7 AI Terms Explained, From Tokens to Hallucinations

Tom’s Guide has published a plain-language glossary of seven AI terms that come up whenever people use or read about chatbots such as ChatGPT, Gemini and Claude. The list covers large language models, generative AI, machine learning, neural networks, hallucinations, prompt engineering and tokens.

The last of those matters for anyone paying for an AI service, because some subscription plans set limits measured in tokens.

The seven terms, in plain English

Large language model (LLM). This is the program behind a chatbot. It is trained on huge amounts of text, including books, websites, articles and research papers, and it learns patterns in human language instead of memorizing answers. Ask it to write a birthday message for a friend and it draws on those patterns to produce one.

Generative AI. This describes AI that creates new content from a prompt: text, images, audio, video and computer code. The stylized photos, skateboarding-dog videos and AI-made songs people see online all come from generative AI.

Machine learning. Computers learn to spot patterns and make predictions from data, rather than following step-by-step instructions from a programmer. Tom’s Guide compares it to Netflix recommendations, which draw on what you watched, liked and skipped. More data produces stronger suggestions.

Neural network. This is a computer system built from layers of connected nodes, loosely modeled on how neurons in the brain link up. It is how an AI learns to recognize something complicated, such as a cat in a photo.

Hallucinations, prompts and tokens

A hallucination is a confident answer that is wrong or invented. Tom’s Guide says an AI tool does not automatically know whether what it says is true, so it can produce made-up statistics and cite sources that do not exist. The author recommends checking its answers more than once, especially for medical, financial or legal questions.

Prompt engineering means writing detailed instructions so a chatbot gives a better result. A prompt is whatever you type into the tool. The article contrasts a bare request like “give me dinner ideas” with a fuller one that specifies the number of ideas, the preparation time, the cost per meal and the supermarket where the ingredients will be bought.

Tokens are the small chunks of text an AI model works with when it reads and writes. A token can be a short word, part of a longer word or a punctuation mark. A chatbot splits the sentence you type into tokens before it builds a reply.

Why it matters: tokens show up on your plan

Some AI models can only handle a limited amount of information in a single conversation. That is why token limits appear in subscription plans: they indicate how much a plan can take in before it reaches its cap. Knowing the term makes those plan descriptions easier to compare.

The hallucination warning has a practical side too. Because chatbots can state false information with confidence, the answers they give on health, money or legal matters are the ones Tom’s Guide says deserve the most checking.

Prompt wording also changes what comes back. A request with specifics gives the model more to work with than a vague one, according to the article’s dinner example.

Who wrote the list

The glossary comes from Elton Jones, who covers AI for Tom’s Guide. He says he could not explain the difference between a chatbot and an AI agent a year ago. He wrote that beginners will not become experts in ChatGPT, Gemini or Claude overnight, and that the list is meant as a base of core definitions for any level of user.

Source: Tom's Guide