Is it safe to use AI tools like ChatGPT and Claude?

Yes, with common sense. Treat them like a conversation in a public café — don't share passwords, confidential work documents, financial data, or personal information you wouldn't say out loud in front of strangers. The tools themselves are not out to get you. Just apply the same judgment you'd use anywhere online.

Does AI save my conversations?

Most free AI tools do save your conversations and may use them to improve their models. You can turn this off in the privacy settings of most major tools including ChatGPT, Claude, and Gemini. It takes about sixty seconds and is worth doing.

What should I never put into an AI tool?

Passwords, financial account details, confidential client information, sensitive work documents, home addresses, and anything your employer would have strong feelings about. When in doubt, leave it out.

Can my boss see what I type into AI tools?

If you're using a free consumer version on your personal device, your employer generally cannot see your conversations. If your workplace provides AI tools or has enterprise agreements, different rules may apply. When in doubt, check your workplace policy before you paste anything.

Is the free version of AI tools less private than paid?

Generally yes. Free tiers typically have broader data collection than paid or enterprise versions. If you're using AI for sensitive work it's worth understanding which version you're on and what that means for your data.

Which AI tool is the most private?

All major AI tools have privacy settings worth checking. Claude, ChatGPT, and Gemini all let you opt out of having your conversations used for training. No tool is completely private but all of them give you meaningful control if you look for it.

What is AI in plain English?

AI is software that can understand language, generate text, analyse information, and assist with tasks. The most common type for everyday use is a large language model — which is just a fancy way of saying it's been trained on an enormous amount of text and learned to have a conversation from it.

Do you need to know coding to use AI?

No. Most modern AI tools work through conversation. You type what you need, it responds. No technical knowledge required.

Is ChatGPT the same as AI?

No. ChatGPT is one example of a conversational AI tool, made by a company called OpenAI. There are many others including Claude, Gemini, and Copilot. They all do similar things but taste slightly different.

What are the best AI tools for beginners?

ChatGPT, Claude, and Gemini are the best starting points for beginners. They're free to use, work through simple conversation, and you really can't break them. ChatGPT is the most well known, Claude is particularly good for thinking things through, and Gemini integrates well with Google tools you probably already use.

How do you write a good AI prompt?

A good prompt tells the AI who you are, what you need, how you want it done, who it's for, what you don't want, and what good looks like. The more detail you give it, the better the result. Vague in, vague out — every single time.

What is artificial intelligence in plain English?

Artificial intelligence is software that can understand language, generate text, analyse information, and assist with everyday tasks. The most common type for everyday use is a large language model — trained on enormous amounts of text and designed to have a conversation with you in plain English.

What is a prompt in AI?

A prompt is the instruction you give an AI tool. The quality of what comes back depends entirely on the quality of what you put in. A good prompt includes context, method, audience, constraints, and an expected result. Vague in, vague out — every single time.

ChatGPT is a conversational AI tool made by OpenAI. It understands plain English, responds to questions and instructions, and can help with writing, research, problem solving, and everyday tasks. It was the first mainstream AI tool of its kind when it launched in 2022 and remains the most widely used.

What is a large language model?

A large language model is a type of AI trained on enormous amounts of text. It learns patterns in language and uses them to generate responses, answer questions, and have conversations. ChatGPT, Claude, and Gemini are all large language models. The name sounds technical but the experience is just a conversation.

Claude is a conversational AI tool made by Anthropic. It is particularly good at thinking things through, writing, and nuanced conversation. Like ChatGPT and Gemini it works through plain English conversation and requires no technical knowledge to use.

Gemini is Google's conversational AI tool. It integrates well with Google products like Gmail, Docs, and Drive, and is available free to anyone with a Google account. Like ChatGPT and Claude it works through plain English conversation.

AI literacy is the ability to understand, use, and think critically about artificial intelligence tools. It does not require technical knowledge or coding skills. It means knowing what AI can and cannot do, how to get useful results from it, and how to protect your privacy when using it.

What is prompt engineering?

