Mentoring and editing of your own work
PhD Assistance describes its role as that of an external mentor who helps a scholar brainstorm an idea and translate it into a research model. For a manuscript, that means the support starts from your draft and your findings. A subject mentor comments on and edits what you have written; you remain the author.
This boundary is not a matter of preference. The ICMJE recommendations set four criteria for authorship, including final approval of the version to be published and agreement to be accountable for all aspects of the work, and list writing assistance, technical editing, language editing and proofreading among the contributions that do not on their own qualify for authorship (International Committee of Medical Journal Editors, 2025). Such help should be acknowledged rather than hidden.
Why one point of contact and a written record matter
Rules on third-party help differ between institutions. Richards and Harwood (2024) note that some university proofreading policies are poorly worded and ambiguous, and that staff and students are often unclear about them. They describe a range of interventions, from the lightest touch, such as correcting typos, to the heaviest, such as wholesale rewriting of content, and argue that academics should decide which of these they allow.
Routing every request and every supervisor change through one person, by email, serves that purpose. You can show your supervisor exactly what was asked for and what was done, and your supervisor's limits can be passed on to your mentor before editing begins rather than after.
What a similarity report can and cannot tell you
Each document goes through a similarity check before delivery, and you receive the report. It is useful evidence, but it needs to be read with care. In a test of 15 text-matching systems across eight languages, Foltýnek et al. (2020) found that some systems help identify some plagiarised content, but none found all of it, and at times they flagged material that was not plagiarised. The systems did best on verbatim copying and poorly on paraphrased or translated text.
Two consequences follow. A low score is not proof of originality, and a high score is not proof of misconduct: matches may be correctly quoted and cited passages, standard method descriptions or your own earlier work. That is why the report goes to you with the manuscript, so you and your supervisor can look at what was matched and why.
AI tools and who stays responsible
Publication-ethics bodies are clear that responsibility for a manuscript cannot be handed to software. COPE states that AI tools cannot be listed as authors because they cannot take responsibility for the submitted work, that authors must disclose how and which AI tools were used, and that authors are fully responsible for the content, including any parts produced by an AI tool (Committee on Publication Ethics, 2023). The ICMJE recommendations likewise ask that AI use for writing assistance be reported in the acknowledgements (International Committee of Medical Journal Editors, 2025).
Our published process describes a similarity check, not an AI-detection step, and readers should be cautious about tools that claim to detect machine-generated text. In a test of such detectors, Weber-Wulff et al. (2023) concluded that the available tools were neither accurate nor reliable, tended to classify text as human-written, and performed worse when content was obfuscated. Honest disclosure by the author, not a detector score, is what the guidance asks for.
