Glossary
The key AI terms, briefly explained in plain language.
Technical terms such as API and tokens appear in the Advanced view.
- Affiliate link
- A link through which a website may earn a fee if you buy something. On our site it never affects the order of tools and we always disclose it.
- AI (artificial intelligence)
- Computer programs that do tasks that normally need human thinking, such as writing text, recognising images or answering questions. They learn this from very many examples.
- AI Act
- The European law on AI. It sets rules based on risk, such as banned uses and a disclosure duty for deepfakes.
- AI agent
- AI that does not just answer but carries out steps to reach a goal, such as looking something up, editing a file or operating an app. Never let an agent do important things unchecked.
- AI assistant (chatbot)
- A program you chat with in plain language: you ask a question or give an instruction and get an answer.
- AI avatar
- A digital presenter that speaks your text in a video, without you filming yourself.
- API
- A connection that lets software talk to a service. Through an API you can use AI in your own app or automation, usually paid per use.
- API key
- A secret code that gives access to an API. Never share it or put it in public code or the browser.
- Automation (workflow)
- A fixed series of steps that runs by itself, for example: new form β data into a spreadsheet β confirmation e-mail.
- Commercial use
- Using something to make money, for example in ads or for clients. Not every plan allows it; read the terms.
- Context window
- How much text a language model can take into account at once: your question, the conversation and any documents. When full, the model βforgetsβ the beginning.
- Credits
- An allowance you use up per AI task, such as per image or per minute of video. Check how many credits you get per month and what extra credits cost.
- Data processing agreement (DPA)
- An agreement between your business and a service on how it handles personal data. Required under the GDPR when a service processes personal data for you.
- Deepfake
- AI-made images, video or audio that look real but are fake. The EU AI Act requires disclosing that such content is AI-generated.
- Dubbing
- Voicing a video in another language, sometimes with matching lip movements.
- Embeddings
- A way to turn text into numbers that capture meaning, so software can search by meaning instead of exact words.
- EU data residency
- Your data is stored on servers inside the European Union. Some organisations require this.
- Fine-tuning
- Training an existing model further on your own examples so it fits a specific task or writing style better.
- Freemium
- A service with a limited free version and paid plans with more features.
- GDPR
- The European privacy law. It sets what companies may do with personal data. Do not enter other peopleβs personal data without good reason.
- Generative AI
- AI that creates something new: text, images, music, video or code, based on your description.
- Hallucination
- When AI confidently states something that is not true, such as an invented source, date or number. Always check important facts.
- Language model (LLM)
- The βbrainβ behind an AI chatbot. It has read enormous amounts of text and predicts a fitting answer word by word. LLM stands for large language model.
- Multimodal
- AI that understands more than one kind of input, for example text and images, sound or video.
- No-code
- Building something, such as a website, app or automation, without programming: by dragging, clicking and plain language.
- Open model (open weights)
- An AI model you may download and run yourself. Check the licence: not every open model may be used commercially.
- Per-user price
- You pay per person using the service. A team of five pays five times the price.
- Prompt
- The question or instruction you give AI. The clearer the prompt, with goal, context and desired format, the better the result.
- RAG (chatting with your own documents)
- A technique where AI first retrieves relevant passages from your documents and bases its answer on them, making answers more specific and checkable.
- Running locally
- Using AI on your own computer instead of over the internet. Your data then stays on your device.
- Text to speech
- Having a written text read aloud by a computer voice.
- Token
- A piece of text (often part of a word) a language model works in. API prices and limits are usually expressed per token.
- Training on your data
- Some services use what you enter to further train their AI. You can often switch it off; check the settings or terms.
- Transcription
- Automatically turning spoken words into text, for example from an interview or meeting.
- Upscaling
- Making an image larger and sharper with AI without it becoming blocky.
- Verified, Supported, Unverified
- Our statuses for prices and facts. Verified: confirmed word for word on the official source. Supported: several sources agree. Unverified: not yet confirmed on the official site.
- Voice cloning
- Creating an AI voice that sounds like a real person. Only do this with your own voice or explicit consent.
- Watermark
- A logo or text a service puts on your image or video, often on free plans.