Sample review · The manuscript, authors, journal, and every citation below are invented to show the form. Identifiers are illustrative and do not resolve. No client work appears here.
A complete review, start to finish
Below is a full decision record in the exact shape of a real one, run against an invented manuscript on vitamin D and cardiovascular risk. It shows what a review finds, how each finding is stated, where the review corrects its own work, and what it deliberately leaves to you. Nothing here is gated. It is yours to read, and yours to forward.
Manuscript Review · Verified at the primary source
The findings
Every finding is a receipt, not a summary: the manuscript’s own words on the left, the source’s own words on the right, then why it matters to a reviewer and the repair available now. The two columns let you judge the gap yourself, which is the whole point.
A causal claim the cited source does not support
As the manuscript has it · Discussion, paragraph 3 (cites reference 7)
“Correcting vitamin D deficiency reduces the incidence of major adverse cardiovascular events.”
At the primary source · reference 7
“In this observational cohort, lower baseline 25(OH)D was associated with a higher incidence of events; the design cannot establish causation, and these data should not be read as evidence that supplementation lowers risk.”
Why it matters
This is the kind of claim a reviewer stops on: an observational source used to support a causal, interventional statement. It is also the most common way a careful paper draws an avoidable rejection.
Repair available now
Either match the language to what the source supports (“is associated with a lower incidence of”), or cite an interventional source that tests supplementation and events. Both are available in the current literature.
The abstract hedges; the discussion does not
In the abstract
“Lower baseline 25(OH)D was associated with a higher incidence of major adverse cardiovascular events.”
In the discussion, same manuscript
“Our findings demonstrate that correcting deficiency lowers cardiovascular risk.” The limitations section commits to the hedged register the abstract uses.
Why it matters
Reviewers read the abstract and the conclusion first, and a mismatch between them reads as over-claiming. The limitations section you already wrote is the register the discussion should match.
Repair available now
Harmonize the discussion to the hedged register the abstract and limitations already use. No new analysis is required.
A listed identifier does not resolve to the cited paper
As cited · reference 3
“Dominguez-Katz and Prahl. Competing-risks regression for cardiovascular endpoints. doi:10.5555/cmp.2026.0130”
At the registry
Author, title, journal, and year all match the intended paper; only the identifier is wrong. As listed, 10.5555/cmp.2026.0130 does not resolve; the correct identifier is 10.5555/cmp.2026.0103, with the final digits transposed.
Why it matters
A reference that does not resolve is a copy-editing flag at best and, at some journals, a desk-check failure. It is quick to fix and worth fixing before a reviewer finds it.
Repair available now
Correct the identifier to the one that resolves to the intended paper. Confirm the corrected DOI against the registry record before submission.
A listed identifier resolves, but not to the cited paper
As cited · reference 4
“Weiss and Colombo. Serial 25(OH)D and arterial stiffness. doi:10.5555/cmp.2026.0104”
At the registry
The record at 10.5555/cmp.2026.0104 is a different article, not the work cited. This review confirmed only that the listed identifier is not the correct one; it did not establish which paper the identifier leads to, and does not claim to.
Why it matters
A citation that points somewhere other than intended is a substantive accuracy issue, not a typo: a reader who follows it lands on the wrong work.
Repair available now
Locate the intended paper, cite its correct identifier, and confirm the match at the registry. If the intended paper cannot be located, the claim it supports should be re-sourced.
A co-author is listed who is not on the record
As cited · reference 5
“Refeld J, Ostroff P, et al. Vitamin D supplementation and cardiovascular events: a meta-analysis.”
At the registry
The record at this identifier lists Refeld J and three co-authors; none of them is Ostroff P.
Why it matters
Author-list mismatches are a citation-integrity flag, and some screening tools surface them automatically after submission.
Repair available now
Correct the author list to match the record at the registry.
A source in a different population is cited for a general claim
As the manuscript has it · Introduction (cites reference 6)
“Cholecalciferol supplementation improves cardiovascular outcomes in adults.”
At the primary source · reference 6
“This randomized trial enrolled 214 adults receiving maintenance hemodialysis.” The finding is real and correctly cited, but the population is narrower than the sentence implies.
