Scientific Writing with AI

The Writing Bottleneck

🔄 Quick Recall: In the previous lesson, you used AI to write analysis code, verify statistical logic, and create publication-quality visualizations. Now you’ll use those results to write the manuscript — with AI as your drafting and editing partner.

Most researchers can analyze data faster than they can write about it. The analysis takes a week; the manuscript takes months. AI doesn’t write your paper for you, but it can break through the blank-page paralysis and accelerate every stage from first draft to final revision.

Section-by-Section Drafting

Each manuscript section benefits from AI differently. Here’s how to prompt for each:

Abstract

Write an abstract for a [journal type] paper:

Title: [your title]
Background: [1-2 sentences on the problem]
Methods: [study design, sample, analysis]
Key results: [main findings with numbers]
Conclusion: [main takeaway]
Word limit: [journal's limit, typically 250-300]

Follow the structure: Background → Methods → Results → Conclusion

Why this works: Abstracts are the most formulaic section. AI excels at condensing information into structured formats. Your job: verify every number, ensure the conclusion matches your actual findings, and check the word count.

Introduction

Draft an Introduction for my paper on [topic]:

The problem: [what gap or question motivates this study]
Why it matters: [practical or theoretical significance]
What's known: [key findings from my literature review]
What's missing: [the specific gap my study addresses]
My approach: [how my study fills the gap]

Target journal: [journal name]
Target length: [word count, typically 500-800 words]
Tone: [formal academic / accessible / field-specific conventions]

✅ Quick Check: Why provide “what’s known” and “what’s missing” rather than letting AI generate the literature context? Because AI might cite papers that don’t exist, misrepresent findings, or omit the specific studies that motivate YOUR work. Your literature review (from Lesson 2) provides accurate context. AI structures it into a compelling narrative.

Methods

Provide maximum detail — AI needs specifics to produce reproducible descriptions:

Write the Methods section for my study:

Participants: [N, demographics, recruitment, inclusion/exclusion criteria]
Design: [experimental, longitudinal, cross-sectional, etc.]
Materials: [instruments, scales, equipment — with citations]
Procedure: [step-by-step protocol]
Analysis: [statistical tests, software, significance criteria]
Ethics: [IRB approval number, consent procedures]

Include subsections as appropriate for [target journal].

Results

Write the Results section based on these findings:

[Paste your statistical output or summarize key findings]

For each result, include:
1. Descriptive statistics (means, SDs, or frequencies)
2. Test statistic, degrees of freedom, p-value
3. Effect size with confidence interval
4. Brief interpretation in context

Follow APA reporting standards for [test type].
Do NOT interpret beyond what the data show.

Discussion

The Discussion is where your expertise matters most. AI can draft, but you must drive the interpretation:

Draft a Discussion section based on:

Main findings: [list your key results]
How they relate to existing literature: [agreement/disagreement with prior studies]
Theoretical implications: [what this means for theory in my field]
Practical implications: [real-world applications, if any]
Limitations: [be specific and honest]
Future directions: [what should be studied next]

Important: Distinguish clearly between what my data supports directly and what is speculative. Use appropriate hedging language.

Maintaining Your Scholarly Voice

AI produces competent academic prose. But competent isn’t the same as yours. Here’s how to keep your voice:

Step 1: Write one paragraph in your natural voice before using AI. This becomes your voice reference.

Step 2: Include voice instructions in your prompts:

Match this writing style: [paste your paragraph]

Specifically:
- Sentence length patterns: [short and direct / longer and nuanced / mixed]
- Vocabulary level: [technical for specialists / accessible for broader audience]
- Use of first person: [we / the authors / passive voice]
- Citation integration: [parenthetical / narrative]

Step 3: After AI drafts, read every sentence aloud. Rewrite any sentence that doesn’t sound like something you would write.

✅ Quick Check: Why is maintaining your voice important? Because reviewers notice voice inconsistency. If your Introduction reads differently from your Discussion, it signals that someone — or something — else wrote part of the paper. Consistent voice also reflects consistent thinking, which strengthens your argument.

AI for Editing and Revision

AI is often more useful for revision than for drafting:

Clarity check:

Review this paragraph for clarity. Identify:
- Sentences longer than 25 words that could be split
- Jargon that could be replaced with simpler terms
- Passive voice that could be made active
- Ambiguous pronouns or references

Logic check:

Read my argument from [section]. Does each paragraph's
conclusion logically follow from its evidence? Flag any
gaps in reasoning or unsupported claims.

Journal compliance:

Review this manuscript against [Journal Name]'s
author guidelines. Check: word count limits, heading
structure, reference format, figure requirements,
and required sections.

Non-Native English Speaker Advantages

AI particularly benefits researchers whose first language isn’t English. This is explicitly supported by most journals:

What AI does well:

• Correcting grammar and awkward phrasing
• Suggesting more precise vocabulary
• Improving sentence flow and transitions
• Adapting writing to journal-specific conventions

Effective prompt for language improvement:

Improve the English in this paragraph while preserving
the exact scientific meaning. Do not change the content,
claims, or level of certainty. Only improve grammar,
word choice, and readability.

[your paragraph]

Key Takeaways

• AI drafts each manuscript section differently — Abstracts benefit from condensation prompts, Methods from detail-rich prompts, Discussion from interpretation-guided prompts
• Maintain your scholarly voice by providing a writing sample as reference and reading every AI-generated sentence aloud
• AI excels at revision: clarity checks, logic verification, and journal compliance review
• For non-native English speakers, AI is an accepted and encouraged tool for language improvement — most journals require disclosure but support the practice
• The ethical line: AI helps you express your ideas clearly, not generate intellectual content

Up Next: You’ll tackle the critical question of AI ethics in research — reproducibility, disclosure requirements, and navigating journal policies that are still evolving.