Prompt engineering is the skill of writing clear, detailed instructions for AI tools to get better results. Despite the technical name it requires no coding knowledge — it is simply learning how to place a good order. The better your prompt, the better the output.

Mostly yes, with the same common sense you'd apply anywhere online. AI tools are genuinely useful and the major ones are built by companies with serious privacy and security obligations. That said, they can get things wrong, fill in gaps with confident-sounding nonsense, and occasionally misread what you're asking. The trick isn't blind trust or blanket suspicion. It's knowing when to verify what comes back.

If I download ChatGPT on my phone, can it see everything?

Short answer: no. ChatGPT can only see what you actively type or paste into it. It doesn't have access to your photos, emails, messages, contacts, or anything else on your phone unless you specifically share it in the conversation. You're in control of what goes in. The same common sense applies as anywhere online though. Don't share passwords, financial details, or anything you wouldn't say out loud in a café.

a stack of three old brown tattered books on a brown wooden chair in front of a bookcase of books

The Sauce Dictionary

AI has a language problem.

It's full of words that sound technical, intimidating, and frankly a bit made up. Most of them aren't that complicated once someone explains them properly.

That's what this page is for. One term at a time, in plain English, with a sauce analogy to make it stick. Bookmark it. Come back when someone says something that makes no sense. We'll be here.

So… what do all these AI words actually mean?

In alphabetical order, because even sauce has standards.

Algorithm

A set of instructions that tells a computer what to do and in what order. Every time you get a recommendation on Netflix, Spotify, or YouTube, an algorithm decided what to show you. It's not magic. It's just a very long, very fast recipe.

Like a recipe that runs itself. You set the ingredients and method once, and it just keeps cooking.

Bias

When an AI produces outputs that consistently favour certain groups, perspectives, or outcomes over others. Usually because the data it was trained on reflected those same patterns. AI doesn't invent bias. It inherits it.

If you only ever tasted one brand of tomato sauce growing up, you'd probably think that's just what tomato sauce tastes like. AI has the same problem.

Chatbot

A computer program designed to have a conversation with a human. Not all chatbots use AI, some just follow a script. The ones that use AI, like ChatGPT and Claude, can understand context and respond more naturally.

The difference between a chatbot that follows a script and one that uses AI is the difference between a vending machine and a good waiter.

Context Window

The amount of text an AI can read and remember within a single conversation. Think of it as the AI's short term memory. Once you go beyond the context window, it starts forgetting what was said earlier.

Like a dinner party where the host can only remember the last twenty minutes of conversation. Everything before that? Gone.

Context Window Degradation

What happens when a conversation gets so long that the AI starts losing track of what was said earlier and its responses become less coherent or go completely off track.

If your AI suddenly starts talking about llamas when you were discussing website design, this is probably why (true story - this actually happened to Elyse)!

The sauce has been on the stove too long. It's starting to burn at the edges.

Deepfake

AI generated video, audio, or images that make it look or sound like someone said or did something they didn't. The technology is genuinely impressive. The implications are genuinely concerning.

Imagine someone photoshopping your face onto someone else's body, except now they can do it with video and your voice too. That's a deepfake.

Generative AI

AI that creates new content (text, images, audio, video, code) rather than just analysing or categorising existing content. ChatGPT, Claude, Gemini, and DALL-E are all generative AI tools.

Regular AI reads the recipe and tells you what's in it. Generative AI reads a thousand recipes and writes you a new one.

Guardrails

Rules and limits built into an AI tool to stop it from producing harmful, dangerous, or inappropriate content. Every major AI tool has them. They vary in strictness depending on who built them and why.

Like the sides on a bowling lane for kids. They don't stop you playing, they just stop the ball going somewhere it really shouldn't.

Hallucination

When an AI confidently states something that is completely made up. It's not lying. It genuinely doesn't know it's wrong. It's filling in gaps in its knowledge with plausible sounding nonsense.

Like a chef who's never been to Italy confidently explaining the authentic regional origins of their pasta dish. Sounds right. Might be completely wrong.

