humanflow

Generative AI

Also called GenAI

Generative AI is the class of systems that produce new content — text, images, audio, video or code — as opposed to systems that classify, rank or predict from content that already exists.

Last reviewed 15 August 2026 · The HumanFlow team

In plain English

The dividing line is whether the output is new material or a judgement about existing material.

A tool that writes you a paragraph is generative. A tool that tells you the paragraph is 87% likely to be AI-written is not — it is a classifier, and it is the other half of this website.

A worked example

Four tools a student might use in one afternoon, sorted by which side of the line they fall on.

Generative                     Not generative
----------                     --------------
ChatGPT writing a paragraph    Turnitin scoring similarity
An AI humanizer rewriting it   An AI detector scoring it
Autocomplete finishing a line  A spellchecker flagging a typo
A translation tool             A search engine ranking results

The right-hand column contains tools nobody thinks of as AI, several of which are, and none of which produce new content.

The middle row is where policies get difficult. Autocomplete generates text, technically, and almost no institution intends to ban it.

That is why well-written policies name behaviours rather than technologies. A rule about "generative AI" has to explain whether predictive text on a phone is included, and most do not.

Why it matters for AI detection

Institutional policies are written in this vocabulary, and the term is doing real work in them. A policy that prohibits "generative AI" and a policy that prohibits "AI tools" cover very different ground, and the second one arguably bans your spellchecker.

The distinction also decides what you have to disclose. Most disclosure requirements are aimed at content that was generated, not at software that checked something you wrote.

If your policy is ambiguous on this point, the ambiguity is the institution's problem to resolve and a reasonable thing to ask about before submission rather than after.

Commonly confused with

Large language model
The LLM is the underlying technology for text. Generative AI is the broader category including image and audio systems built on quite different architectures.
AI detector
A detector is the opposite kind of system: it consumes text and outputs a judgement. It generates nothing, and confusing the two makes policy documents incoherent.

Read next

Part of the AI detection glossary.