Use ChatGPT to understand your electricity bill

An electricity bill contains several different stories: how much energy the meter recorded, what the supplier charged for it, and how payments changed the account balance. An AI assistant can help put those parts into a readable table. The most useful result is an explanation you can check against the bill, with the original figures still visible.[1]

This guide uses ChatGPT’s documented file-analysis features and an original, deliberately inconsistent practice bill. We calculated the example independently; we did not test an uploaded household bill in a signed-in ChatGPT account. The exercise concerns single-rate domestic electricity in Great Britain.

Give the assistant the bill’s numbers and dates

ChatGPT supports document and text uploads on Free and paid plans, subject to account limits, on the web and supported mobile apps. Attach the relevant PDF or paste its text into the conversation. A short bill does not need a special-purpose agent: name the information you want extracted and ask follow-up questions about it.[2]

Remove your name, address, account number, payment details and supply identifiers from the working copy. Retain the billing dates, reading dates, units, rates and balance labels. OpenAI’s Data controls setting lets you turn off Improve the model for everyone for new conversations; this does not delete the conversation or remove it from history.[3]

Prefer selectable text from the bill to a screenshot of tiny numbers. OpenAI warns that exact-value extraction from scans and complex image-based tables can be unreliable. If you type the figures yourself, check them back against the original before asking for arithmetic. One missing decimal point can make a perfectly executed calculation useless.[1]

A practice bill with one inconsistency

Imagine a fictional bill for 1–30 September 2026. Its actual meter readings are 12,480 kWh at the start of September and 12,720 kWh at the start of October. It nevertheless bills 250 kWh. Its invented prices are 25p per kWh and 50p per day, both before VAT. There are 30 charged days, £40 opening credit and an £80 payment. No other adjustments apply.

The meter difference is 240 kWh: 12,720 minus 12,480. The printed bill has therefore charged for ten more units than its stated readings imply. Its remaining arithmetic can be internally consistent while its starting quantity is wrong. Ask the assistant to retain both versions so you can see the disagreement.

The reconstructed charge is £2.63 lower. That figure includes the different rounded VAT amount. On a real account, return to the bill before calling this an overcharge: a copied reading might be wrong, a rate period might be missing, or an adjustment might appear elsewhere. The example explicitly rules those possibilities out to make its arithmetic inspectable.

Tax belongs to the billing period

The September example uses the earlier 5% domestic rate. HMRC’s dated measure introduced a temporary zero rate for qualifying domestic electricity supplies in England, Scotland and Wales from 1 October 2026 to 31 March 2027; other domestic fuels remain at 5%. Giving an assistant the country and supply dates prevents it from applying a remembered tax rate to the wrong period.[5]

Also distinguish prices before VAT from prices that already include it. If a bill gives an inclusive unit price, adding the tax again would count it twice. Ask the assistant to reproduce the bill’s stated basis before doing any multiplication. Where a bill crosses a price or tax change, keep its separate dated lines separate.

A payment is different from a charge

The £80 direct debit in the example does not establish the cost of September’s electricity. Fixed monthly payments commonly spread estimated annual energy spending across the year, and a payment can also reflect an account balance. Citizens Advice recommends asking the supplier to explain a changed direct debit and the readings behind it.[7]

For the corrected example, £40 opening credit plus £80 paid minus £78.75 new charges leaves £41.25 credit. Label that number as credit throughout. If an assistant presents the same positive figure as money you owe, the arithmetic may look right while the meaning has reversed. Ofgem explains that paying more than the energy used can build a credit balance.[4]

For comparisons between bills, request kilowatt hours per day as well as pounds. In this exercise, 240 kWh over 30 days is 8 kWh a day. A longer billing period can produce a larger total without higher daily use. Keep actual and estimated readings identifiable; an actual reading after several estimates can reveal earlier consumption that had been understated.[6]

Turn the result into a precise question

The practical output for this invented case is a short query: “Your bill shows readings of 12,480 and 12,720 kWh, a difference of 240 kWh, but the consumption line charges 250 kWh. Please explain the additional ten units and how the corrected reading affects the charges and closing credit.” It specifies the discrepancy without asking AI to invent a cause.

On a real bill, keep the relevant pages and dated meter evidence with that question. Citizens Advice recommends giving the supplier the information supporting a suspected error and retaining the correspondence. If the underlying readings are estimated, obtaining and submitting an actual reading may be the useful next step.[6]

For Economy 7 or another multirate tariff, request one calculation per register and rate period: Ofgem distinguishes these from a single-rate bill. The same habit applies whatever the assistant: preserve the original quantities, explain each transformation, and finish with the specific item that needs clarification.[8]

Report an error or suggest a correction