Aug 10, 2026

How to Set Inclusion and Exclusion Criteria for a Literature Review: Screening and Study Selection

By JournalLabs Research Team

Introduction

A literature review depends on more than finding a large number of papers.

Researchers also need a consistent way to decide which studies belong in the review and which do not. Without explicit inclusion and exclusion criteria, study selection can become inconsistent, difficult to explain, and vulnerable to bias.

For example, a researcher reviewing digital mental-health interventions may find studies involving different age groups, technologies, clinical conditions, study designs, and outcome measures. Some papers may appear relevant from the title but fail to address the review question when the full text is examined. Others may report useful evidence but use a population, intervention, or publication type outside the intended scope.

Inclusion criteria define what a study must meet to enter the review, while exclusion criteria identify conditions that make it unsuitable. Together, they convert a broad question into a practical screening framework.

They should be established before final selection, tested against real results, and applied consistently during title-and-abstract and full-text screening.

This guide explains how to define criteria, screen studies, record exclusions, handle borderline cases, and prepare the final evidence set.

What Are Inclusion and Exclusion Criteria?

Inclusion criteria describe the required characteristics of eligible studies.

They may define:

• Population or sample
• Intervention, exposure, or phenomenon
• Comparator
• Outcomes
• Study design
• Research setting
• Publication period
• Language
• Publication type
• Geographic scope

Exclusion criteria describe characteristics that disqualify a study.

Examples include:

• The wrong population
• An unrelated outcome
• A noneligible study design
• No accessible full text
• Duplicate publication
• Conference abstract without sufficient data
• Editorial, commentary, or protocol when primary evidence is required
• Publication outside the defined date range

The two lists do not need to mirror every statement. If “adults aged 18 years or older” is an inclusion criterion, repeating “participants younger than 18 years” may be unnecessary unless mixed-age samples require a rule.

The purpose is to make decisions operational, consistent, and traceable.

Why Study-Selection Criteria Are Difficult to Set

Research Questions Are Often Broader Than Screening Decisions

A question such as:

How does remote work affect employee well-being?

does not yet specify whether the review includes hybrid work, self-employed workers, students, qualitative studies, preprints, or papers measuring job satisfaction rather than well-being.

Screening requires more precise boundaries than the initial topic statement.

Relevant Papers May Describe the Same Concept Differently

A study may use telework, distributed work, home-based work, or flexible work arrangements instead of remote work.

Eligibility criteria should define the concept, not depend only on one keyword.

Criteria Can Become Too Broad or Too Narrow

Broad criteria may create an unmanageable evidence set containing studies that are difficult to compare.

Narrow criteria may exclude important evidence before the researcher understands the field.

The objective is not to maximize the number of included studies. It is to create an evidence set that can answer the review question.

Abstracts Do Not Always Provide Enough Information

An abstract may omit ages, measurement details, publication status, subgroup definitions, or study design. Some decisions therefore require full-text review.

Decisions Can Change After Results Are Seen

Researchers may be tempted to modify criteria because a paper supports or contradicts an expected conclusion.

Criteria can be refined when genuine ambiguities emerge, but changes should be documented and applied consistently to all studies.

The Traditional Study-Screening Workflow

A conventional screening process usually follows five stages:

1. Define the review question and scope.
2. Translate the scope into explicit eligibility criteria.
3. Remove duplicates and screen titles and abstracts.
4. Review the full text of potentially eligible studies.
5. Record included studies and reasons for full-text exclusion.

For formal reviews, multiple reviewers may screen records independently and resolve disagreements through discussion or a third reviewer.

The workflow is demanding when results are numerous, abstracts are incomplete, and studies are borderline.

How AI Can Support Screening Without Making the Final Decision

AI can help organize the screening process, but it should not determine eligibility without researcher oversight.

Possible uses include:

• Converting a review question into candidate eligibility categories
• Structuring criteria into a consistent checklist
• Summarizing titles and abstracts
• Highlighting missing eligibility information
• Grouping records by likely relevance
• Flagging possible duplicates
• Organizing full-text exclusion reasons
• Identifying inconsistent screening decisions

AI may still misclassify a study, overlook a qualifying detail, or infer unstated information. Final decisions should remain with researchers and use the full text when necessary.

Traditional vs AI-Assisted Study Selection

AI can improve organization and speed, but it does not replace source verification or a documented screening protocol.

Step-by-Step Guide to Setting Inclusion and Exclusion Criteria

1. Start with a Focused Review Question

A useful question defines the main concepts that should determine eligibility.

Instead of:

AI in education

use:

How do generative AI writing tools affect writing performance among university students?

This question suggests several screening dimensions:

• Population: university students
• Intervention or exposure: generative AI writing tools
• Outcome: writing performance
• Setting: higher education
• Study purpose: effects or associations

2. Define the Population

Specify who or what must be studied.

