PhD Literature Review: How to Plan, Structure, Write, and Manage Yours

A useful way to think about the entire process is:

Research question

↓

Search strategy

↓

Discovery

↓

Screening

↓

Extraction

↓

Comparison

↓

Synthesis

↓

Research gap

↓

Research argument

↓

Research question / objectives / proposal

Your literature review should become more focused as you move down this chain.

If it remains a giant collection of papers until the day you start writing, the workflow has failed.

Step 1: Start With the Research Question, Not the Papers

One of the most common mistakes is beginning the literature review with:

“Let me search my topic.”

The problem is that “topic” is usually too broad.

Suppose your broad area is:

Autonomous mobile robots.

That could produce an enormous literature set covering:

• navigation;
• localization;
• SLAM;
• perception;
• planning;
• manipulation;
• multi-robot systems;
• human-robot interaction;
• reinforcement learning;
• control;
• sensing;
• energy efficiency;
• field robotics;
• agricultural robotics;
• warehouse robotics;
• search and rescue;
• and hundreds of adjacent areas.

You cannot meaningfully review all of that.

Instead, define the research problem you are investigating.

For example:

How can a resource-constrained mobile robot maintain localization when visual sensing becomes unreliable in degraded environments?

Now your literature search has boundaries.

You can identify conceptual dimensions such as:

• resource-constrained robotics;
• localization;
• visual degradation;
• alternative sensing;
• sensor fusion;
• robustness;
• field environments.

The literature review becomes a map of a problem rather than a dump of everything related to a topic.

A useful question before every search

Ask:

What decision will this search help me make?

If you cannot answer that, your search is probably too broad.

Step 2: Define Your Search Concepts

Once the research question is sufficiently focused, break it into conceptual blocks.

For example:

Do not assume that one keyword captures an entire research concept.

Researchers use different terminology for similar ideas.

A paper may never use your preferred phrase even though it is directly relevant.

This is where synonym mapping becomes valuable.

Create a working vocabulary containing:

• synonyms;
• abbreviations;
• alternative spellings;
• technical terminology;
• older terminology;
• disciplinary terminology;
• competing definitions.

Your initial search vocabulary should therefore evolve as you learn the field.

Step 3: Build a Search Strategy Instead of Searching Randomly

Your first search is not supposed to find everything.

It is supposed to teach you how the field describes itself.

Start broad enough to identify:

• landmark papers;
• recent reviews;
• major researchers;
• competing terminology;
• important conferences;
• dominant methods;
• foundational theories.

Then, recursively narrow it down.

A practical layered progression is:

Search Layer 1 — Landscape

broad concept + broad problem

Search Layer 2 — Methods

problem + methodology

Search Layer 3 — Debate

concept + competing approaches

Search Layer 4 — Limitations

approach + limitations

Search Layer 5 — Gap

approach + unresolved problem

This creates a progressively more targeted literature set.

Step 4: Use Seed Papers to Expand the Literature

Once you find a genuinely useful paper, stop treating it as a single source.

Treat it as a node in the literature network.

Two techniques are particularly useful:

Backward citation chaining

Look at the references cited by the paper.

This helps you identify:

• foundational work;
• earlier theories;
• original methods;
• seminal experiments.

Forward citation chaining

Look for later papers that cite the paper.

This helps you identify:

• follow-up studies;
• criticisms;
• extensions;
• competing methods;
• newer findings.

Citation chaining is particularly useful because keyword searches can miss papers that use different terminology. University research guides explicitly recommend backward and forward citation chaining as a literature-search technique.

Your literature review should therefore not be:

Search → Download → Repeat

It should become:

Search → Find seed paper → Follow references → Follow citations → Compare → Discover new concepts → Search again

That is a much more powerful workflow.

Step 5: Do Not Read Every Paper the Same Way

This is one of the biggest productivity traps in doctoral research.

Not every paper deserves the same amount of time.

A useful three-level reading strategy is:

Level 1: Screening

Determine whether the paper is worth deeper attention.

Look at:

• title;
• abstract;
• keywords;
• figures;
• conclusion;
• publication date;
• venue;
• references.

Ask:

Is this actually relevant to my research question?

If not, move on.

Level 2: Analytical reading

For relevant papers, examine:

• research question;
• theoretical framing;
• methodology;
• dataset or population;
• experimental conditions;
• key findings;
• limitations;
• assumptions;
• comparison with previous work.

Level 3: Deep reading

Reserve full, detailed reading for papers that are:

• foundational;
• methodologically central;
• directly related to your proposed contribution;
• highly influential;
• contradictory to your emerging argument;
• or essential for defending a claim.

The objective is not to read every paper from page 1 to page 40.

The objective is to understand the evidence you need to build your argument.

