Skip to content
OctymPowered By Proofig AI
What we check

Octym runs every integrity and quality check, in one review.

The complete package, grouped into three categories: what's inside the manuscript, how it stands against the published record and the requirements of wherever it's headed, and whether the science is sound. Every finding is a signal backed by evidence, never a verdict.

10 checks

Internal manuscript integrity

Issues that originate inside the manuscript itself: manipulated or duplicated figures, numbers that don't add up, captions that don't match. The clear, must-fix problems, surfaced first.

AI-generated text signals

Calibrated signals of AI-generated or paraphrased writing, aware that non-native English is not evidence of anything. A signal for a person to weigh, never a determination about an author.

Image manipulation

Pixel-level forensics that flag where a figure has been cloned, spliced, or retouched, tuned for the image types that matter most in life science.

Powered by Proofig AI

Clone & spliceWestern-blot manipulationMicroscopy edits

Image duplication

Compares every figure against the others to catch reused or overlapping panels, including FACS / flow-cytometry.

Powered by Proofig AI

Within-paper duplicationFACS / flow cytometry

AI-generated images

Identifies synthetic, AI-generated scientific figures, surfaced as a signal with the evidence behind it.

Powered by Proofig AI

Graph & chart duplication

Finds data-visualizations reused or repurposed across the manuscript. A distinct forensic from photographic images.

Powered by Proofig AI

Statistical & analytical consistency

Checks the reported statistics against each other: p-values, test statistics and degrees of freedom that are not consistent with one another, percentages and counts that are impossible for the stated sample, confidence intervals that cannot be right, and results that disagree between text, tables and figures. Concrete arithmetic inconsistencies, quoted with the reported values.

p-value & test-statistic consistencyImpossible percentages, counts & intervalsText, table & figure matchEffect-size & uncertainty reporting

Figure–legend consistency

Reads figures, their legends and the text that cites them together, and flags mismatches: legends that describe something different from what is shown, missing labels or units that make a figure uninterpretable, captions that contradict the methods or results.

Legend, figure & text match

Claim & citation consistency

Flags statements that need a source but lack one, and citations that don't align with the claim they support, each shown in context.

Powered by RefGuard, Proofig AI's reference integrity product

Claim ↔ citation alignmentUncited-claim flag

Internal consistency

Reads the document as a whole and flags where it contradicts itself: definitions and eligibility rules that change between sections, registered or stated plans the results don't follow, and numbers that differ between text, tables and supplements. For grant applications, checks that the proposal, CV, budget and governance material describe the same project.

Plan-to-report consistencyCross-section consistency (applications)

Prior publication & reuse disclosure

Reviews what the manuscript says about preprints, earlier publications, reused datasets and its relationship to the authors' related work, and flags results that appear to be reused from related same-author work without a clear explanation.

Preprint & prior-publication statementsUndisclosed related-work reuse
10 checks

External checks & requirements

Checks that compare the work against everything outside it: the published literature, authoritative records, and the requirements of wherever it's headed.

Image & graph plagiarism

Compares images and data-visualizations against the published literature to surface figures reused from other papers.

Powered by Proofig AI, the largest image-plagiarism database in the industry

Retracted reference detection

Flags citations to retracted papers, with the retraction date, nature and reason attached, so you can decide whether to replace the source.

Powered by RefGuard, Proofig AI's reference integrity product

Hallucinated / non-existent references

Compares each reference with the scholarly publishing record to confirm the article is real. References that can't be matched to any real publication are flagged for you to check: this is where invented or hallucinated citations surface.

Powered by RefGuard, Proofig AI's reference integrity product

Incomplete & broken references

Shows each reference's published details (authors, year, journal) beside the citation as written, so a wrong year, a missing author or a broken link is easy to spot.

Powered by RefGuard, Proofig AI's reference integrity product

Reference analytics

Analyzes the reference list as a whole: how many references were verified, how citation counts and publication years are distributed, how much of the cited work is open access, and which journals and work types dominate. Signals for a person to weigh, not verdicts.

