Google’s AI may be smart enough to write code, but spelling?

Imagine asking Google how many “p” letters are in the word “Google” and getting the wrong answer from Google itself. Sounds like a meme, right? Well, the internet is currently having fun with exactly that situation. Google’s AI Overviews feature is once again going viral, but not for the reasons the tech giant would probably like. Many users across social media are sharing screenshots showing Google’s AI confidently failing basic spelling and letter-counting questions. Also Read: Google Health replaces Fitbit after 14 years with AI-powered wellness features

And yes, some of the examples are genuinely funny. Also Read: 7 AI Image Generators Everyone Is Talking About

In one response, the AI reportedly claimed there is a 1 “f” letter in “Google.” In another, it said the word “Gemini” contains two “m” letters. Even in “Gmail,” Google AI claims to have 1 “o” letter. The list goes on. 

Google can’t spell right; but why? 

The strange part is that modern AI models can already write code, summarise articles, generate images, and solve fairly advanced problems. So how does spelling suddenly become difficult? 

The answer lies in how large language models (LLMs) actually work. AI systems do not “read” words like humans do. Instead, they process chunks of text called tokens. These tokens may represent full words, syllables, or smaller text fragments rather than individual letters.

That means the AI often understands the meaning of a word without properly understanding the exact arrangement of letters inside it. So while humans naturally count letters one by one, AI models sometimes struggle because they are not technically designed around spelling logic in the first place.

Netizens reacted on social media

Unsurprisingly, users on Reddit and X have turned this into another AI meme moment. Many people are joking that Google’s AI can probably build an app faster than most humans but still might fail a kindergarten spelling test. Others are pointing out the bigger issue underneath the humour – trust. 

Search engines are where millions of people go for quick answers every day. If AI-generated results confidently get basic spelling wrong, people naturally start wondering what else might also be inaccurate.

And this is not the first time AI Overviews has faced criticism.

Google’s AI Overviews faced issues since the beginning 

Soon after Google aggressively pushed AI summaries into Search, users discovered several bizarre responses. Some examples included AI-generated advice suggesting people eat rocks or add glue to pizza recipes. There were also cases where satirical content was treated as factual information.

Now the spelling issue is once again fuelling criticism around AI Overviews, especially among users who already feel that Search has become more cluttered with AI-generated summaries.

Add Techlusive as a Preferred Source

Google has acknowledged the issue. The company reportedly said that “counting within words has been a known challenge for LLMs” and that it is working on fixing it.