The estimate and final charge differ

POM AI labels estimates as approximate because the estimate is calculated before the task runs, while the final charge reflects the execution.

Common reasons

• The prompt, requested word count, item selection, image settings, or audio duration changed after estimation.
• A multi-step writing workflow generated a structure and then separate sections.
• A batch processed a different number of items than first selected.
• The generated output used more or less processing than predicted.
• Part of a batch completed before a later item failed.

Compare the same operation inputs first. The estimate’s credits_needed is not a reservation; the execution’s credits_needed and remaining_credits are the authoritative returned values.

Review a discrepancy

1. Note the feature, time, and site.
2. Record the displayed estimate.
3. Record the final charge and balance without exposing the license.
4. Count completed items or sections.
5. Check for retries, duplicate submissions, or a lost browser response.
6. Review the account usage history.

Do not retry merely to reproduce the problem; another execution can consume more credits. If the difference remains unexplained, share the redacted values and public error/result status with support.

Related guides: credit estimate contract, credit consumption contract, and review unexpected credit use.