One name, two habits
"How much have I spent on Uber" is two questions. Rides are transportation: commuting, airport runs, nights out. Uber Eats is food. They tell you different things about your spending, but both land on your card statement under the Uber name.
That matters when an AI agent answers the question. BankBridge's get_merchant_historytool matches the text you give it anywhere in the merchant name or bank description. "uber" is inside "Uber Eats", so a search for Uber returns both. That's a feature when you want the combined number and a trap when you don't.
The split in the demo account
From the BankBridge demo account (a made-up household, fictional data), last 12 months, pulled 2026-10-15:
| Search term | Description on the statement | Charges | Total | Average |
|---|---|---|---|---|
uber | Both of the below | 60 | $1,406.65 | $23.44 |
uber trip | UBER TRIP | 48 | $1,136.57 | $23.68 |
uber eats | UBER EATS ROSA'S CAFE | 12 | $270.08 | $22.51 |
The averages are almost identical, so you can't tell rides from food by amount. You need the description. Your own statements may word it differently, so start by asking the agent what it found:
List every distinct description that matches "uber" in the last 12 months, with a count and total for each.
Prompts for each question
- Rides only, this year:“How much have I spent on Uber rides since January 1? Leave out Uber Eats.”
- Food only:“How much have I spent on Uber Eats this year, and which restaurants appear most?” Eats descriptions can include the restaurant, as the demo's does.
- All delivery apps:“Uber Eats plus DoorDash, by month, for the last 12 months.” See the DoorDash guide for the restaurant breakdown.
- All rideshare:“Uber rides plus Lyft, by month.”
- Last year:“Uber rides and Uber Eats for January through December of last year.” Banks provide up to 24 months.
Ask in Claude · Ask in ChatGPT (with BankBridge connected)
Give dates when you mean the calendar year. The tool's default window is the last 12 months.
Rides you can expense
If some rides were for work, the bank data narrows the search fast:
List my Uber rides between March 3 and March 7 and on any weekday before 7am this quarter.
The agent returns dates and amounts. Match them to Uber's receipts for the report itself. Our guide on expense reports with an AI agent covers the rest of that workflow.
When you need trip-level detail
Bank data answers "how much" across every card. It doesn't answer "where to". Uber offers a personal data download that, according to its help center, includes details on each trip with times and locations, and Uber Eats order details including items, prices and order times (Uber Help, "Download your Uber personal data", checked 2026-10-15). Use that for detail and the bank side for totals across every card you've paid with.
Things that throw the total off
- Pending charges.A ride or order that hasn't posted is included, and its amount can change, for example after a tip. Say "posted only" to match your statement.
- Refunds and credits.These come back as negative amounts and aren't subtracted from the tool's total. Ask for the net.
- Memberships.If you pay for an Uber membership, it can match "uber" too. Ask the agent to separate same-amount monthly charges.
- Cards you haven't connected.Rides on a work card or a partner's card only count if that account is connected.
Want to see it before connecting anything? The demo account has a year of a made-up household’s transactions. Add it to your agent from the demo dashboard, then ask: “How much have I spent on Uber rides and on Uber Eats in the last 12 months, separately?” (what’s in the demo)
On your own accounts, BankBridge is read-only and fetches live from your bank on every question. It costs $5 a month or $50 a year per connected bank.