaces: venue voting, catering tracking and interactive seating charts with draggable guest badges.
• Households: weekly meal coordination, school pickup organization and research on major purchases.
• Chores: a tracker that becomes a game with point bounties, streak multipliers and a shared reward goal.
• Projects and study: exam preparation, volunteer cleanup events and workplace launch planning. The cleanup example adds an interactive tide timeline for choosing volunteer shifts.

The chore-tracker example deserves attention because it shows the product's gamification direction. Google describes point bounties, streak multipliers and a shared reward goal — mechanics imported from consumer games, applied to household coordination. The tide-timeline example shows the opposite pull, toward practical scheduling constrained by real-world conditions.

Who gets access, and how fast?

Access is narrow at launch. The rollout is limited to the United States, restricted to users aged 18 and over, and described by Google as invite-only. New tables open every week as waitlisted users are admitted. That schedule provides a gradual route to wider access rather than an open launch, and it matches the pattern Google has used for other Labs experiments: constrained supply, staged admission, iteration on observed behavior.

The weekly-admission model also has a practical function. Tabletop spaces are shared objects, and Google is admitting groups, not just individuals — a creator must distribute an invite link through a group chat for the space to fill. Staged admission lets Google watch how multi-participant generation behaves under real usage before scaling.

Why does this experiment matter?

Tabletop's central test, as the announcement frames it, is whether tools that can be generated and revised this quickly can keep people synchronized as their plans change. Every group has felt the failure mode the product targets: the itinerary in one chat, the budget in another, the poll in a third, and no single view of what the group actually decided.

The competitive context matters here. Google already operates the most widely used collaborative editing suite in Docs, Sheets and related products. Tabletop extends that collaboration model to AI-generated tools that did not exist before the group needed them — and that mutate as the group's requirements do. If the experiment graduates from Labs, it would give Google a consumer-facing answer to the question of what generative AI does for groups rather than individuals.

For now, the answer is partial and waitlisted. The demonstrated uses span trips, clubs, parties, homes and projects; the revision model is proven in demos such as Pickle Pong and the interactive seating chart; and the admission cadence is weekly. Whether synchronous groups adopt AI-generated tools at scale, and whether those tools stay coherent under constant revision by multiple participants, is the question this experiment is built to answer.