AI-powered research integrity

Scientific Integrity,
Automated.

Detect conflicts of interest, funding bias, and predatory practices in scientific literature with our advanced AI-powered analyzer.

How it Works
01 / MANIFESTO

An act of scientific activism.

The proliferation of predatory journals and undisclosed conflicts of interest has turned academic publishing into a market of simulated legitimacy. RIA is the counterforce.

Industry funds science. Sometimes too much.

Sponsor-funded papers are systemically biased toward results that favor the sponsor. We surface that asymmetry, line by line.

4.5× more likely to favor sponsors

15,500+ predatory journals. Tracked.

Hijacked lookalikes, pay-to-publish mills, fake indexing — every venue checked against an offline database that grows monthly.

15,500+ venues catalogued

"Truth as a public good."

Every detected conflict of interest becomes a step toward reclaiming truth from corporate, political, and ideological capture.

1.6 s average audit time
00  /  ANATOMY OF A PAPER

First, the paper becomes evidence.

RIA reads the PDF structure, extracts claims, affiliations, citations, funding statements, and every disclosure line before a score is ever shown.

01  /  AUTHORSHIP

Who really wrote this.

Affiliations, ghost authors, and undisclosed roles. We cross-check every byline against the paper's funding and prior work.

2 conflicts surfaced
02  /  FUNDING

Follow the money.

Industry grants, sponsorships, and silent backers — flagged when their interests align with the paper's conclusions.

$1.2M sponsor disclosed
03  /  METHODOLOGY

Does the science hold up?

Sample size, statistical power, p-hacking signals, predatory venue indicators — audited against five integrity dimensions.

3 weak-evidence flags
04  /  JOURNAL SIGNALS

The venue is part of the evidence.

We compare journal metadata, indexing traces, publisher history, and predatory-pattern signals before trusting the publication context.

Venue risk elevated
05  /  OUTSIDE THE PDF

The paper is checked against the world around it.

External verification looks for retractions, watchdog mentions, publisher anomalies, and independent records that confirm or challenge the document.

6 external traces matched
06  /  VERDICT

One score. Full evidence.

Every flag is cited back to the source line. Export the report as PDF, share it with reviewers, or revisit the audit anytime.

Report generated · 32 / 100
A Randomized Trial of Compound X for Chronic Inflammation
Authorship · Conflict
Funding · Sponsor
Methodology · Weak power
Venue · Publisher signal
External · Watchlist trace
/ HOW IT WORKS

Three stages.
One verdict.

PDF Input
AI Audit
Risk Score
32/100
High risk · 4 flags · 1.6s
/ EVIDENCE TRAIL

Not a black box.
A trail you can inspect.

As the score forms, RIA keeps the reasoning visible: source line, risk rule, external signal, and reviewer action. The story ends in a report, not a mystery number.

01 · Source line “Funding provided by Compound X Therapeutics.”
02 · Integrity rule Financial interest aligns with positive conclusion.
03 · External check Publisher and sponsor records matched.
04 · Reviewer note Needs disclosure review before citation.
32/100

Exportable integrity report with cited evidence.

03 / THE LIVING DATABASE

Static blacklists are obsolete.

We start with historical knowledge — Beall's list, watchlists, ISSN registries — and feed it through automated scraping into a single normalized source of truth.

Phase 1 · Foundation

Two layers, one verdict.

Every uploaded paper is checked against the internal database first. If the venue is unknown, an LLM-driven web scan looks for behavioural red flags in real time — fake peer review, hidden fees, hijacked indexing.

Phase 2 · Detection

Two human verifiers must agree.

When the AI flags a discrepancy, expert verifiers — researchers, librarians, meta-scientists — review the evidence. A journal joins the database only after two independent validations.

Phase 3 · Human-in-the-loop

A database that improves itself.

Validated journals are absorbed back into the core. Every analysis and every verification grows the system stronger for the entire community — a living database, evolving as fast as predatory publishers do.

Phase 4 · Closed loop
04 / METHODOLOGY

How RIA actually audits a paper.

Every analysis runs through the same five-stage pipeline. Deterministic rules where rules are robust; AI semantic extraction where context matters; everything cited back to the source line.

01 · Ingestion & Preprocessing

PDF parsing, OCR fallback, sectioning, citation-graph extraction, author-affiliation linking. The paper becomes structured evidence.

~150 ms · per paper

02 · Semantic Extraction

LLM passes pull funding statements, declared conflicts, methods, sample sizes, and statistical claims — converted into machine-checkable facts.

~700 ms · LLM passes

! · Predatory Journal Module

Cross-checks the venue against a 15.5k+ database of known predatory and questionable publishers, including hijacked journal lookalikes.

offline DB · 15,500+ venues

03 · Rule Application

Five integrity dimensions evaluated against deterministic rule sets — authorship, funding, methodology, conclusions, venue. Every rule cites the line that triggered it.

deterministic · 5 dimensions

04 · Score & Report

Weighted aggregation across dimensions yields a 0–100 score. Output: gauge, dimension breakdown, evidence trail, exportable PDF.

~50 ms · final aggregation
04 / COMMITMENTS

Ethical AI.

AI is a tool, not an oracle. RIA surfaces what a human reviewer would otherwise miss — and cites the line that triggered every flag.

Open Methodology.

Every rule, threshold, and database is reviewable. Transparency is not a feature — it is the prerequisite for trust.

Public Good.

Knowledge is a commons. The tool is free. Donations cover infrastructure; no commercial gates, ever.

Built by one engineer.

RIA is the product of a personal commitment to scientific integrity. Open to contributors, reviewers, and translators.

05 / SUPPORT

Servers run on donations.

Every analysis ingests a PDF, runs OCR, queries a 15.5k-venue database, and serves the report. Hosting costs about $50 a month — and grows with usage.

AI is the heaviest line.

Semantic extraction passes run on commercial LLM APIs. ~$100/mo covers roughly a thousand analyses. Every euro past that goes to expanding capacity.

Anything above keeps growing the rules.

New detection rules. New domain experts on the team. Expanded predatory- journal coverage. Translation. Open-source code refactors. Donors fund the roadmap.

Free forever. Donor-funded.

100% non-profit. No ads, no upsell, no enterprise tier. Pick a way to help.

Donate via PayPal
01 / DEMO

See it in action.

Watch how the Research Integrity Project helps you verify scientific literature in seconds.

02 / CAPABILITIES

Built by researchers,
for researchers.

Five dimensions of scientific integrity, audited in seconds. Evidence-cited, peer-reviewable, exportable.

Deep Analysis

Scans for hidden funding, undisclosed affiliations, and textual bias using 5 dimensions of integrity. Our AI goes beyond simple keyword matching to understand context.

Instant Results

Get a comprehensive risk score and detailed report in seconds.

Privacy First

Your data is secure. We believe in transparency and reproducibility.

Comprehensive Reporting

Download detailed PDF reports with evidence citations, perfect for peer review or institutional auditing.

READY?

Start auditing.

Free forever. No signup required to analyze your first paper.

How it works
1.6 s end-to-end audit
15,500+ predatory venues
Evidence cited to source line