Based on
Review the feature, the policy, and the process, not one percentage
The VUST runtime gives GrammarBot a narrow job: fix grammar, spelling, and punctuation, preserve meaning, and avoid explanations or new content. Turnitin recommends treating an AI score as one conversation input, while Penn State and Harvard separately emphasize detector limits and the value of drafts and version history.
This is not legal advice or a promise about a specific institution's decision.
Correction, generation, and evidence are different things
| Action | What changes | What to verify |
|---|---|---|
| Spelling and punctuation | The form of your source text | Whether editing is allowed |
| Grammar correction | Incorrect forms and agreement | Whether meaning and voice remain |
| Paragraph rewrite | Wording and sometimes structure | Whether AI assistance must be disclosed |
| Detector score | A probabilistic signal, not authorship | Drafts and supporting evidence |
If the work is flagged
- 1. Check the assignment rule: Some courses allow spelling and grammar assistance; others prohibit any AI-enabled tool. There is no universal institution rule.
- 2. Keep process evidence: Retain the original draft, version history, and narrow correction output. These show how the work developed better than an argument about a detector percentage.
- 3. Separate correction from generation: Fixing a typo or agreement error is not the same act as generating an argument, example, or paragraph. A content rewrite is a different level of assistance.
- 4. Request human review: If the work is flagged, explain the process calmly, show drafts, and ask for the total evidence to be reviewed under the institution's policy.
Reviewed basis: Turnitin, Penn State, and Harvard Bok Center.
- Based on
- The VUST runtime contract and institutional guidance, with no detector-accuracy claim.
- Best for
- Students and ESL writers using narrow correction where it is permitted.
- Not ideal for
- Concealing prohibited generation or manipulating a detector score.
- In short
- Drafts and the assignment rule matter more than one automated flag.