Audit Subscription Renewals with ChatGPT
Start with evidence, not a grand total
A ChatGPT subscription renewal audit can turn billing notes into a reviewable table. The useful distinction is between a monthly planning amount, a stated upcoming charge, and a charge whose status still needs evidence. A cancellation request alone does not prove cancellation.
OpenAI’s October 7, 2026 release describes comparisons and calculators through Intelligent UI in ChatGPT. This audit uses an ordinary text table, so the same checks work when an interactive format is unavailable. Collect the account portal status, invoice or renewal notice, and cancellation confirmation when available. Give each record a source ID and record when it was updated.
The example below is editorial fiction with independently checked arithmetic. It is not a real account audit or a reported ChatGPT run.
Preserve the original records
These five fictional records are the complete inputs. Keep their dates, amounts, currencies, and status wording intact so every conclusion can be traced back to evidence.
S1 = 'Service A (fictional): active, auto-renew on; USD 12.00 per calendar month; next stated charge 2026-10-20, USD 12.00. Price includes all stated taxes and fees.'
S2 = 'Service B (fictional): active, auto-renew on; USD 96.00 per year; next stated charge 2026-11-05, USD 96.00. Price includes all stated taxes and fees.'
S3 = 'Service C (fictional): active, auto-renew on; EUR 4.50 per calendar month; next stated charge 2026-11-20, EUR 4.50. Price includes all stated taxes and fees.'
S4 = 'Service D (fictional): a 2026-09-30 record says active, USD 9.00 per calendar month, next charge 2026-11-01. Cancellation requested 2026-10-10; no confirmation or final charge outcome supplied.'
S5 = 'Service E (fictional): free trial ends 2026-10-25; listed post-trial price USD 6.00 per calendar month. Auto-renew status and first charge date are not supplied.'Confirm the date window and extracted inputs
Use the user-confirmed Asia/Seoul date basis: 2026-10-13 inclusive through 2026-11-12 exclusive. That covers 30 calendar dates, October 13 through November 11. It does not describe exact payment timestamps. Preserve date-only records as supplied; ask about timezone or locale ambiguity before interpreting an affected date.
Ask ChatGPT to extract the records first and wait for your confirmation before producing final calculations. Clarify only unknowns that affect a row or total. Missing optional information need not block confirmed rows. Merge repeated evidence for the same subscription and account; ask if apparently similar records might represent separate subscriptions.
Keep a calendar month distinct from a four-week billing cycle. For another interval or variable usage, confirm the planning method rather than forcing it into the monthly or annual rule used here.
Separate planning from upcoming cash charges
After the extracted inputs are confirmed, this is the expected audit. Pending means unresolved; it never means zero. Only the supplied next charge is evaluated for each confirmed active item.
Confirmed-active-items-only monthly planning is USD 20.00 (12.00 + 96.00 ÷ 12), with EUR 4.50 separately. This excludes pending S4 and S5 and is not a complete account budget. Service B’s USD 8.00 is a planning allocation; its annual cash charge remains USD 96.00.
The partial known subtotal of stated charges inside the window is USD 108.00 (12.00 + 96.00). S3 has no stated charge inside this window. S4 and S5 remain pending, so USD 108.00 is not a complete expected-payment total. Do not add USD and EUR or convert them without a user-provided conversion basis.
On a small screen, swipe the table to see all columns.
| Evidence | Renewal status | Monthly planning | Next stated charge | Window assessment |
|---|---|---|---|---|
| S1 · Service A | Active; auto-renew on | USD 12.00 | 2026-10-20 · USD 12.00 | Included · USD 12.00 |
| S2 · Service B | Active; auto-renew on | USD 8.00 = 96.00 ÷ 12 | 2026-11-05 · USD 96.00 | Included · USD 96.00 |
| S3 · Service C | Active; auto-renew on | EUR 4.50 | 2026-11-20 · EUR 4.50 | Outside the window |
| S4 · Service D | Cancellation requested; outcome pending | Pending; historical USD 9.00/month | Historical record: 2026-11-01; amount outcome unconfirmed | Pending; request is not cancellation proof |
| S5 · Service E | Trial; auto-renew unknown | Pending; listed post-trial USD 6.00/month | Unknown; trial ends 2026-10-25 | Pending; trial end is not a confirmed charge |
Resolve uncertain renewals without guessing
For S4, find the cancellation confirmation or current portal status and check the final billing outcome. The September 30 record and October 10 request form a chronology, not proof of a future November 1 charge or successful cancellation. For S5, confirm auto-renew status and the first charge date; a trial end proves neither automatic billing nor automatic cancellation.
If taxes, fees, or a variable price are unresolved, keep the affected total pending. Do not invent later recurring charges beyond supplied dates. Do not claim savings from a request, an uncertain charge, or a negative bill or refund. Even a confirmed cancellation should retain any separately stated final charge.
Check the result before acting
Copy only the service label, amount, currency, billing interval, status and relevant dates. Omit card or account numbers, passwords, tokens and unnecessary personal identifiers. Use aliases for service or account labels where helpful. Available Data Controls depend on your account and workspace; turning off model training does not delete saved chats.
Match each row to its source ID, status wording, and update chronology. Reconcile conflicting evidence against the invoice and account portal. Independently check the annual division, window boundaries, and each currency subtotal. Finalize confirmed rows and leave unresolved rows visibly pending.
Use the companion prompt to repeat the process with your own records. This workflow does not cancel, subscribe, pay, connect accounts, schedule reminders, or send email. ChatGPT can make mistakes, so verify dates and calculations before making a billing decision.
Open the subscription renewal audit prompt
Related guides
Official OpenAI references checked on October 13, 2026. The five service records are editorial fiction, and the example amounts and calendar dates were independently checked. No real account audit, ChatGPT run or measured savings are represented.
OpenAI: ChatGPT release notes · OpenAI: Does ChatGPT tell the truth? · OpenAI: Data controls in ChatGPT
