Manusights vs Rigorous AI Review: Production Platform vs Research Project

Rigorous AI Review was created by Dr. Robert Jakob and Dr. Kevin O'Sullivan, postdoctoral researchers at ETH Zurich and co-directors of the Agentic Systems Lab. The about page says the project is run in their personal capacity and is independent from their ETH research roles. The GitHub repository is open source under an MIT license.

The tool provides AI-generated feedback on methodology, clarity, and impact. Users upload a manuscript, specify a target journal, and receive a structured report. The web version (v0.2) launched at rigorous.review with an interactive interface and progress tracking.

Rigorous is genuinely interesting as an academic AI-review project. The homepage says it has already been tested by researchers from ETH Zurich, Stanford, Harvard, MIT, the University of Bern, and the University of St. Gallen. The focus on methodology, clarity, and impact feedback addresses a real drafting-stage gap.

But Rigorous is not the same kind of product as Manusights. Its own terms make clear that feedback is automatically generated and does not constitute formal peer review. That distinction matters for two reasons: data handling and submission-decision reliability.

This is the most consequential distinction between the two platforms. Read Rigorous's terms carefully before uploading any manuscript.

What Rigorous's terms of service say:

1. Manuscripts are "temporarily stored on secure servers provided by Backblaze, a third-party cloud storage service, to facilitate processing"
2. Manuscript content is "processed using large language models (LLMs) such as OpenAI APIs to generate feedback", meaning data is "sent to and processed by OpenAI's systems"
3. The service "may utilize anonymized and aggregated data from our service for research purposes" with findings potentially published in academic journals
4. The privacy policy says manuscript files are not stored longer than necessary to complete processing and provide feedback, but it does not define a concrete deletion window
5. The terms put responsibility on users to avoid sharing confidential, sensitive, or proprietary information

That last point is worth pausing on. If you are working with unpublished clinical trial data, proprietary pharmaceutical methods, patentable inventions, embargoed results, or institutional data under compliance requirements, the current Rigorous terms are not written like a confidential manuscript-review contract.

What Manusights's data handling looks like:

• Manuscripts are not used for model training; Anthropic-processed content is handled under provider-side zero-retention terms
• TLS 1.2+ encryption in transit, AES-256 encryption at rest
• Manusights retains only the operational records and review artifacts needed to deliver, support, secure, bill, and audit the requested product

This is not about which AI model is smarter. It is about what happens to unpublished work after processing. For a researcher uploading a draft six weeks before submitting to Nature Medicine, the difference between Backblaze plus OpenAI API processing and no model training with bounded Manusights-side retention may be the comparison that matters most.

Beyond privacy, the analytical scope differs substantially.

Rigorous provides:

• AI-generated comments on methodology, clarity, and impact
• Structured feedback report
• Interactive v0.2 web interface with progress tracking
• Free access on the current public web tool

Rigorous does not provide:

• Citation verification against any database
• Vision-based figure analysis
• Journal-specific readiness scoring or desk-reject risk
• Quantified readiness score
• Prioritized revision checklist (A/B/C by impact)
• Human expert escalation
• Any formal guarantee about output quality (their terms state the output "does not constitute formal peer review")

Manusights provides (at each tier):

• Free scan ($0): Readiness score (0-100), desk-reject risk, top issues, journal-fit verdict, about two to three minutes
• Full Review ($39): Citation-integrity checks against scholarly metadata sources (CrossRef, PubMed, OpenAlex, Semantic Scholar, bioRxiv, medRxiv), vision-based figure analysis, journal-specific scoring across 5 dimensions, prioritized A/B/C revision checklist, 30 minutes
• Expert review ($1,000+): Named field expert or CNS-level editor, full NDA, 3-7 day turnaround

In our analysis of manuscripts that reach the pre-submission stage, the drafts tempted by Rigorous usually split into two buckets.

The first bucket is early-stage academic drafting, where the team wants a free second opinion on methods and is not especially worried about confidentiality. Rigorous can make sense there.

The second bucket is much riskier: authors with selective-journal targets, unpublished data, or commercial sensitivity who are attracted by the free price but have not matched that choice to the manuscript's risk profile. That is where the decision usually gets made on the wrong variable. The real question is not whether zero dollars beats thirty-nine dollars. It is whether a research project with OpenAI processing and unspecified deletion is the right place to upload this particular draft.

