SciSpace Academy
Tutorial

Using AI in Scientific Writing: Rigor, Evaluation and Integrity

Dr. Ethan James GantanaEnglishAcademic writing

TL;DR

1-min read

This tutorial is for researchers, reviewers and academic authors, particularly in the health sciences, who use or evaluate AI-assisted writing. It is led by Dr. Ethan James Gantana, a specialist haematopathologist at the National Health Laboratory Service in Cape Town and Senior Lecturer at Stellenbosch University. He explains how SciSpace can support scientific writing workflows and, just as importantly, how to judge what AI tools produce before it reaches a journal.

Key takeaways

  • SciSpace can support several stages of a scientific writing workflow, from preparing material to refining drafts.
  • AI-generated text has real limits, so its reliability needs to be assessed rather than assumed.
  • A practical check is to compare every factual statement and statistic with the original sources and to watch for vague or overgeneralised wording.
  • Citation integrity means each reference is real, accurately described and genuinely supports the point it is attached to, and AI tools can produce plausible but non-existent or mismatched citations.
  • Methodological rigor has to be maintained whenever AI contributes to a manuscript, and a co-author or colleague review of key sections adds another safeguard.
  • Best practices are offered for authors and reviewers who want the benefits of AI assistance while staying fully accountable for their scientific work.
Apply it

To apply this, treat any SciSpace-assisted draft as unverified, confirm each reference in the original paper before submission, and keep responsibility for every claim in your manuscript.

How to do it

  1. Map SciSpace to your writing stages

    Identify which stages of your scientific writing workflow SciSpace's AI Writer can support. This shows where AI in scientific writing genuinely helps.

  2. Evaluate AI-generated text critically

    Assess each AI-generated passage for reliability and limitations before keeping it. AI output needs careful evaluation before it reaches a journal.

  3. Protect methodological rigor

    Check that any AI contribution to methods or results is accurate and justified. Rigor must not slip because text was drafted quickly.

  4. Verify every citation

    Use the citation generator, then confirm each reference exists and supports the claim. This avoids unsupported or fabricated references.

  5. Apply responsible-use practices

    Follow best practices for responsible AI use, whether you are the author or the reviewer. This protects the integrity of the manuscript.

Watch the tutorial33 min

Using AI in Scientific Writing Without Fake CitationsPlay video: Using AI in Scientific Writing Without Fake Citations

Chapters

  1. Welcome and speaker introduction
  2. Can AI-generated scientific content be trusted?
  3. Can AI be an author on a research paper? The ICMJE rule
  4. Can AI write a review article? Testing ChatGPT, DeepSeek, Jenny AI and SciSpace
  5. What is the SANRA scale for narrative review quality?
  6. AI vs human review results: ChatGPT's unverifiable references
  7. Why does AI make up fake citations? AI hallucination explained
  8. Real references, wrong claims: how AI misattributes citations
  9. Can ChatGPT write a research paper? What the study found
  10. Will AI replace review articles? Peer review and AI disclosure policies
  11. How to use SciSpace for literature review as a starting point
  12. Can AI do a systematic review? SciSpace agent on a real protocol
  13. AI title and abstract screening and data extraction in SciSpace
  14. Fact-checking a 69-page AI-generated systematic review
  15. What is lazy (secondary) referencing? Citation errors in AI output
  16. Can an AI-generated review count as a systematic review? The full-text problem
  17. How to keep AI-assisted research rigorous: three safeguards
  18. How to use AI in scientific writing responsibly

Try this on SciSpace

Frequently asked questions

How can I check whether AI-generated scientific text is reliable?

Compare every factual statement and statistic with the original sources, confirm that cited references exist and actually support the claim, and look for vague or overgeneralised wording. Ask a co-author or colleague to review key sections. AI text should be treated as a draft that requires the same scrutiny as any unverified source.

What is citation integrity?

Citation integrity means each reference is real, accurately described and genuinely supports the point it is attached to. AI tools can produce plausible but non-existent or mismatched citations, so authors should verify every reference against the original paper before submission. Accurate citation also gives proper credit to other researchers.

Related tutorials