Knowledge Cut-off

The date after which an AI tool has no information. It was trained on data up to a certain point, and it doesn't know what happened after that. Ask it about recent events and it'll either tell you it doesn't know or (worse) make something up.

The sauce was bottled in 2024. Anything that happened after that isn't in the bottle.

Large Language Model (LLM)

The type of AI behind tools like ChatGPT, Claude, and Gemini (your tomato sauces). Trained on enormous (large) amounts of text (language), it learned to understand and generate language by finding patterns across billions of words. The name is technical. The experience is just a conversation.

Imagine reading every book, article, and website ever written and learning language from all of it. That's roughly what an LLM has done.

Machine Learning

A way of training AI by showing it enormous amounts of data and letting it find the patterns itself, rather than programming in specific rules. Most modern AI uses machine learning.

Instead of teaching a child the rules of grammar, you just read them ten thousand books and let them figure it out. That's machine learning.

Model

The underlying AI system that's been trained to do a specific job. When companies release a "new model" they've trained a new version that's smarter, faster, or better in some way. ChatGPT runs on GPT models. Claude runs on Claude models.

Think of it like a new vintage. Same winery, new year, slightly different result. Sometimes better. Sometimes not.

Multimodal AI

An AI that can work with more than one type of input (text, images, audio, video) rather than just words. Most of the major AI tools are now multimodal to some degree.

A sauce that works on everything. Pizza, pasta, sandwiches, eggs. Doesn't matter what you put it on, it handles it.

Natural Language Processing (NLP)

The branch of AI that deals with understanding and generating human language. Every time an AI reads your message and responds in plain English, NLP is doing the work.

It's what stops the AI from needing you to speak in code. You talk normally. It figures out what you mean.

Neural Network

A computing system loosely inspired by the structure of the human brain. It processes information through layers of connected nodes, finding patterns in data the way neurons find patterns in experience.

Not actually a brain. More like a very complicated flowchart that learned to think for itself.

Prompt

The instruction or question you give an AI tool. The quality of the response depends almost entirely on the quality of the prompt. Vague in, vague out, every single time. Check out The Perfect Prompt over on the Getting Saucy page for more info!

Your prompt is your order. The more specific the order, the better the meal.

Temperature

A setting that controls how creative or unpredictable an AI's responses are. Low temperature means more consistent, predictable answers. High temperature means more creative, surprising, occasionally unhinged ones.

Turn up the heat, get spicier results. Turn it down, get something safer and more reliable. Just like actual cooking.

Token

The unit of text an AI processes, roughly three to four characters or about three quarters of a word. AI tools think in tokens rather than words, and most have limits on how many tokens they can process at once.

This is why, when everyone asked ChatGPT how many Rs were in the word strawberry, it confidently said two. It wasn't reading the word the way you do. It was processing it as fragments (straw and berry) and counting from there. Three Rs, right in front of it, and it missed them completely.

The AI isn't reading your message the way you read it. It's breaking it into tiny pieces and processing each one. Tokens are the pieces. And sometimes the pieces don't add up the way you'd expect.

Training Data

The information an AI was trained on: the text, images, or other data it learned from. The quality and diversity of training data has an enormous impact on how well the AI performs and what biases it might carry.

What goes into the sauce determines what comes out. Rubbish ingredients, rubbish result.

Use Case

A use case is simply a situation in your actual life where AI could genuinely help you get something done, and despite what the internet would have you believe, it doesn't need to be complicated or impressive to count.

Writing an email you've been putting off, summarising a document that's been gathering dust in your inbox since Tuesday, figuring out what to cook with whatever's left in the fridge before the weekly shop, or drafting a message you're not quite sure how to word. Those are all use cases, and they're a lot more useful than whatever the tech bros are demonstrating on LinkedIn this week.

Most people struggle to find theirs at first because they've been doing things on autopilot for so long that they don't even notice the friction anymore. The trick is to pay attention to the moments in your day where you think "ugh, I really don't want to do that" or "I have no idea where to even start with this" because that's your use case quietly waving at you from across the kitchen.

Turns out most of us have about seventeen of them before breakfast. We just haven't been looking.