The pattern we see is specific. A draft can get useful methodology comments and still fail at Nature Medicine because Figure 3 lacks the control needed for the main claim, or at JAMA because the cohort endpoint does not support the clinical conclusion, or at Cell because the novelty paragraph ignores a recent competing paper. Those are not generic "impact" comments. They are claim-level, figure-level, and journal-bar problems.

Free methodology feedback on non-sensitive manuscripts. For a PhD student checking whether their experimental design has obvious flaws before showing the paper to their advisor, free is hard to beat. If the manuscript doesn't contain anything confidential and you're OK with the data handling, Rigorous provides useful methodology comments at zero cost.

Academic credibility. ETH Zurich is a world-class research institution. The project reflects genuine interest in how AI can support peer review, and the open-source approach (MIT license) promotes transparency.

Exploratory use. For researchers curious about what AI peer review looks like, Rigorous is a low-stakes way to experiment, provided the manuscript isn't sensitive.

Upcoming features. Rigorous's GitHub shows Agent2_Outlet_Fit (journal/conference fit evaluation) in development. If shipped, this would address one of the current gaps. Worth watching, though it's not available yet.

Rigorous falls short when the manuscript is no longer an early methods draft and the next decision is submission. It does not verify citations, read figure panels, score a specific journal target, or provide a human expert escalation path. Those are the checks that matter when the risk is an avoidable desk rejection rather than a better-written methods paragraph.

Methodology-feedback false confidence. The methods critique is useful, but the manuscript can still fail because the Results section overclaims what the figures show or because the Discussion does not handle the obvious reviewer objection.

Free-tool privacy mismatch. A zero-cost review can be a good drafting aid for non-sensitive work. It is a poor default for unpublished clinical data, patentable methods, or industry-adjacent results when the terms disclose third-party storage and model processing.

Formal-review false confidence. Rigorous explicitly says its output is not formal peer review. Authors should not treat a positive AI report as evidence that the manuscript is ready for a selective journal.

Journal-fit blind spot. Agent2_Outlet_Fit is still described as in development on GitHub. Until journal/conference fit is live and verified, Rigorous should not be treated as a target-journal readiness tool.

When privacy matters. Any manuscript with unpublished data, proprietary methods, patentable material, or institutional compliance requirements should use a service with explicit privacy boundaries. Rigorous's current terms disclose Backblaze storage, OpenAI API processing, and research use of anonymized aggregated data. Manusights does not train models on manuscripts, bounds Manusights-side retention, and sends Anthropic-processed content under provider-side zero-retention terms.

When you need more than methodology feedback. Rigorous provides methodology, clarity, and impact comments. That's one layer of review. Manusights adds citation verification, figure analysis, journal-specific scoring, and a prioritized revision checklist. For a manuscript 4 weeks from submission, you need all of these.

When you need a production-grade assessment. Rigorous explicitly says its output is not formal peer review. Manusights delivers a structured diagnostic report designed to inform submission decisions with quantified risk scoring.

When the stakes are high. For career-defining submissions where judgment calls about novelty and positioning can determine acceptance, human expert review matters. Manusights provides a path to named field experts ($1,000+) and CNS editors ($1,500-$2,000) with full NDA protection. Rigorous is AI-only.

Choose Rigorous if:

• The manuscript is not privacy-sensitive
• You want free AI methodology feedback
• You're comfortable with OpenAI processing and Backblaze storage
• You don't need journal-specific scoring, citation verification, or figure analysis
• You're in the early drafting stage and want a second opinion on experimental design

Choose Manusights if:

• The manuscript contains unpublished or proprietary data
• You need citation verification, figure analysis, or journal-specific scoring
• You need a quantified readiness assessment for a selective journal
• You want the option to escalate to human expert review
• You need bounded operational retention and no model training on manuscript content

If your manuscript is not privacy-sensitive, try Rigorous for free methodology feedback. Then manuscript readiness check (about two to three minutes) for readiness scoring and journal-fit assessment. If you need citation verification, figure analysis, or a prioritized fix list, add the $39 Manusights diagnostic.

If your manuscript is privacy-sensitive, skip Rigorous entirely. Start with Manusights. The free scan and $39 diagnostic both use no-training manuscript handling, bounded Manusights-side retention, and provider-side zero-retention terms for Anthropic-processed content.

• manuscript readiness check

Rigorous product, privacy, and repository claims on this page reflect public information checked on 2026-06-24. AI-review projects can change quickly; verify current upload terms, data handling, and supported modules before uploading confidential unpublished work.