Practical writing on AI manuscript review, research integrity, and getting work ready to submit.
AI manuscript review examines a paper for integrity and quality across three territories, then returns findings as evidence a person judges. How it works.
ChatGPT helps with language, not integrity: fabricated citations, no evidence trail, real confidentiality risk. What grounded manuscript review does instead.
Language models invent citations that look real, and they survive review. How hallucinated references reach published papers, and how to catch yours first.
Retraction happens after citation, so a valid reference can be withdrawn before you submit. What citing retracted papers costs and how to screen for it.
Roughly 20 to 35% of manuscripts are flagged for image issues, yet acceptance is rescinded for only 1 to 8%. What image screening actually finds.
Generative tools produce figures experts cannot reliably spot by eye. What AI-generated image detection can honestly promise, and what publishers do next.
Figure reuse across different papers is invisible to manual review because no one holds the literature in their head. How image plagiarism screening works.
The issues that cause desk rejections are mostly catchable before you submit. A practical pre-submission manuscript checklist, dimension by dimension.
Desk screening decides whether a paper reaches a reviewer at all. What journals check before peer review, and what happens when screening flags something.
n = 28 · whole-number responses
Reported means can be mathematically impossible given the sample size. How the GRIM test works, what an inconsistency signals, and why it is not proof.
Automated integrity checks catch pattern classes well and cannot judge intent or merit. An honest account of the limits, and what they mean for tooling.
More than 70% of surveyed researchers failed to reproduce another lab's work. Where irreproducibility comes from and the completeness lever that helps.
ICMJE requires disclosing AI-assisted technologies and bars chatbots as authors. What to disclose, where to put it, and what undisclosed use costs.
Funders expect applications to address rigor explicitly: design, power, authentication, reporting. What NIH and others require, and where sections fail.
Contact us, or log in.