Research integrity and responsible AI

Credible work makes its reasoning traceable.

Practical principles for using evidence, sources and AI tools without surrendering accuracy, authorship or accountability.

Ten operating principles

A baseline for trustworthy research and writing.

Adapt these principles to the discipline, study design, audience and rules governing each project.

  1. 01

    Keep the question visible

    Let the research question—not a convenient dataset or preferred answer—drive the design and interpretation.

  2. 02

    Record the method

    Document searches, inclusion decisions, instruments, transformations and analytical choices well enough to audit the work.

  3. 03

    Represent sources fairly

    Read the original source, preserve its context and cite it for the exact claim it supports.

  4. 04

    Separate findings from inference

    Distinguish what the data show from interpretation, explanation and recommendation.

  5. 05

    Report uncertainty

    Describe limitations, missing data, alternative explanations and the boundaries of generalization.

  6. 06

    Respect authorship

    Credit substantive contributions and do not assign authorship to people or tools that cannot take responsibility for the work.

  7. 07

    Use AI transparently

    Follow relevant disclosure rules, verify every output and never invent or rely on unchecked citations.

  8. 08

    Protect confidential material

    Do not submit private data, unpublished manuscripts or restricted material to an unapproved system.

  9. 09

    Correct the record

    Respond openly to errors and make material corrections visible to readers.

  10. 10

    Follow the applicable standard

    Use the reporting, ethics, citation and institutional requirements that govern the specific work.

Integrity checks

Pause before publishing when…

  • A central claim is not supported by a source you have checked.
  • The method cannot be reconstructed from the available record.
  • Results and interpretation are blended in a way that hides uncertainty.
  • Confidential material has entered an unapproved system.
  • AI contributions, conflicts or material limitations are undisclosed.
Continue with primary guidance

Independent, non-affiliate sources

COPE publication ethics guidance ↗EQUATOR reporting guidelines ↗Purdue OWL research and citation guidance ↗Browse the full resource library →