Powered by RefGuard, Proofig AI's reference integrity product

Data & code availability

Resolves the data, code and materials a manuscript says are available, confirming the links actually lead somewhere, and checks that availability statements are present and specific.

Paper-mill signals

Surfaces the signals that cluster in paper-mill submissions, each from its own check: figures reused from other papers, authorship and affiliation inconsistencies, undisclosed reuse of related work, references that don't support the claims, and missing disclosure of the original data files behind key images. Each is a signal with its evidence attached, never a conclusion about the authors.

Authorship & affiliation consistency

Checks author names, affiliations, correspondence details and contribution statements for completeness and internal consistency, and flags the authorship and accountability gaps that often accompany paper-mill submissions.

Funder requirements

Reviews an application against the funder's own criteria: scheme fit and eligibility, required structure and annexes, rigor of the approach and of the preliminary evidence, budget justification, data-management and sharing plans, ethics, and the declarations the scheme requires. Checks are built for NIH R01 and R21 and for ERC Starting, Consolidator and Advanced grants.

Scheme fit & eligibilityStructure, annexes & registrationsRigor of approach & preliminary evidenceBudget, timeline & delivery riskData management & sharing planEthics, security & declarations

Submission requirements & completeness

Checks the submission against what editors otherwise verify by hand: scope and article type, completeness, required statements and declarations, ethics and registration, reporting completeness, and rights and permits. Journal-specific rules are added to your review profile by your Octym team at onboarding.

Scope & article typeRequired statements & declarationsRegistration & protocol transparencyReporting completenessRights, permits & dual-use
REVIEWMethodsRigorData
6 checks

Research quality

Analysis of the research itself: whether the methods are sound, the work could be reproduced, and the science holds up.

Methods & rigor

Evaluates the soundness of the design (randomization, blinding, controls, and whether the sample size is justified) and whether the methods are reported clearly enough to be appraised.

Randomization, blinding & controlsSample-size & power reportingMeasurement & resource validityReporting completeness & methodological clarity

Reproducibility

Checks that the protocol, computational environment, and analysis are described completely enough to reproduce the result.

Protocol completenessComputational environment

Replicability signals

Reads the signals that bear on whether a finding would hold up: whether the sample and information are adequate, whether the analysis is robust to its assumptions and choices, and whether the claims stay within what the evidence supports. Signals for a person to weigh, never a prediction.

Sample & information adequacyAnalytical robustnessClaim calibration

Research ethics

Checks that ethics / IRB approval, informed consent, conflicts of interest, funding, and AI use are disclosed and internally consistent, with a named human accountable for the work.

IRB & consentCoI & fundingAI-use disclosure & human accountability

Novelty & contribution

Checks whether the manuscript states its research question, the gap it addresses and the contribution it claims, and whether that framing is consistent with the evidence and citations it presents. For grant applications, whether the case for importance and innovation is made against current practice.

Question, gap & contributionConsistency with the evidence presented

Peer-review-grade analysis

Synthesizes every check into a structured, reviewer-style Assessment: major issues first, each comment anchored to the passage it's about, alongside a Summary of the manuscript. Decision-support for a human, not an accept/reject verdict.

Signals, not verdicts. Octym surfaces findings with evidence and traces each to its source. A person always decides what matters. Octym never accuses anyone of misconduct.

FAQ

The checks, answered.

Octym reviews a manuscript across three categories: Internal manuscript integrity (AI-generated text signals, Image manipulation, Image duplication, AI-generated images, and more); External checks & requirements (Image & graph plagiarism, Retracted reference detection, Hallucinated / non-existent references, Incomplete & broken references, and more); Research quality (Methods & rigor, Reproducibility, Replicability signals, Research ethics, and more). Every check runs in one review, and every finding is traced to its source.

