In short
AI use isn't yes/no, it's a scale with six levels: from level 0 (no AI) through spelling help, rewriting, research assistance and AI-assisted structuring up to level 5 (full text generation by AI). The scale is deliberately discipline-agnostic — it works for a school essay just as much as for a marketing text.
Why a scale instead of a line
"Is AI use allowed?" rarely has a useful yes/no answer. AI-assisted spellchecking is fundamentally different from a fully AI-generated paper — yet both often get lumped together. A level scale makes the actual question discussable: not whether, but how much and for what.
THE SCALE IN DETAIL
The 6 levels of use
No AI
The text is produced entirely without AI tools — the reference level every other level is compared against.
Spelling & grammar
AI corrects typos and grammar without changing content or phrasing — comparable to a classic spellchecker.
Rewriting & style
AI helps make already-existing sentences clearer or more polished — the content stays entirely the human's own.
Research assistance
AI helps gather, summarize or contextualize information — the actual drafting and phrasing stays with the human.
(Partial) content and structure by AI
AI suggests an outline or structure and produces a draft of the article. The human takes over parts of it entirely in places, but rewrites most of it. AI drafts, the human rewrites.
Full text generation
AI produces the complete text, the human adopts it largely unchanged. The level most discussions about "AI use" are actually about — without naming it that way.
What this scale is really for: it makes AI use transparent instead of just bannable. An examination policy, an editorial guideline, or a usage declaration can set concretely up to which level something is allowed - instead of a blanket, barely enforceable "no AI" rule that nobody can check consistently.
Free PDF summary
The 6 levels of use as a print-ready A4 one-pager - to print, post, or use as the basis for your own policy.
Frequently asked questions
That depends on context — an examination policy or editorial guideline sets which levels are allowed. The scale is what makes that decision discussable in the first place; it doesn't make the decision itself.
No, deliberately not. The same level 3 (research assistance) describes both a student using ChatGPT for outline ideas and a content team using AI for research summaries.
Set, per context, up to which level use is allowed, and require disclosure above a certain level. The grading rubric and usage declaration templates build directly on this scale.
By Simon Schmidlin