Artificial intelligence can help at almost every stage of scientific research — from idea formulation to final text editing. But the scientific community has set clear boundaries here: AI can be a powerful assistant but never an author or responsible party. In this guide we look at how to properly apply AI in research and which line must not be crossed.
Stage 1: Literature search and analysis
The most time-consuming part of research is studying existing literature. AI plays three roles here: search, summarization, and comparison.
Ready prompt (for a literature map):
I am researching the topic "Water resource management in Central Asia."
Break the topic into 5 small directions and for each direction give:
1) Key terms (in English and Russian), 2) 3 scientific databases for searching
(what query to write in Scopus, Web of Science, Google Scholar),
3) A list of expected key questions for each direction.
Found article abstracts can be uploaded to the AI and summarized:
Read the abstracts of the following 8 articles and build a table: columns —
author/year, method, main conclusion, relevance to my topic (high/medium/low).
At the end: are there contradictory conclusions among these articles — if so,
which ones?
[Abstracts text]
Important limitation: AI can "read" full article texts and draw conclusions, but it can also invent sources — citing a nonexistent article ("hallucination") is the most dangerous error in scientific writing. Verify every citation in the original source.
Stage 2: Research design and methodology
AI acts as an "interlocutor" in refining the research question and choosing a method.
Ready prompt:
My research question: "The impact of remote work on employee productivity
in Tashkent IT companies." Break this question into 3 measurable sub-questions.
For each, suggest a suitable method (survey/interview/observation) and justify
the sample size. Critically point out 2 weak spots in the methodology.
The last sentence matters — asking AI to criticize its own proposal prevents blind trust. AI also helps compose survey questions, but testing questions on real respondents (pilot) is an irreplaceable stage.
Stage 3: Data analysis
In statistical analysis, AI helps write code (Python/R) and interpret results — but responsibility for the analysis's correctness rests fully with the researcher.
Ready prompt:
Write Python code for the following analysis: survey data of 200 respondents
(CSV: age, work_experience, remote_days, productivity_score).
1) Data cleaning and descriptive statistics, 2) correlation between
remote_days and productivity_score, 3) a 3-sentence conclusion interpreting
the result. Write a comment for each step in the code.
Golden rule: don't use AI-written analysis code without understanding it. You must know what each line does — at a defense or review, the answer "AI did it like that" is not accepted.
Stage 4: Scientific writing — allowed and forbidden boundaries
This is the most delicate topic. The position of major publishers and ethics committees is as follows:
Allowed (disclosure usually not required): fixing grammar and spelling, improving style, translation — i.e., polishing text you wrote.
Allowed (disclosure required): help formulating ideas, generating text parts, literature search — these must be shown in the article's "Methods" or "Acknowledgements" section. For example, Springer Nature journals require documenting LLM use in the Methods section.
Forbidden: listing AI as an author (COPE and ICMJE are unanimous: AI cannot take responsibility, so it cannot be an author); uploading a confidential manuscript to AI during peer review; creating fake data or sources.
Ready prompt (for lawful editing):
Edit the English of the following scientific text: improve grammar, academic
style, logical coherence. Don't change scientific terms or content. Give a
separate list of the places you changed.
[Text]
"For every sentence written with AI's help, responsibility remains with the author. A journal editor won't accept the excuse 'AI wrote it' — the error counts as yours. So check AI text even more strictly than your own."
Stage 5: Ethics checklist — before publication
Before sending the manuscript to a journal, answer these questions:
- Did I disclose? Did I read the target journal's AI policy and show the usage in the required section?
- Is authorship correct? Are only people who made a real contribution listed? Is AI absent from the author list?
- Are sources real? Is every citation verified in the original source? Are there no AI-"invented" sources?
- Is data protected? Were research participants' personal data not uploaded to the AI service?
- Was review confidentiality breached? When reviewing others' manuscripts, did I not upload them to AI? (ICMJE considers this a confidentiality breach.)
- Are facts verified? Is every AI-generated claim confirmed by an independent source?
Additional point in the Uzbek context: the question of AI use in dissertations under HAC (Higher Attestation Commission) requirements is not yet fully regulated — agree in advance with your supervisor and council.
Which tool at which stage: an AI map for the researcher
Beyond general chatbots (ChatGPT, Claude, Gemini), there are special tools for specific research stages:
Literature search: Elicit and Consensus — find evidence from scientific articles for your question; Scite.ai — shows how an article was later cited (supported or criticized); ResearchRabbit and Connected Papers — starting from one article, draw a "map" of related works.
Data analysis: ChatGPT's code-writing ability (Python/R) or Claude's large-table analysis function — for statistical calculations and visualization.
Writing and editing: Grammarly or Paperpal — for academic English; DeepL Write — for rephrasing sentences.
Plagiarism and originality: Turnitin, iThenticate — standard tools for checking text before submission.
Important warning: these tools can also err. For example, literature-search tools sometimes "find" a nonexistent article. Checking every find in the original database (Scopus, Web of Science, Google Scholar) is an irreplaceable habit.
Extra advice for Uzbek researchers: if access to international databases is limited, use access provided through your university library — many universities have institutional subscriptions to Scopus and Springer.
Grant applications and scientific presentations
Beyond the article, a researcher has two more important written works: grant applications and conference presentations.
For a grant application: AI is strong at checking an application's structure. Ready prompt:
Evaluate the following grant application draft (criteria: problem statement
clarity, methodology soundness, budget realism, measurability of expected
results). Score 1–5 on each criterion and concretely suggest how to improve
the 2 weakest spots.
[Application text]
For a presentation: turning a 15-minute speech text into a slide plan:
From the following talk text, build a plan for 10 slides: for each slide
a title, 3 key points, a description of the shown chart/table.
Mark places where complex jargon should be simplified.
[Text]
Reminder: grant organizations have also started having AI policies — if AI was used in the application, disclose as the instructions require. Some foundations don't accept fully AI-generated applications.
Practical conclusion
Do today: (1) Identify which stage of your research you're at and try one of the above prompts suited to that stage. (2) Find and read your target journal's AI policy page — it takes 10 minutes but will save you from big problems in the future. (3) Open an "AI usage log": record which tool, when, for what purpose you used — when submitting the article, writing the disclosure section will be a 5-minute job. Remember: AI accelerates research, but scientific integrity is solely your responsibility.
Agreement with your supervisor
Hiding AI use from your supervisor is the worst strategy. Instead, talk openly at the start of the research:
- Say at which stages you want to use AI (literature search, language editing, code writing).
- Ask about the supervisor's boundaries — some oppose text generation but agree to editing.
- Record the agreement in writing (at least via email).
This conversation takes 15 minutes but lets you confidently answer "did you use AI?" at the defense. In many Uzbek universities this question is still at the discussion stage — your openness can be an example for the department too.
Final reminder: AI tools update every month — a method correct today may be outdated tomorrow. Recheck journal policies before each submission and exchange experience with colleagues in your field. Scientific integrity is an unchanging value; the ways of applying it update with the times.
