Our Methodology

How Paper Integrity analyzes research papers — and why you can trust the results.

How Analysis Works

Paper Integrity uses a multi-stage AI pipeline to evaluate research papers across six dimensions of scientific rigor:

  1. Content Extraction — We retrieve the full text, metadata, and supplementary data from the paper using specialized academic APIs and direct publisher access.
  2. Structural Analysis — Our AI evaluates study design, methodology, sample sizes, statistical approaches, and control groups against established standards for the paper's field.
  3. Conflict Detection — We cross-reference author affiliations, funding sources, and disclosure statements to identify potential conflicts of interest.
  4. Retraction & Correction Check — Every paper is checked against retraction databases and correction notices.
  5. Citation Context — We analyze how the paper has been received by the scientific community, including replication attempts and critical commentary.
  6. Risk Assessment — All findings are synthesized into an overall Citation Risk score with specific, evidence-based flags.

Six Dimensions of Integrity

Paper Status

Publication status, retraction history, corrections, and errata. Is this paper still considered valid by the publishing journal?

Study Design

Appropriateness of methodology for the research question. Sample sizes, controls, blinding, randomization, and measurement validity.

Analysis Integrity

Statistical methods, p-value reporting, effect sizes, multiple comparisons corrections, and data availability.

Transparency

Pre-registration, data sharing, code availability, methodology detail, and reproducibility commitments.

Independence & Conflicts

Funding sources, author affiliations, industry ties, disclosure completeness, and sponsor involvement in study design.

External Validation

Independent replications, meta-analyses, citation sentiment, expert commentary, and field consensus.

Our AI Stack

Paper Integrity is powered by Claude (Anthropic) for analysis and Perplexity Sonar Pro for real-time content retrieval. We chose these models specifically for their accuracy, reasoning depth, and resistance to hallucination.

Unlike generic AI summarizers, our analysis pipeline uses structured prompts developed with input from researchers and science journalists. Each report follows a consistent evaluation framework — not a freeform summary.

We do not train on your queries or analyzed papers. Your research stays private.

What We Can't Do (Yet)

Intellectual honesty is core to our mission. Here's what Paper Integrity does not claim to do:

  • We cannot detect sophisticated data fabrication that passes statistical tests
  • We cannot access papers behind paywalls unless you upload the PDF
  • We cannot replace expert domain knowledge in highly specialized fields
  • Our AI can make mistakes — always verify critical claims against primary sources
  • Citation risk scores are assessments, not verdicts. Use them as one input in your editorial judgment

Validation

We continuously test Paper Integrity against known problematic papers, including:

  • Papers that have been formally retracted
  • Papers flagged by Retraction Watch and PubPeer
  • Papers involved in well-documented scientific controversies
  • Clean, well-conducted studies (to verify we don't over-flag)

Our sample reports page showcases analyses of both problematic and exemplary papers — from the retracted Wakefield MMR study to well-conducted clinical trials.

Who We Are

Paper Integrity is built by Sage AI LLC. We're a small team obsessed with making research verification accessible to anyone who cites scientific papers — journalists, researchers, policy analysts, and the public. Questions about our methodology? Contact us at support@paperintegrity.com.