Farid A.
I am extremely frustrated with the way…
I am submitting a formal complaint regarding the consumption of my paid Tokens as a result of AI behavior that was outside the scope of the task I requested. I want to make one principle absolutely clear: A user should not be required to pay for errors, hallucinations, unnecessary processing, repeated analysis, or inefficient behavior generated by the AI system itself. I provided Bolt with a precise, limited, and clearly restricted instruction. I explicitly told the Agent not to audit the project, not to redesign anything, not to analyze unrelated files, and not to perform unnecessary work. The requested modification was extremely small and targeted. Despite this, approximately 212K Tokens out of my 300K daily allowance were consumed in a very short period of time. This is completely disproportionate to the actual task performed. The fact that an AI system may require Context, reasoning, file analysis, retries, or internal processing does not automatically make it reasonable for the user to bear the full cost of that processing — especially when the excessive consumption results from AI errors, hallucinations, repeated operations, unnecessary analysis, or failure to follow explicit instructions. I paid for Tokens to accomplish development work — not to finance uncontrolled AI processing or compensate for mistakes made by the system. I therefore require Bolt to address the following: 1. Explain exactly where the approximately 212K Tokens were consumed. 2. Provide, if available, a detailed usage breakdown for this specific session. 3. Confirm what files, Context, operations, and Agent actions generated this consumption. 4. Confirm whether the Agent repeatedly processed the same files or Context. 5. Confirm whether unrelated parts of my project were analyzed. 6. Explain why the system continued consuming such a large amount of Tokens despite my explicit instruction to keep the intervention minimal. 7. Clarify what safeguards Bolt has to prevent users from losing large amounts of paid Tokens because of AI errors or inefficient Agent behavior. 8. Review the session logs from a technical perspective rather than sending me a generic explanation about AI Context. 9. Restore or refund any Tokens that were consumed unnecessarily, abnormally, or as a result of AI/system behavior rather than legitimate work required to complete my request. This is not simply a complaint about AI performance. The important issue is the financial consequence of AI failure. If the Agent makes mistakes, hallucinates, enters unnecessary reasoning loops, repeatedly analyzes information, or performs work that the user explicitly instructed it not to perform, the resulting Token consumption should not automatically become the user's financial responsibility. Otherwise, users are effectively paying for the system's own errors. That is not a reasonable model of accountability for a paid service. I am therefore requesting a formal investigation of this specific session, including the actual usage data and Agent activity. Please escalate this matter to the appropriate technical and billing team, and provide me with a Ticket/Case Number. I also request confirmation that my usage logs will be reviewed and that Bolt will determine whether the approximately 212K Tokens were reasonably and necessarily consumed for the task I requested. If the investigation confirms excessive, unnecessary, abnormal, or system-generated consumption, I expect the corresponding Tokens/Credits to be fully restored or appropriately refunded. Please do not send me a generic explanation about AI, Context windows, or how Agents work. I am asking you to investigate the actual session, identify where the Tokens went, determine whether that consumption was justified, and correct any unjustified financial loss to the user. A paid AI service must have accountability for the financial consequences of its own AI behavior. Users should pay for the work they request — not for the AI's mistakes.
Response from Bolt.newOct 5, 2026
Seeing roughly 212K tokens go on a small, targeted change after you told the Agent to keep it minimal is frustrating, and I'm sorry that happened. I don't have account info attached to this review, so I can't track down your session on my end just yet. We've reached out through Trustpilot to get the info we'd need to find you. Once we have that, someone on our team will look into your actual session with you rather than send a general explanation. - Erica, Bolt CX