Using AI as Your Research Assistant for Literature Reviews

The AI Research Assistant Prompt

Here's the single most powerful prompt template you can use with any AI assistant to run a doctoral-level literature review:

You are an expert research librarian and PhD-level academic investigator
with agentic browser capabilities. Your task is to conduct a
comprehensive, doctoral-level literature review on [TOPIC] by actively
searching and retrieving sources from authoritative academic databases.

Research Databases to Search:
Scopus, Web of Science, ProQuest, JSTOR, SpringerLink, ScienceDirect,
Google Scholar, and specialized databases like PhilPapers or ATLA
Religion Database.

Research Objectives:
- Map the intellectual landscape of [TOPIC]
- Identify seminal works and recent high-impact publications
- Synthesize methodological approaches
- Surface research gaps that could anchor an original thesis argument

Conduct the Following:
1. Active Search — Query each database with Boolean search strings
2. Synthesis — Organize findings into thematic clusters
3. Critical Evaluation — Note contribution, methodology, citation impact
4. Hyperlinked Reference List — APA 7th edition, organized by theme

Your 5-Step Core Research Workflow

Step 1: Define Your Search Scope

Before opening any tool, write down your topic in one clear sentence, then break it into 2–4 key concepts. For example:

• Topic: "The impact of remote work on employee mental health in sub-Saharan Africa"
• Concepts: remote work / employee mental health / sub-Saharan Africa

Each concept becomes a keyword cluster you will combine with Boolean operators. The clearer your concepts, the better your searches — and the better your AI assistant's output.

Step 2: Run Broad Discovery with Perplexity

Start with Perplexity (perplexity.ai) — free to use — because it searches the live web AND academic sources, then answers with numbered citations. Ask it to map the field:

Map the current research landscape on [TOPIC]. Identify:
1. The 5 most-cited seminal papers
2. 3 recent (last 3 years) high-impact studies
3. The main theoretical frameworks used
4. Any systematic reviews or meta-analyses
Cite every claim with a source URL.

Then use Brave Search (search.brave.com) as a second opinion — it returns pure, ad-free results and is excellent for finding PDFs, preprints, and reports that Google buries.

Step 3: Deepen with Google Scholar

Google Scholar (scholar.google.com) remains the backbone of any literature review:

• Search each keyword cluster with AND combinations
• Click "Cited by" on any key paper to walk forward in the citation chain
• Click "Related articles" to find siblings you didn't know existed
• Use the date filter (left sidebar) to limit to the last 3–5 years
• Create an alert for your topic so new papers email you automatically

Step 4: Read Faster with NotebookLM

Upload your shortlisted PDFs into NotebookLM (notebooklm.google.com). It creates a private, source-grounded notebook where you can:

• Ask questions that are answered only from YOUR sources (no hallucinations)
• Generate summaries, timelines, and study guides
• Find connections across papers automatically
• Export citations in your chosen format

Step 5: Track Everything

Create a simple tracking table before you start — a spreadsheet or Notion page with these columns:

| Source     | Theme       | Method       | Key finding                 | Citation (APA) | Notes        |
|------------|-------------|--------------|-----------------------------|----------------|--------------|
| Author2024 | Measurement | Longitudinal | Effect size d = 0.42...     | ...            | Include in §3 |

Record which databases you searched, which terms you used, and how many results you got. This "search diary" is what makes your methodology chapter defensible at viva.

Your AI Research Assistant: Choose & Configure

Any of these free AI assistants can run the whole workflow above — pick the one you already use:

• DeepSeek — chat.deepseek.com (free, strong for long documents)
• ChatGPT — chatgpt.com
• Grok — grok.com
• Gemini — gemini.google.com/app
• MiniMax — chat.minimax.io

Activate Web Search & Think Longer

Before researching, turn on the assistant's web search toggle (so it pulls live sources instead of training memory) and, where available, enable "think longer" / deep-thinking mode for research questions. This single step is what separates a real literature search from a hallucinated summary.

The Thesis Structure Prompt

When you're structuring a thesis or dissertation, paste your university's table-of-contents template (or a rough outline) into the chat with this prompt:

You are my thesis advisor. I am about to research [TOPIC].
Copy and paste the table of contents structure we will follow and
turn it into a complete research plan:
1. Reorganize the sections in a logical order
2. Under each chapter, list the sub-sections I will need
3. Note which chapters map to which part of my research method
4. Flag any chapter that will need the most sources

Boolean Search Mastery

Boolean operators are the difference between 50,000 irrelevant results and 50 perfect ones. Use these variations when searching databases directly:

• "[TOPIC]" AND "literature review" AND methodology — narrows to methodology-focused work
• [KEYWORD1] AND [KEYWORD2] AND "empirical study" — finds only empirical evidence
• [TOPIC] AND framework AND theoretical — surfaces theory-building papers
• [TOPIC] AND (challenges OR barriers OR obstacles) — captures synonyms in one search
• [TOPIC] AND NOT "grey literature" — excludes unwanted result types
• [TOPIC] AND (2020 OR 2021 OR 2022) — date-bounded searching without filters

The Citation Matrix Method

Once you have 15–30 core papers, build a citation matrix to find the most influential works and spot conversation clusters:

1. Create a spreadsheet with papers as rows and themes as columns
2. For each paper, mark which themes it addresses and who it cites
3. Sort by "times cited by papers in your matrix" — the most-connected papers are your anchor sources
4. Look for citation gaps: papers that never cite each other but should. That gap is often your thesis gap.