Step 6: Build a Literature Review Database

This is where a structured database becomes dramatically more useful than a folder full of PDFs.

At minimum, track:

The important field is not simply:

“Summary”

It is:

“What does this paper contribute to my understanding of the research problem?”

That distinction prevents your database from becoming an annotated bibliography.

A Literature Review Matrix Is Not the Literature Review

A literature review matrix is a working tool.

It is not the finished chapter.

This distinction is worth emphasizing because many scholars make the following mistake:

1. Create a spreadsheet.
2. Add 100 papers.
3. Fill in the columns.
4. Sort by author.
5. Start turning rows into paragraphs.

That usually produces a descriptive literature review.

The matrix exists to help you see relationships between sources.

The final review should be organized around those relationships.

For example:

Several studies demonstrate that approach A performs well under controlled conditions. However, studies conducted in dynamic environments report significantly lower reliability. This divergence appears to be associated with…

That is synthesis.

The papers are supporting evidence for the argument.

They are not the argument itself.

Step 7: Organize Papers by Themes, Not Authors

One of the strongest signals that a literature review is still immature is an author-by-author structure:

Smith found X.

Jones found Y.

Patel found Z.

Kumar found A.

A stronger structure is:

Theme 1: Approach A

Discuss the collective evidence.

Theme 2: Approach B

Compare the alternative.

Theme 3: Conditions Under Which the Approaches Fail

Now the limitations become visible.

Theme 4: Emerging Approaches

This shows the direction of the field.

Theme 5: Unresolved Problem

Now the reader can see where your research enters the conversation.

This is why literature-review guidance commonly recommends organizing the body around ideas, themes or relationships rather than simply listing studies.

Step 8: Look for Patterns, Not Just Findings

Once you have enough papers, stop asking:

“What did each paper find?”

Start asking:

“What pattern emerges across the papers?”

But what do these patterns look like in real life? Well, they can be agreements, contradictions, methodological dependence, gaps, etc. and we have discussed this at length in the self-help book PhD Scholar’s Compass: From Burnout to breakthrough.

Step 9: Identify What the Literature Does Not Yet Explain

This is where the literature review begins producing research value.

A research gap is not simply:

“Nobody has studied X.”

That claim is often difficult to defend.

Instead, look for gaps in Knowledge, methodology, context, theory, application, integration or, even, reproducibility. Again, you can read these at length in the PhD Scholar’s Compass book and implement at your own pace.

A good literature review helps you determine which of these, if any, genuinely exists.

Step 10: Use the “Claim → Evidence → Limitation → Opportunity” Framework

Here is a practical framework you can apply to individual themes.

1. Claim

What does the literature currently suggest?

2. Evidence

Which studies support that claim?

3. Limitation

Under what conditions does that evidence become weaker?

4. Opportunity

What remains unresolved?

For example:

Existing studies demonstrate that visual SLAM can achieve reliable localization in structured environments.

Evidence: Multiple experimental studies report high localization accuracy under controlled illumination and sufficient visual texture.

Limitation: Performance degrades when illumination changes rapidly, visual texture decreases, or the environment contains substantial occlusion.

Opportunity: This creates a need to investigate sensing strategies that remain reliable under degraded visual conditions.

Notice what happened.

The research gap did not appear from imagination.

It emerged from the boundary of existing evidence.

That is what you want your literature review to accomplish.

Step 11: Separate “Unknown” From “Unimportant”

Not every gap deserves a PhD.

This is a crucial distinction.

You can find thousands of unanswered questions.

The existence of an unanswered question does not automatically make it:

• significant;
• novel;
• feasible;
• fundable;
• technically achievable;
• or worthy of doctoral research.

Evaluate a potential gap using at least four dimensions:

This is where your literature review connects directly to your research-topic strategy.

If your proposed topic cannot survive this test, the problem may not be your writing.

The problem may be your research direction.

Step 12: Don’t Manufacture a Research Gap

This deserves its own warning.

A weak literature review often contains sentences like:

“Few studies have investigated…”

“There is a lack of research on…”

“No research has examined…”

These claims are dangerous unless you have genuinely established the search boundaries that support them.

Instead, use evidence-based language:

“Within the literature reviewed here…”

“Existing studies predominantly focus on…”

“The evidence remains limited regarding…”

“A consistent limitation across these studies is…”

“The literature provides comparatively little evidence about…”

Your job is not to manufacture novelty.

Your job is to demonstrate where the existing evidence stops being sufficient.

Step 13: Distinguish a Research Gap From a Research Problem

These are related but not identical.

A research problem is the issue that motivates investigation.

A research gap is the deficiency in existing knowledge or evidence concerning that problem.

For example:

Problem

Autonomous robots struggle to maintain reliable localization in degraded environments.

Existing knowledge

Several localization methods work effectively under normal sensing conditions.

