Samples & Verification
This page shows theverification stepof the platform's output, abridged (example topic names are generalised; bibliographic entries are publicly available records). The verification report is shown as it is, includingentries that could not be confirmed and entries that were inconsistent— the point of verification is to let researchers see the problems, not to make the output look flawless.
Sample 1 ｜ Chinese review (immune-related adverse events in cancer immunotherapy)
Citation verification report (abridged)
27 citations checked: 27 passed, 0 failed or pending human review.
| Citation | Status and basis |
|---|---|
[#1] | ✅ In the retrieved set (Global cancer statistics 2022: GLOBOCAN estimates…) |
[#5] | ✅ In the retrieved set (2023 ESH Guidelines for the management of arterial hypertension…) |
[#8] | ✅ In the retrieved set (Macrophages in immunoregulation and therapeutics…) |
[#25] | ✅ In the retrieved set (Enfortumab Vedotin and Pembrolizumab in Advanced Urothelial Cancer…) |
[#36] | ✅ In the retrieved set (The cancer-immunity cycle: Indication, genotype, and immunotype…) |
[#39] | ✅ In the retrieved set (Neoadjuvant–Adjuvant or Adjuvant-Only Pembrolizumab in Advanced Melanoma…) |
[#92] | ✅ In the retrieved set (Bladder cancer — Nature Reviews Disease Primers) |
Note: verification checks whether each citation maps to a real record retrieved in this task (existence + semantic consistency). It shows that the source exists — not that the way it was summarised is necessarily correct.
Spot-check of citation–content consistency (abridged)
| Citation | Result |
|---|---|
[102] | Consistent — the statement that in CheckMate-901 nivolumab plus chemotherapy significantly prolonged overall survival versus chemotherapy alone, with more treatment-related adverse events, matches the abstract. |
[5] | Uncertain — the abstract was not present in the supplied list, so the corresponding statement could not be verified. |
[1] | Inconsistent — the abstract does not match what the text claims at that citation, and the content is not cited or discussed there. |
These three entries come from real outputs of different documents. Verification marks “unable to confirm” and “inconsistent” results explicitly for human judgement; the platform does not hide them and does not claim citations are always correct. If fewer than 70% pass overall, the generation is not charged.
Sample 2 ｜ English review (Sacral Neuromodulation for LUTD)
22 citations checked: 22 passed, 0 failed or pending human review.
| Citation | Status and basis |
|---|---|
[#6] | ✅ In the retrieved set (Neuromodulation for Lower Urinary Tract Dysfunction…) |
[#7] | ✅ In the retrieved set (Remote programming in stage I sacral neuromodulation…) |
[#14] | ✅ In the retrieved set (Efficacy and Safety of Sacral Neuromodulation…) |
[#17] | ✅ In the retrieved set (Sacral Neurostimulation in Neurogenic Lower Urinary Tract Dysfunction…) |
[#19] | ✅ In the retrieved set (Should We Consider Sacral Nerve Stimulation…) |
Deep-mode output (plan → section-by-section drafting → adversarial review → revision) also includes a structured quality scorecard and a list of evidence gaps.
Full sample (downloadable · English)
This is a real platform output (2026-09-07, topic: myeloid cell mechanisms in neurodegenerative disease),unedited— download it to see the section structure, reference formatting and the verification report exactly as generated.
| Scale | 120 of 221 papers included (quality-first screening) ｜ 89,839 characters of body text ｜ 19 sections |
|---|---|
| Citation verification | 67 of 67 passed verification (existence + semantic consistency) |
| Section structure | Introduction｜Microglia: Guardians and Threats｜Monocyte-Derived Macrophages｜Phagocytic Clearance of Pathological Proteins｜Metabolic Reprogramming｜Inflammasome Signaling and Pyroptosis｜Trophic Factor Secretion｜Extracellular Vesicles｜Gut–Brain Axis｜Aging and Immunosenescence｜Single-Cell Technologies｜Therapeutic Implications |
| Download | 📥 Word（DOCX） · Markdown |
The download carries the AI-generated content label (a label section in the body plus file metadata). It is an AI-assisted first draft, shown only to illustrate the output format, and does not constitute any scholarly conclusion.
What one generation task gives you
- First draft: Word (DOCX) and Markdown
- References: RIS / BibTeX / EndNote XML / PMID list, rendered programmatically from your target journal's template
- Citation verification report and a consistency spot-check record
- A structured quality scorecard (heuristic prompts, not a formal risk-of-bias assessment)
- Exports carry the AI-generated content label (in the interface and in file metadata)
Scope and limits
- All of the above areAI-assisted first drafts, to be verified by the researcher before use;
- they do not replace the standard workflow of a systematic review or meta-analysis, and are not medical advice or a basis for clinical decisions;
- follow your target journal's and institution's AI-use and disclosure rules, and do not remove or obscure the AI label before sharing.