No. Every finding is a signal backed by evidence, never a verdict. Octym shows what it found and where; a person decides what it means. It does not accuse anyone of misconduct.

Image and figure integrity is powered by Proofig AI, a trusted leader in scientific image integrity, adopted by the Science family of journals and trusted by the world's top researchers, publishers, and research institutes.

They are all available on every paper; your review profile decides what runs and what leads. You decide, and your Octym team makes it happen.

Funders: yes. NIH R01 and R21 and ERC Starting, Consolidator and Advanced applications are checked against the scheme's own criteria, in the same review. Journals: for research institutions and departments, your Octym team configures the review at onboarding to the journals your researchers submit to; for publishers, to each journal's own instructions for authors.

Octym flags signs of manipulation, duplication, AI generation and image plagiarism in a manuscript's figures, using image forensics powered by Proofig AI. Octym finds cloned, spliced or retouched regions in Western blots and microscopy; images reused within the manuscript, even when resized, rotated or partly overlapping, FACS and flow-cytometry panels included; AI-generated figures; and duplicated graphs and charts. Every figure is also compared against the published literature, powered by Proofig AI, the largest image-plagiarism database in the industry. A person decides on each signal.

Octym catches p-values, test statistics and degrees of freedom that disagree, percentages and counts impossible for the sample, and confidence intervals that cannot be right. Octym also flags results that differ between the text, tables and figures, and gaps in effect-size and uncertainty reporting, quoting each inconsistency with the reported values. Its methods and rigor check asks whether sample size and power are justified. Octym checks the numbers against each other, in the same review that reads the text, references and figures. A person decides what each signal means.

Yes, Octym reads the whole manuscript in context, so it catches where a figure, its legend, the tables and the text disagree. Octym flags legends that describe something other than what is shown, missing labels or units, and captions that contradict the methods or results. It catches numbers that differ between text, tables and supplements, stated plans the results do not follow, and definitions that change between sections. A tool that reads only the text never sees the figures. A person judges each mismatch.

Octym flags passages that read as AI-generated or paraphrased, as calibrated signals for a person to weigh, never as a determination about an author. The check is built knowing that non-native English is not evidence of anything. Octym reviews the entire paper across three categories of checks, so a text signal sits beside AI-generated image detection, powered by Proofig AI, and the authors' AI-use disclosure. No flag is proof, and a person decides.

Yes, Octym's AI agent explains in plain language why a figure was flagged, with every answer grounded in the manuscript and linked to its source. Each finding is already marked at its exact location with the evidence behind it; the agent lets you go further: summarize the reproducibility gaps or check whether the ethics statement is sufficient. Asked why, a general-purpose AI chatbot answers without grounding, and a single-purpose tool returns its flag without an agent you can question. The agent comes with every Octym package.

Yes, Octym checks that a manuscript discloses its use of AI consistently, with a named human accountable for the work, in its research ethics check. The same check covers ethics approval, consent, conflicts of interest and funding, and the submission requirements check flags missing required statements and declarations. Each gap is marked at its exact location, with the evidence behind it. A person decides what each finding means.

Octym reviews scientific manuscripts and grant applications across three categories of checks, covering text, statistics, methods, references and required statements. Octym's image forensics, powered by Proofig AI, are tuned for the image types that matter most in life science, such as Western blots and microscopy, and graph and chart checks cover plotted data. Its reference checks are powered by RefGuard, Proofig AI's reference integrity product. For institutions and publishers, journal profiles include clinical, preclinical and observational titles, measured against CONSORT, ARRIVE and STROBE. A person decides on every finding.

Octym finds citations to retracted papers, hallucinated references, incomplete or broken references, claims missing a citation and citations that do not support their claims. These checks are powered by RefGuard, Proofig AI's reference integrity product, and run together in one review. Each retracted citation shows its retraction date, nature and reason, and an overview of the bibliography shows publication years, citation counts and open-access share. Every finding is marked in context for a person to review.

See every check on any manuscript.

Contact us, or log in.