A Verifiable AI-Assisted Research Workflow: How Prof. Jefferson Moreno Guaicha Ran His Scoping Review

In short: Prof. Jefferson A. Moreno Guaicha used SciSpace throughout a scoping review on research integrity and AI ethics. He built the evidence base himself from Scopus, Web of Science and Dimensions, used AI to compare papers and draft, checked every reference twice, and rewrote the manuscript in his own voice. His rule: use AI only where you understand the subject well enough to challenge its output.

Who is Prof. Jefferson Moreno Guaicha?

Dr. Jefferson A. Moreno Guaicha is a Research Professor in the Department of Human Development at Yachay Experimental Technology Research University (Yachay Tech) in Urcuquí, Ecuador. He works on philosophy of education and artificial intelligence, teaches scientific research methodology, and has taught Ethics and Regulation of AI in Yachay Tech's Master's in Artificial Intelligence. His earlier work includes a systematic review of AI tools in teaching strategies (Revista Andina de Educación, 2025).

He recently used SciSpace on a scoping review on research integrity and AI ethics, from organising the literature to answering peer reviewers. Profiles: ORCID · Google Scholar · ResearchGate.

For him, AI is useful only with a clear division of responsibility. The researcher defines the question, understands the subject, chooses the methodology, decides which evidence belongs in the study and judges whether the conclusions hold. AI supports the scale and structure of that work. It can't supply the expertise needed to defend it.

AI helps process and structure the research; the researcher stays responsible for understanding, verifying and defending it. The result wasn't an automated review. Every AI-assisted step was followed by a decision, check or revision of his own.

1. Start with a research question and a defined body of evidence

His workflow doesn't begin with an open-ended instruction to find and write everything about a topic. He starts with his own research question and searches established academic databases: Scopus, Web of Science and Dimensions. He selects the relevant records, exports them and uploads that literature to SciSpace.

“I prefer to have my own research databases using high-quality academic sources, and then upload that curated corpus to SciSpace. From there, I can compare studies, extract and structure information. This gives me greater control over the process and helps strengthen its transparency, traceability, and reproducibility.”

This keeps the review grounded in a corpus he chose, rather than letting the AI decide the evidence base. Within that collection, SciSpace helps him compare papers, extract relevant information, spot patterns and organise the state of the art. Deciding which studies belong in the review stays with him.

It's also the order he teaches his methodology students: first learn to search academic databases, then use SciSpace to extract and compare evidence from the literature you gathered. If you're new to the method, see our systematic literature review agent and the guides linked at the end.

2. Move from the evidence to a working draft

He only starts writing once three things are in place: the research design and methodological criteria, the selected literature uploaded to SciSpace, and his own ideas and interpretations developed during the research.

He then uses the SciSpace Agent to produce a working draft. His prompt gives structured instructions on the manuscript's organisation, the methodological criteria, the evidence, and the ideas and interpretations he wants developed. The AI structures his thinking rather than replacing it.

“I like the first draft to be based on my information, my ideas, and the literature I have already selected.”

He doesn't treat that draft as a finished manuscript. He moves it to AI Writer to review it point by point, refine his ideas and verify any additional information or literature introduced during drafting. Then he edits the text so its reasoning, tone and language are his own.

3. Verify the bibliography twice

Source verification isn't a final formality in his workflow; it's built into the process. His rule for students is simple: “They need to do a double check.”

The first check combines his own ongoing review of references and claims with SciSpace's automated bibliography audit (the same check the Find, Add & Audit Citation agent runs). The audit separates references it could access and review from papers that need more attention.

The second check is manual. Some sources sit behind paywalls or need a journal or database subscription. When the audit reports that a paper couldn't be accessed, he doesn't count the reference as verified. He finds the original paper himself and reads it. His process:

1. Review the references and supporting claims, and run the bibliography audit.
2. Check which references were verified and which were inaccessible.
3. Open the unresolved papers independently.
4. Confirm the bibliographic details (authors, year, journal, DOI) and the claim each source is being used to support.

The audit organises the work. The researcher still decides whether the evidence is valid.

4. Return the manuscript to the researcher's voice

After checking the evidence, he goes back to the writing and revises the AI-assisted material in his own words and from his own point of view. He uses SciSpace's AI Detector as an editing checkpoint.

For him, the detector isn't proof of who wrote a manuscript. It's a prompt to revisit passages that still sound generic, predictable or shaped by AI structure. The aim isn't to disguise AI use; it's to make sure the final structure, reasoning and language reflect the researcher's understanding.

“You start with your idea, you use AI to support your ideas, and then... you are working again and editing all the process to have a result that is completely yours.”

He kept using the same workflow when the scoping review reached peer review. Rather than treating submission as the end, he went back to the evidence, the manuscript and the verification steps while addressing the reviewers' comments.

The boundary: use AI only where you can evaluate it

He believes AI can support many stages of research, with one firm condition: the researcher must understand the subject well enough to challenge the output. That means knowing the relevant theories, methodology and research process before asking AI to interpret evidence.

“If it is not your own idea, and if you don't understand the topic and you are taking that information as true, please don't use AI.”

This matters most when discussing results:

“If we don't understand what we get in the results and we are trying to discuss that information just with AI, we are not doing research.”

AI can organise evidence and suggest connections. It can't decide whether an interpretation is scientifically or conceptually sound when the researcher lacks the knowledge to assess it.

Prof. Moreno Guaicha's workflow at a glance

1. Define a research question you understand.
2. Search academic databases (he uses Scopus, Web of Science and Dimensions) and select the relevant literature.
3. Upload that researcher-selected corpus to SciSpace.
4. Compare papers, extract evidence and organise the state of the art.
5. Generate a working draft with the SciSpace Agent from your own design, ideas and bibliography.
6. Review the draft point by point in AI Writer, and verify every source added during drafting.
7. Run a bibliography audit.
8. Manually check inaccessible or uncertain papers, including their bibliographic details and the claims they support.
9. Rewrite the manuscript in your own voice, using the AI Detector to find passages to revisit.
10. Return to the evidence when addressing reviewer feedback.

The value isn't that AI finishes the research faster on its own. It's that AI helps the researcher work through the evidence, while the researcher stays responsible for what enters the study, what supports the argument and what can ultimately be defended.

Build a verifiable research workflow with SciSpace

Work from your own database searches, compare evidence, audit references and revise your manuscript, while keeping source verification in your hands.

Frequently asked questions

Related guides

注意转载授权 · 阅读原文