HumanCite · Methodology

How we verify.

Purpose-built software screens every manuscript the moment it arrives. Experts make every judgment. Nothing leaves without a human signature. This is the full method — what we check, how we grade it, who stands behind it, and what you receive.

  1. ReceivedManuscript in. A case opens.
  2. EnginesRetraction · quality · screening · language.
  3. Human judgmentEvery citation verified by a person.
  4. SignedDouble-checked, then issued.
01WHAT WE CHECK

Software screens at scale. A human judges every citation.

Human review

Source Verification

The core of every review. An expert human finds the original source behind each citation and confirms it is real — it exists, it can be reached, and the authors, title, year, and publisher match the official record.

Every verdict is double-checked before your report is issued.

Automated

Retraction Check

Has anything you cite been retracted, corrected, or seriously questioned since it was published? Every citation is checked against a continuously updated corpus of retraction and correction notices. A person confirms anything the engine finds.

Automated

Source Quality Check

Are your sources reputable? We assess the standing of the journals, outlets, and institutions behind your citations — including known predatory publishers — and report it as context beside each citation.

Applies to journal articles, institutional reports, and news sources.

Automated

Problematic Paper Screening

Some published papers are known to be unreliable. We screen your cited papers against research-integrity warning databases and flag any match for human attention.

Covers every cited paper with a registered identifier — and we tell you exactly which ones could not be screened.

Automated

Manuscript Language Scan

Your full text is scanned for “tortured phrases” — garbled wording that is a known fingerprint of auto-generated writing.

Requires the full manuscript.

Optional add-on

In-Text Citation Check

Does every reference in your list actually appear in your text? We trace each one into the body, count how often it is cited, and show whether any single author dominates.

02THE H-STANDARD

Four verdicts. Each one a position you can defend.

Every check feeds one decision: the verdict each citation receives.

01
H-Verified
CONFIRMED
Six blocks match the publisher record. The reference is sound.
02
H-Flagged
CORRECTABLE
Source exists; one or more blocks disagree. Author is asked to correct.
03
H-Fabrication
NOT REAL
Positive evidence the source does not exist — not just absence of evidence.
04
H-Unresolved
INACCESSIBLE
Cannot confirm. Cannot evidence fabrication. We say so.
SAMPLE REFERENCE

Hartmann, K., & Reyes, D. (2021). Reviewer cognition under temporal pressure. JSP, 47(3), 211–238.

SIX-BLOCK BREAKDOWN
Creator
Title
Place
Date
Publisher
Specific item
03WHO SIGNS

Real people, independent of the work, sign every report.

A person makes the call on every citation — every time. If it hasn’t been read by a human, it hasn’t been H-Verified. And the people checking are independent: not the author, not the publisher, not the tool vendor.

  • One standardEvery reviewer works to the same published standard — the four verdicts of the H-Standard — and the quality of their work is continuously reviewed.
  • Double-checkedEvery verdict gets a second pair of eyes before the report is signed.
  • If you disagreeOpen a dispute from the report itself. The case is re-examined and the report reissued.

An audit only has weight if the auditor is accountable, the method is published, and the conclusions are reproducible.

04A SAMPLE REPORT

What a verified report delivers.

Every manuscript comes back as a signed report: a verdict for each reference, the evidence behind it, with source quality and retraction status reported separately. The document below is illustrative — same structure, synthetic data.