Why it matters
Scope drift across populations is a claim-support issue a reviewer in the field will notice, because dialysis physiology is not the general community population your study addresses.
Repair available now
Either scope the sentence explicitly to the dialysis population, or cite a source in the community population your manuscript concerns.
A prevalence figure the cited source does not report
As the manuscript has it · Introduction (cites reference 8)
“Roughly 40 percent of adults are vitamin D deficient.”
At the primary source · reference 8
“Using a 25(OH)D threshold of 30 nmol/L, deficiency prevalence in the sample was 18 percent.” The thresholds differ, and the figure the source reports is not 40 percent.
Why it matters
A specific prevalence figure attached to the wrong source is exactly the kind of claim a source review exists to catch: it looks sourced, and it is not.
Repair available now
Cite the 40 percent figure to a source that reports it at the threshold you state, or revise the figure to the one your cited source actually reports.
A cited preprint has been withdrawn
As the manuscript has it · Discussion (cites reference 9, a preprint)
“Vitamin D receptor signaling suppresses vascular inflammation.”
At the source · reference 9
The preprint at this identifier carries a withdrawal notice from its authors, so the claim cannot be confirmed against it. Reported as could-not-confirm: the review does not assert the claim is wrong, only that it cannot be verified against the source as cited.
Why it matters
A withdrawn source cannot carry a load-bearing claim into submission, and a reviewer who checks will find the withdrawal notice.
Repair available now
Replace the citation with a source that can be confirmed, or remove the claim. If a peer-reviewed version of the withdrawn work now exists, verify it directly before citing it.
Every check, at its primary source
All sixteen references, each resolved and read where it lives. Identifiers here are illustrative and do not resolve; in a real review each links to the live registry record so you can check it yourself.
| Reference as listed | What we found at the source | State |
|---|---|---|
| 1. Prahl, Achterberg Cohort profile of the community study 10.5555/cmp.2026.0101 | Resolved. Authors, title, year, and journal match the record. | Verified at source |
| 2. Achterberg et al. 25(OH)D immunoassay standardization 10.5555/cmp.2026.0102 | Resolved. Cited accurately as a methods reference. | Verified at source |
| 3. Dominguez-Katz, Prahl Competing-risks regression methods 10.5555/cmp.2026.0130 | Listed as ...0130 does not resolve; the correct identifier is ...0103 (final digits transposed). | Needs your decision |
| 4. Weiss, Colombo Serial 25(OH)D and arterial stiffness 10.5555/cmp.2026.0104 | Identifier resolves, but not to the cited work. Correct identifier not established. | Needs your decision |
| 5. Refeld, Ostroff et al. Supplementation meta-analysis 10.5555/cmp.2025.0410 | Record has a different author list; one listed co-author is not on it. | Needs your decision |
| 6. Ibarra et al. Cholecalciferol in hemodialysis 10.5555/cmp.2025.0288 | Resolves and is accurate; population (dialysis) is narrower than the citing sentence. | Needs your decision |
| 7. Santos-Reyes et al. Observational 25(OH)D and incident events 10.5555/cmp.2025.0173 | Resolves; is observational and cautions against causal inference. Cited for a causal claim. | Needs your decision |
| 8. (prevalence source) Deficiency prevalence estimate 10.5555/cmp.2024.0512 | Resolves; reports a different figure at a different threshold than the manuscript states. | Needs your decision |
| 9. Okafor et al. VDR signaling and vascular inflammation (preprint) preprint, withdrawn | Withdrawn by authors. Claim cannot be confirmed against it. | Could not confirm |
| 10. (clinical guideline) Vitamin D guideline, current edition 10.5555/cmp.2026.0007 | Resolved. Edition and population match the citing sentence. | Verified at source |
| 11. Lindqvist et al. Mechanistic review, cited as background 10.5555/cmp.2025.0061 | Resolved. Framed as background, not as clinical evidence. | Verified at source |
| 12. Barros et al. Randomized supplementation trial 10.5555/cmp.2024.0399 | Resolved. Cited accurately for its actual (null) primary result. | Verified at source |
| 13. Halvorsen, Prahl Cohort methodology reference 10.5555/cmp.2023.0140 | Resolved. Accurate. | Verified at source |
| 14. (statistical software) Analysis software citation 10.5555/cmp.2022.0031 | Resolved. Standard methods citation, accurate. | Verified at source |
| 15. Nakamura et al. Prior systematic review 10.5555/cmp.2024.0250 | Resolved. Cited accurately for its stated conclusion. | Verified at source |
| 16. Achterberg, Weiss Assay reproducibility reference 10.5555/cmp.2025.0119 | Resolved. Accurate. | Verified at source |
Nine verified clean, six needing a decision, one could not be confirmed. That is the honest tally, carried up from the record above without rounding it into a single word.
