AI Review
AI review reads your entire manuscript and evaluates every checklist item automatically, returning an item-by-item compliance report, a pass rate, identified issues, and an overall recommendation.
How AI review runs
When you select AI review, the module works through these steps:
- Loading checklist from database -- the checklist for your study type is fetched from the data service. If no checklist exists, the review stops with an error.
- Preparing manuscript for analysis -- your PDF is converted to base64.
- Analyzing manuscript against checklist criteria -- the manuscript and checklist are sent to the platform's server-side analysis function, which runs the AI evaluation.
- Processing results -- the returned evaluation is assembled into the review result.
The analysis sends three things to the server: the manuscript (as base64), the study type, and the checklist questions for that type.
What the AI returns per item
For each checklist item, the AI returns:
| Field | Meaning |
|---|---|
| question | The checklist item evaluated |
| answer | Whether the manuscript adequately addresses it (true/false) |
| justification | The AI's explanation for its judgement |
| confidence | An optional confidence value (%) for that item |
The justification makes the review auditable: for every item, you can read why the AI reached its conclusion.
The overall result
Across all items, the AI review produces:
| Output | Contents |
|---|---|
| Pass rate | Passed items / total items as percentage |
| Item-by-item list | Every checklist item with answer, justification, and confidence |
| Major issues | Significant reporting gaps identified |
| Minor issues | Smaller reporting gaps identified |
| Overall assessment | Narrative summary of compliance |
| Recommendation | Accept / Minor Revisions / Major Revisions / Reject |
The pass rate is the proportion of checklist items the manuscript satisfied. The recommendation defaults to Major Revisions if the analysis does not return an explicit one.
Exporting the report
The AI review result can be exported as a PDF report containing the recommendation, overall assessment, major and minor issues, and the full item-by-item list with justifications. This gives you a shareable compliance report for submission materials or revision guidance.
The Review Chat
After a review (AI or manual), a Review Chat is available to discuss findings. It is study-type-aware: it builds context describing the study type and specific issues found, so guidance is tailored to your manuscript.
The chat assembles context such as "This is a review of a [study type] with the following issues..." and adds study-type-specific guidance hints. You can ask:
- Why a particular item matters
- How to address a specific gap
- What the relevant guideline expects
The assistant helps you understand review feedback and how to address it -- constructive guidance that maintains academic standards.
AI review vs manual review
| Manual review | AI review | |
|---|---|---|
| Item judgement | You answer yes/no per item | AI answers per item with justification |
| Compliance score | Implicit (issues by severity) | Explicit pass rate |
| Recommendation source | Computed from escalation rules | Returned by AI analysis |
| Issue lists | Major/minor from item categories | Major/minor identified by analysis |
| Export | On-screen results | On-screen + PDF export |
The AI review's per-item judgements should be checked against the manuscript using the justifications provided. The issue lists and recommendation focus your attention but do not substitute for final human judgement.
The complete picture
Upload PDF
|
Determine study type (survey / manual / AI)
|
Fetch matching checklist
|
Choose review mode:
- Manual review: you judge each item
- AI review: the AI judges each item
|
Compliance report (pass rate / issues / recommendation)
|
Review Chat (discuss and plan revisions)
|
Revise and optionally re-review
The module matches your manuscript to the right reporting guideline for its design and checks it item by item -- so reporting gaps are caught before submission, not after.