Tool-by-Tool Comparison

Perplexity (Free + Premium Education)

Perplexity is an AI-powered search engine that provides answers with cited sources. Use the "Focus" feature to restrict results to Academic, Reddit, or the Web. For literature discovery and "state of the art" questions, it's the fastest tool you own.

Research Rabbit

A free tool that helps you discover related papers through citation mapping. Start with a paper you know, and it will show you connected research you might have missed — including papers that CITE yours (which Google Scholar hides behind a login).

sciSpace

Lets you interact with research papers directly — ask questions about a specific paper, get plain-English explanations of complex sections, and find related work. Its Chrome extension adds an "Explain" button to any paper.

NotebookLM

Google's AI research tool. Upload your documents and it can summarize, answer questions, and generate study materials grounded ONLY in your sources.

Elicit (for systematic reviews)

Elicit.org extracts structured data from papers automatically — participants, sample sizes, effect sizes, and findings — into a table you can sort. Ideal when you need to compare 50 papers side by side.

Prompt Engineering for Research-Grade Output

• Give the AI a role: "You are a systematic reviewer with 20 years' experience in public health."
• Chain your requests: ask it to first find sources, then critique them, then synthesize — one step at a time.
• Request structure: "Respond with a table: Author, Year, Method, Sample, Finding, Limitations."
• Force citations: "Only use sources you can cite with a DOI or URL. Mark anything uncertain as [UNVERIFIED]."
• Ask for the counterargument: "What would a critic of this approach say?" — this produces the critical evaluation your supervisor wants.

Ready-to-Use Prompt Library

Template 1: Seminal Works Finder

Identify the 10 most seminal works in [FIELD] published before 2015.
For each: title, author(s), year, journal, and a one-paragraph summary of
why it became foundational. Rank them by citation impact and explain
how each shaped later research. Cite sources.

Template 2: Research Gap Hunter

Based on this list of papers: [PASTE LIST]. Identify 5 research gaps,
each with: (a) what we know, (b) what we don't know, (c) why it matters,
(d) what methodology could answer it. Rank by feasibility for a
doctoral thesis.

Template 3: Methodology Comparison

Compare the methodological approaches used in these 3 papers:
[PAPER A], [PAPER B], [PAPER C]. For each: design, sample, measures,
analysis, strengths, weaknesses. Recommend the strongest approach for
replicating this research in a different country.

Template 4: Thematic Synthesis

Group these 30 papers into 4-6 thematic clusters: [PASTE PAPERS].
For each cluster give: the core debate, the key authors, how positions
have shifted over time, and 3 papers that best represent it. Output as
structured headings I can drop into my literature review chapter.

The Complete Literature Review Checklist

• Scope defined — topic sentence + 2–4 concepts written down
• Search diary started — databases, dates, terms, result counts logged
• Broad search done — Perplexity + Brave Search + Google Scholar
• Boolean strings tested — at least 5 operator combinations per concept
• Citation chain walked — "Cited by" + "Related articles" on every anchor paper
• Citation matrix built — 15–30 papers scored against themes
• PDFs shortlisted — maximum 30 for a chapter-scale review
• PDFs uploaded to NotebookLM — summaries generated per paper
• Research gaps listed — at least 3 candidate thesis angles
• Reference list started — APA 7th, organized by theme, with DOIs
• Counter-arguments noted — limitations of each major position
• Supervisor preview ready — one-page executive summary written

Output Structure for Your Literature Review

1. Executive overview — the field and its major scholarly conversations
2. Thematic synthesis — literature organized by major research streams
3. Methodological landscape — approaches used across the discipline
4. Research gaps — potential thesis angles and original contributions
5. Complete reference list — APA 7th edition, organized by theme, with DOIs/URLs

Troubleshooting

• "AI keeps inventing sources" — Always add "cite with DOI or URL, mark uncertain as [UNVERIFIED]" to your prompt, and spot-check every citation in Google Scholar.
• "Too many results" — Add AND + "empirical study", limit the year range, and require your keyword in the title: intitle:[KEYWORD].
• "Too few results" — Remove the most restrictive term, use OR for synonyms, and try the wildcard *.
• "Paywalled papers" — Search the exact title in Brave Search; authors often host PDFs on ResearchGate or university sites. Use your KNUST library proxy for institutional access.
• "Can't access Scopus/Web of Science" — KNUST students get free access through the library portal. Log in with your student credentials before searching.