Why Google's own AI struggles to spell 'Google' correctly

Google's artificial intelligence systems have been caught making embarrassing spelling errors, including misspelling the company's own name. The issue highlights ongoing limitations in large language models when it comes to basic character-level tasks. Experts say this reveals a fundamental flaw in how AI language systems process text.

Google's AI tools have once again come under scrutiny for an embarrassing and surprisingly basic flaw: an inability to reliably spell words correctly — including, at times, the name "Google" itself. The issue has resurfaced in tech circles, prompting renewed debate about the real-world limitations of even the most advanced AI language models available today.

How AI reads — and misreads — text

The root of the problem lies in how large language models (LLMs) process text. Rather than reading individual characters the way humans do, these systems work with "tokens" — chunks of text that may represent whole words, parts of words, or fragments of phrases. This means that when asked to count letters or spell out a word character by character, AI models are essentially being asked to do something they were never architecturally designed for.

This architectural blind spot means that tasks which a child in primary school could complete without effort — such as spelling a word letter by letter or counting how many times a specific letter appears — can trip up the most powerful AI systems on the market. Google's Gemini, OpenAI's ChatGPT, and similar tools have all demonstrated this weakness under the right conditions.

Not a bug — a design limitation

Researchers and AI engineers have noted that this is not a bug to be patched in the next software update, but rather a fundamental characteristic of transformer-based language models. These systems are optimised to predict the most statistically likely next token, making them excellent at generating coherent prose or summarising complex documents — but poorly suited for precise character-level reasoning.

The irony of Google's AI struggling to spell its creator's own name has not been lost on critics, who argue it serves as a useful reminder that AI tools, despite their impressive capabilities, remain unreliable for tasks requiring strict logical or sequential precision. For everyday users, the lesson is clear: trust AI for broad strokes, but verify the fine print yourself.