Is ChatGPT really THAT bad at math – or could I be clearer in my prompts?

Understanding ChatGPT’s Limitations in Math and Data Accuracy: A User’s Perspective

As a user subscribed to the £20 plan of ChatGPT, I initially turned to this tool to generate mock-up images, where it proved to be highly effective. My primary daily applications, however, are more straightforward: maintaining two trackers—one for credit card spending (tracking running totals and transaction splits) and another for calorie intake (monitoring daily targets and calculating daily and weekly over/under totals).

Despite its impressive versatility, I have frequently encountered issues where ChatGPT provides incorrect summaries or calculations—even after it initially outputs accurate data. For example, today I entered my Sunday calorie intake, and it generated a table indicating I was approximately 1,500 calories over my goal; however, its summary later claimed I was about 2,500 calories over. Curious about this discrepancy, I asked it to explain how it arrived at the 2,500-calorie figure. The assistant shared its calculations but acknowledged it was considering my tracker’s 1,500 calorie overage as the correct figure. When I requested further clarification on its logic, it then recognized that the 1,500 figure was accurate all along, revealing an inconsistency in its processing.

This isn’t an isolated occurrence. I have experienced similar inconsistencies with my financial tracker, where ChatGPT randomly assigns values to categories or “buckets” that I’ve never mentioned. It tends to defend these arbitrary assignments until I ask for a detailed, step-by-step explanation, at which point it often reorganizes its reasoning.

My initial consideration was to leverage ChatGPT for advice—particularly on financial matters. However, given its difficulties in precisely tracking calories and spending, I am left questioning how others derive significant value from this tool. The frequency of errors and contradictions makes me cautious about relying on it for critical data analysis or decision-making.

In summary, while ChatGPT delivers impressive capabilities in generating content and providing general guidance, its limitations in maintaining data accuracy and consistent reasoning—especially concerning numerical or logical tasks—are evident. Users should approach it with awareness of these potential pitfalls and verify outputs when precision is essential.