Correction or generation
The Boundary Isn't the Word AI. It's What the Tool Does.
A checker can use AI internally and still perform correction only. A generative tool creates ideas, structure or wording that was not present in the input.
The useful rule
Classify the action, not the brand
One product may fix a typo, rewrite a sentence or create a paragraph. Academic policies should name the allowed depth of intervention and any disclosure requirement.
Decision table
| Tool | Action | New content | Typical risk | VUST |
|---|---|---|---|---|
| Spelling correction | Minimal | No | Low, but policy wins | GrammarBot |
| Rewriting | Medium | Usually no new facts | Gray zone | Humanizer / Rephrase |
| Generation | High | Yes | Often needs permission/disclosure | Post / AI chat |
This is a classification framework, not a universal academic-integrity rule.
Three levels of intervention
One AI label can hide very different actions.
Frequently asked questions
Is spell check AI?
Sometimes, technically: a modern checker may use rules, statistics or an LLM. For policy decisions, the important boundary is the task — correcting supplied text or creating new content.
Does GrammarBot generate an essay?
No. Its product job is to correct grammar, spelling and punctuation in text you supplied, show the changes and briefly explain them.
Is paraphrasing generative AI?
It is a gray zone because the intervention is stronger than fixing a typo. Check the course or workplace policy and disclose tool use where required.
Can I use a grammar checker at university?
There is no universal answer. Policies vary by institution, course and assignment; this framework classifies the action but does not override local rules.
How should an educator write the rule?
Allow or prohibit named actions: spelling-only correction, sentence rewriting, idea generation, drafting and source creation. That is clearer than a blanket AI label.