Why AI Can’t Do Math

This data was originally featured in the June 25th, 2025 newsletter featured here: INBOX INSIGHTS, June 25, 2025: When AI-First Goes Wrong Part 5, Why AI Can’t Do Math

In this week’s Data Diaries, let’s explain some math – in particular, why generative AI tools like ChatGPT struggle with math.

At its heart, math – like addition, subtraction, multiplication, and division – isn’t like written prose. Calculation and statistics may both use numbers, but they do so very differently.

Natural language is predictive and probabilistic in nature. If I say, “I pledge allegiance to the”, most Americans would auto-complete that sentence mentally with the word “flag”. The probability of the next word being flag is nearly 100%.

Mathematical language is symbolic and deterministic in nature. If I say, “4 + 5 =”, we know that the numbers are symbols (symbolizing a quantity) and the + sign is an operator that mentally groups two things together. An AI tool might look at 4 and 5 and conclude the next most probable character is 6, which would be probability-wise correct but computationally wrong.

Imagine this. Suppose you ask AI for a brownie recipe. It’s seen millions of brownie recipes, and almost all of them are in standard size pans. It gives you a solid recipe. Then you tell it that you don’t have a 9×13 pan. In fact, the only thing you have is a 13 inch frying pan. If you ask it to recompute its ingredients based on that, it’s going to go off the rails because there are almost certainly no recipes based on a non-standard size frying pan.

A human would go, okay, 9×13 is 117 square inches. A 13 inch frying pan has a radius of 6.5 inches, so its area is 132.73 inches, so we need to adjust the recipe upwards by 13% for each ingredient.

AI will try to make adjustments based on what it’s seen, but since it hasn’t seen that much, chances are its math is going to yield inedible brownies.

What this means is that under the hood, AI is incapable of performing the actual math, the actual calculation. It can’t do it, and at least for now, with the current AI architectures, under the hood it will NEVER do math well. What software makers do behind the scenes is have AI write CODE (which is a language) that it executes and then does the math, returning the result to the language model. It’s an inelegant but effective hack.

This is why we always advise to hand finished results to tools like ChatGPT. Share the report, not the raw data. The AI tools will write good prose around the finished results because you’re then using numbers as language, like saying “Website traffic was down 41% last month”, rather than as symbols of computation.

Don’t let AI do math.

Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.