Google Gives Gemini Read-Only Access to Wallet Spending Data

Google announced on November 13, 2026 that all Gemini app users in the United States can now connect Google Wallet, letting the assistant answer personal-finance questions using their real spending data. The integration is read-only: Gemini can analyze transactions but cannot move, manage or invest money. That boundary is the load-bearing wall of the entire feature, and Google repeats it at every layer of the announcement.

Users invoke the feature by tagging @Google Wallet in a chat, on the web or in the Gemini mobile apps for Android and iOS. Wallet joins Gmail, Photos, YouTube and Search on the list of apps Google makes available through Personal Intelligence, its connected-app personalization system. Google expanded that system on March 17, 2026 across AI Mode in Search, the Gemini app and Gemini in Chrome. The connected experiences cover personal Google accounts only; Workspace business, enterprise and education users remain outside the program.

What can Gemini actually do with your transactions?

The pitch is analysis, not action. Google's examples center on questions a spreadsheet could answer, delivered conversationally. A user reviewing the previous quarter's spending while aiming to save $1,000 over three months can receive dynamic in-chat charts covering cash flow, asset allocation and budgets.

Gemini can also find recurring charges across linked accounts and flag subscriptions whose prices increased during the year. Another cited use case: examining the previous year's gift and travel transactions to suggest a personalized holiday budget.

Wallet data alone can inform those responses. For deeper insights, users can link credit-card and bank accounts to Wallet through Plaid, the third-party financial-data intermediary. Google frames Plaid linking as optional, but the phrasing concedes the point skeptics will press: the baseline Wallet connection offers less than the full picture. The takeaway is that the feature's usefulness scales with how much financial data a user volunteers.

Read-only by design — with a separate disclaimer

The integration is opt-in through Connected Apps in Personal Intelligence settings, and users can switch it off at any time. Gemini receives read-only access. It cannot initiate payments, manage accounts or invest funds.

Google's own Connected Apps documentation adds a caveat the marketing copy does not foreground: users should not rely on Gemini's output for financial advice and should consult a professional. A generative assistant summarizing a household budget is not a fiduciary, and Google says so in writing.

The scope of what gets shared is broader than transactions alone. Shared Wallet information can include payment methods, data from Google services saved in Wallet, and financial-account information linked through services such as Plaid. Data added to a connected app in the future also becomes accessible to Gemini automatically — a forward-looking permission users grant once and may not revisit.

What happens to the data afterward?

Here the story shifts from convenience to training policy. When Keep Activity is on, Gemini Apps activity — including data from connected apps — can be used to improve Google services and train generative AI models.

Google draws a careful distinction. It says it does not train those models directly on Google Wallet and payments services. It does train on summaries, excerpts and inferences derived from relevant transactions, plus prompts, responses, suggestions and information about how those responses and suggestions were generated. In practice, the raw transaction record stays out of training; the assistant's derivative output built on it does not. Whether that line feels meaningful depends on how much of a user's financial life lives inside those summaries.

A subset of Wallet-related interactions, summaries, excerpts and inferences is reviewed by human reviewers, including trained service providers, for debugging, safety and service improvement. Google says that before trained service providers review interactions personalized through Connected Apps, it disconnects them from Google Accounts and reduces personal information — unless it is addressing abuse or harm.

Deletion is also split across layers. Disconnecting Wallet, or deleting Wallet data, does not delete Gemini Apps Activity. Users must manage or delete that activity separately, and deleting it can prevent its use for future product improvement only unless trained reviewers have already reviewed the data. The caveat: once a human reviewer has seen the material, deletion loses much of its force. The plain-word takeaway is that turning the feature off and erasing its footprint are two different tasks.

Memory Summary: an editable record of what Gemini knows about you

Alongside Wallet support, Google introduced Memory Summary, a readable snapshot of saved preferences, interests and ongoing goals. The Memory Summary help page says users can add details, correct inaccuracies, delete information and inspect a chronological Memory history showing when each piece of information was added or modified.

That transparency mechanism matters because Wallet and Memory work in combination. Gemini now holds transaction information alongside an editable record of preferences and goals — the raw material for genuinely personalized financial guidance, or for a much richer behavioral profile, depending on your vantage point.

Eligibility comes with conditions: a personal account, an age of at least 18, Keep Activity enabled and Memory turned on. Memory Summary is available on the web and mobile wherever Personal Intelligence is offered, but not in the European Economic Area, Nigeria, Switzerland or the United Kingdom. The geographic carve-out mirrors the pattern of features that depend on broad data retention, which GDPR-style regimes make harder to ship.

Why this matters

The finance experience rests on a layered privacy model, and each layer has its own switch: the initial connection, activity retention, model training and saved memory. Google has kept money movement outside the chatbot's reach while pulling money visibility inside it — a defensible design, but one where every promise of control arrives with an asterisk about training, review or separate deletion paths.

The commercial logic is straightforward. Personal finance is among the highest-engagement, highest-stakes use cases for an assistant, and Google is claiming it without the regulatory weight of moving funds. The open question is whether users treat the distinction between analyzing money and managing it as firmly as Google's architecture does.