HumanCite Verification ReportSample · illustrative data
CaseHC-0626-0148
SubjectSubmitted manuscript · 24 references checked
SignedReviewed and signed · 04 Jun 2026
H-Verified19
H-Flagged3
H-Fabrication1
H-Unresolved1
Issues identified
  • 1 fabricated reference
  • 1 retracted source
  • 1 predatory journal
References by publication year
ReferencesShowing 6 of 24 references
ConfirmedStated mismatchWrong / unfindableNot applicable
01Mensah G., Okafor L. — Adaptive load balancing in distributed sensor networks. Journal of Systems Engineering, 2022.H-Verified
CreatorConfirmedTitleConfirmedPlaceConfirmedDateConfirmedPublisherConfirmedSpecific itemConfirmed
Where we found itExact match — DOI 10.1016/j.jse.2022.04.011 resolves to the cited article.
Why we ruled this wayAll six fields match the publisher record. DOI resolves; authors, title, year and volume confirmed.
Source qualityQ1 · Scopus indexed · SJR 2.41 · h-index 188SCImago — descriptive context, not a verdict
02Adeyemi K. — Nanofiltration efficiency in arid-zone desalination. Global Journal of Applied Engineering, 2021.H-Flagged
Source-quality concern
CreatorConfirmedTitleConfirmedPlaceConfirmedDateConfirmedPublisherConfirmedSpecific itemConfirmed
Why we ruled this wayThe reference is accurately cited, but the journal appears on Beall’s list of potentially predatory publishers. Surfaced as descriptive context so the author can decide.
Predatory listListed (Beall’s)Descriptive source-quality flag — the reference itself is accurately cited.
03Vasquez R., Lindqvist H. — Reinforcement priors for low-resource translation. Computational Linguistics Review, 2023.H-Fabrication
Blended hallucination
CreatorWrong / unfindableTitleWrong / unfindablePlaceConfirmedDateWrong / unfindablePublisherConfirmedSpecific itemMissing
Where we found itNo record matches the cited DOI (10.1080/clr.2023.5567) — it does not resolve. Computational Linguistics Review is a real journal, but has never published an article under this title or by these authors.
Why we ruled this wayA real journal stitched to fabricated authors and a non-resolving DOI — a blended hallucination. We won’t call a citation fabricated without proof: confirmed against the publisher’s full index and Crossref, with at least two independent sources agreeing no such article exists.
04Halloran T. — Microbiome diversity and metabolic outcomes. Clinical Nutrition & Metabolism, 2019.H-Verified
CreatorConfirmedTitleConfirmedPlaceConfirmedDateConfirmedPublisherConfirmedSpecific itemConfirmed
Why we ruled this wayCorrectly cited — but the source was retracted after publication. Retraction is reported separately from the verdict: a citation can be H-Verified and retracted.
Retracted — retraction notice, 14 Mar 2022. Reason: image duplication.Screened against a continuously updated Retraction Watch corpus.
05Working paper — National Institute for Labour Studies, “Wage dynamics in informal markets”, 2024.H-Unresolved
CreatorConfirmedTitleConfirmedPlaceConfirmedDateConfirmedPublisherNot applicableSpecific itemWrong / unfindable
Where we found itA working paper with no DOI and no indexed record. The institution is real, but the document itself cannot be independently confirmed.
Why we ruled this wayGenuinely unverifiable — returned honestly as H-Unresolved, not fabricated. We won’t call a citation fake unless we can prove it.
06News analysis — “Regional grid failures and demand spikes”, The Continental Post, 2023.H-Flagged
Misrepresentation
CreatorConfirmedTitleConfirmedPlaceConfirmedDateConfirmedPublisherConfirmedSpecific itemConfirmed
Why we ruled this wayThe outlet and article are real, but the figure attributed to them is not stated in the source. Flagged for misrepresentation; source-quality context shown for transparency.
Source assessmentFactual reporting — Mixed · Bias — CenterMedia Bias / Fact Check — third-party, descriptive

What’s in every report.

Every source read by an expert humanVerified against its real source — not an automated confidence score or a database lookup.
Checked twiceA second reviewer examines every verdict before the report is signed. No verdict rests on one person.
We won’t call a citation fabricated unless we can prove itH-Fabrication requires positive evidence from at least two independent sources; the genuinely unverifiable comes back honestly as H-Unresolved.
Checked field-by-field across six blocksCreator, Title, Place, Date, Publisher, Specific item — with retraction status and source quality reported separately.

Ready to verify your references?

Get started →Talk to us →

HUMANCITE · INFO@HUMANCITE.COM