Limitation

Their performance decreases when visual information becomes unreliable.

Gap

There is insufficient understanding of how adaptive multimodal sensing can maintain localization robustness under changing sensing conditions.

Research question

How can a mobile robot dynamically select and combine sensing modalities to maintain localization reliability under degraded visual conditions?

This progression is exactly what a good literature review should help you build.

Step 14: Understand the Difference Between a Narrative Review and a Systematic Review

Do not automatically turn your PhD literature review into a systematic review simply because “systematic” sounds more rigorous.

They serve different purposes.

A traditional or narrative literature review may be appropriate when your objective is to:

• understand a field;
• synthesize theories;
• map debates;
• establish conceptual context;
• compare approaches;
• develop an argument.

A systematic review follows a much more explicitly defined methodology for searching, selecting, evaluating and synthesizing evidence.

PRISMA 2020 is a reporting guideline for systematic reviews; it provides structured reporting recommendations, including a checklist and flow diagram. It is not a universal requirement for every literature review in every PhD thesis.

Scoping reviews are another distinct approach and are particularly useful when the purpose is to map the extent, nature or characteristics of literature around a broad or complex area.

Therefore:

First determine what kind of review your research question requires. Then choose the methodology.

Do not choose the methodology because a template told you to.

Step 15: Build a Search Log

Your future self will thank you for this.

Record:

• database searched;
• date searched;
• search string;
• filters used;
• number of results;
• important papers discovered;
• inclusion criteria;
• exclusion criteria;
• notes about terminology.

Why?

Because six months later you may discover a new paper and wonder:

“Why didn’t I find this earlier?”

A search log lets you reconstruct your process.

It also becomes particularly important if your literature review has systematic elements or if you need to explain how your evidence base was assembled.

Step 16: Keep Track of Search Saturation

At some point, additional searches begin returning the same papers.

This is useful information.

Suppose you run five different searches and repeatedly encounter the same:

• seminal papers;
• review articles;
• research groups;
• methodologies;
• datasets.

You are beginning to understand the structure of the field.

This does not necessarily mean you have found everything.

It means your search vocabulary is becoming mature.

Now deliberately search for:

• contradictory evidence;
• criticism;
• alternative terminology;
• negative findings;
• failed approaches;
• recent papers;
• adjacent disciplines.

This is where you prevent confirmation bias.

Step 17: Search for Disagreement on Purpose

A surprisingly effective literature-review strategy is:

Search for evidence that could prove your emerging argument wrong.

If your current interpretation is:

“Method A is the dominant solution.”

Search:

“limitations of Method A”

“Method A failure”

“Method A comparison”

“Method A versus Method B”

“Method A robustness”

“Method A real-world performance”

This creates a more balanced literature review.

A PhD thesis is not supposed to read like a marketing brochure for your preferred hypothesis.

Your credibility comes from demonstrating that you understand the strongest evidence against your own position as well as the evidence supporting it.

Step 18: Use AI Carefully

AI can be useful during literature review.

But it should not replace scholarly judgment.

Useful applications include:

• brainstorming search terms;
• generating synonym lists;
• clustering your own notes;
• identifying recurring themes;
• converting your database into preliminary summaries;
• helping compare methodological descriptions;
• generating questions to investigate;
• checking whether an argument is logically structured.

High-risk applications include:

• asking an AI to invent references;
• trusting AI-generated citations without verification;
• treating summaries as substitutes for reading important papers;
• asking AI to write the entire literature review;
• accepting claims without checking the original source.

A doctoral literature review is ultimately an exercise in human scholarly judgment.

AI can reduce administrative workload.

It cannot outsource responsibility for what you claim the literature says.

Step 19: Verify Every Important Claim Against the Original Paper

Never rely solely on:

• a paper’s abstract;
• another paper’s summary;
• an AI-generated summary;
• a citation database;
• or a review article’s characterization of an original study

when the claim is central to your argument.

For important claims, go back to the original source.

Check:

• what the researchers actually tested;
• under what conditions;
• what data they used;
• what the reported limitations were;
• whether the conclusion supports the interpretation you intend to make.

This is especially important when writing statements such as:

“Previous research has demonstrated…”

The next question should always be:

“Exactly where did it demonstrate that?”

Step 20: Turn Your Database Into a Synthesis Map

Once your database contains enough high-quality sources, stop adding columns indefinitely.

Start mapping relationships.

For each major theme, ask:

What do we know?

What do we disagree about?

What methods dominate?

What assumptions are shared?

What contexts are missing?

What limitations recur?

What findings contradict each other?

What questions remain unanswered?

Which unanswered questions matter?

This produces your synthesis map.

Your final literature-review structure should emerge from this map—not from the order in which you happened to read the papers.