The review corrects its own work
“On reference 4, an earlier draft of this review wrote that the wrong identifier ‘resolves to an unrelated article on a different topic.’ A check of the review against itself, before release, caught that this claimed more than had been verified: calling the other article ‘unrelated’ is a judgment about a paper the review never read. Corrected: the review confirmed only that the listed identifier is not the correct one for the cited work. It did not establish which paper the wrong identifier leads to, and does not claim to.”
Before release, every review is checked against itself, and any place an earlier draft over-stated is disclosed here rather than quietly fixed. A review that shows you its own catch is the one you can trust with your manuscript. We would rather show you a catch than claim we never need one.
Two columns, kept separate
We verify references, citations, and claim support against their primary sources. We do not judge the science’s merit. That is yours.
How the check was done
Every reference was resolved and read at the registries where it lives, field by field: does the paper exist, is the identifier correct, and does the source say what your sentence attributes to it, at the same scope and population. Abstracts were read; title-only matching was not used. Journal requirements were checked at the journal’s own live author instructions, not from memory. The whole manuscript was covered, including references inside tables and figures.
The verification software is AI-assisted, and we assume it can err. So the discipline is built around independence: every load-bearing finding is re-checked through a separate pass that does not share the first pass’s inputs, with each identifier re-resolved at the registry, and a person re-derives it before release. Independence here is a property of the input chain, not of who does the checking. Where something cannot be confirmed, the review says could not confirm. It never assumes, because a false “verified” is the one failure we treat as unacceptable.
A person reads the complete record, re-derives the findings the paper rests on, and releases the review by hand. The software is structurally incapable of sending anything on its own.
What it did and did not assess
This review assessed citation fidelity, claim support, identifier resolution, internal consistency across the manuscript’s own sections, and fit to the target journal’s instructions. It did not assess the merit of the science, the correctness of the statistics beyond whether the citations support them, the ethics, or the authors’ intent. Those are the authors’.
If any sentence in this review does not match its cited primary source, that is a fidelity error we want to correct. Cross-checking against the cited sources is invited.
Every review ends by inviting the author to check it. A review that invites its own falsification is the point.
Released
Read in full, re-derived at the primary source, and released by hand.
Ryan Gruzen · ASI Source Review
Every review ends this way: a named person is accountable for it, and the software is structurally incapable of sending anything on its own. On a live review this block carries the releaser’s signature and the date; here it is illustrative.
How you would interface with it
You follow the review inside a private portal while it is underway. Nothing releases itself; a person does, by hand. These are illustrative views of that interface partway through an engagement, before the release shown above. The data is invented.
Verification record
Cited: randomized trial, null primary result. Found at source: matches, cited accurately.
Cited: meta-analysis, listed co-author not on record. Correction proposed for your approval.
Cited: preprint. Found at source: withdrawn by authors. Reported as could not confirm.
Where it stands
Manuscript entrusted
Confidentiality agreement signed first. Then the near-final manuscript and target journal.
Received and scoped
Coverage confirmed across body, tables, figures, and reference list.
In verification
Every citation resolved and read at its primary registry.
A person reads and releases
A human re-derives the load-bearing findings and releases the review by hand. Never automated, never skipped.
Released to you
The complete decision record, yours to check line by line.
The delivered record is also portable: it can be copied as Markdown or plain text, or printed to PDF, so it travels with you into whatever you write next. Nothing about the review is ever released without a person reading it first.
This is what you receive
A full review plus one re-review of your revision. If you have a high-stakes submission ahead, send word. The confidentiality agreement comes first; the manuscript only after it is signed. This sample is yours to keep, and yours to pass to a colleague with a submission coming up.