Possible criteria include:

• Age range
• Diagnosis or condition
• Occupation
• Educational level
• Geographic location
• Dataset type
• Species or experimental model

Also decide how to handle mixed populations. A study may include both eligible and ineligible participants. The protocol should state whether subgroup data must be reported separately.

3. Define the Intervention, Exposure, or Phenomenon

Clarify what must be examined.

For example, a review of generative AI writing tools may include systems that generate or revise text but exclude automated grammar checkers that do not use generative models.

Avoid definitions that rely only on brand names or current terminology. Describe the underlying function.

4. Define Comparators and Outcomes

Some reviews require a specific comparator, such as usual care, placebo, in-person instruction, or no intervention.

Outcomes should also be operationalized.

“Writing performance” could refer to:

• Rubric scores
• Error rates
• Revision quality
• Argument structure
• Instructor evaluation
• Standardized assessment results

Decide whether self-reported perceptions count as outcomes or only objective performance measures are eligible.

5. Define Eligible Study Designs

Specify whether the review includes:

• Randomized controlled trials
• Cohort studies
• Cross-sectional studies
• Case-control studies
• Qualitative studies
• Mixed-methods studies
• Simulation studies
• Systematic reviews
• Meta-analyses

The appropriate design depends on the review question. A review of effectiveness may prioritize comparative studies, while a review of user experience may require qualitative evidence.

6. Set Publication and Practical Boundaries

Possible boundaries include:

• Publication years
• Languages
• Peer-review status
• Full-text availability
• Countries or regions
• Publication types

Each restriction needs a reason. A recent date range may suit fast-changing technology, while language restrictions may introduce bias.

7. Write Operational Rules

Criteria should support clear decisions.

Weak criterion:

Relevant studies about AI and writing

Operational criterion:

Empirical studies evaluating the use of generative AI tools for producing or revising written work among students enrolled in higher-education institutions.

Operational definitions reduce inconsistent interpretation.

8. Pilot the Criteria

Apply the draft criteria to a small sample of clearly relevant, clearly irrelevant, and borderline records.

Ask:

• Are important studies being excluded?
• Are unrelated studies being included?
• Are any criteria ambiguous?
• Can two reviewers apply the rules consistently?
• Is required information available in abstracts?

Revise unclear rules before screening the full result set.

9. Document Every Change

Record the original and revised criteria, why and when they changed, and whether earlier records were reassessed. This prevents undocumented post hoc decisions.

How to Screen Titles, Abstracts, and Full Texts

Title Screening

Title screening removes clearly irrelevant records.

Exclude only when the title makes ineligibility obvious. A vague title should usually proceed to abstract screening.

Abstract Screening

Review the abstract against the eligibility checklist.

Classify each record as:

• Include for full-text review
• Exclude
• Unclear

Use “unclear” when essential information is missing. Do not infer that an unstated criterion has been met.

Full-Text Screening

Full-text screening confirms final eligibility.

Check:

• Population
• Intervention or exposure
• Comparator
• Outcome
• Study design
• Publication type
• Data availability
• Duplicate or overlapping samples

When excluding a full-text article, record one clear primary reason.

Examples include:

• Wrong population
• Wrong intervention
• Wrong outcome
• Ineligible design
• No usable results
• Duplicate dataset
• Not an empirical study

Consistent reasons make the selection process easier to report and audit.

How to Handle Borderline Studies and Screening Disagreements

Create decision rules for recurring borderline issues, such as:

• Mixed populations
• Partially relevant interventions
• Multiple reports from the same study
• Unclear publication status
• Outcomes reported only in supplementary material

When reviewers disagree, compare the decision with the written rule. Resolve it through discussion, a third reviewer, clarification of an ambiguous rule, or author contact when essential information is unavailable.

Apply revised rules to all comparable records.

Common Inclusion and Exclusion Criteria Mistakes

Creating Criteria After Reading the Results

This can introduce selection bias. Define the framework before final screening.

Using Vague Language

Terms such as relevant, high quality, or appropriate are not operational criteria unless they are clearly defined.

Excluding Studies Because of Their Findings

Eligibility should depend on the question, population, methods, and evidence type, not whether results support an expected conclusion.

Applying Too Many Restrictions

Unnecessary language, date, geography, or publication-type limits may remove useful evidence.

Confusing Search Filters with Eligibility Criteria

A database filter controls retrieval. An eligibility criterion determines whether a study belongs in the review.

A paper missed by a search filter cannot be screened, so retrieval restrictions should be used cautiously.

Excluding Unclear Abstracts Too Early

Missing information is not proof of ineligibility. Move uncertain records to full-text review.

Recording Inconsistent Exclusion Reasons

Use a controlled set of primary reasons so similar studies are treated consistently.

Failing to Reassess Earlier Decisions

When criteria change, previously screened records may need to be reviewed again.

How JournalLabs Supports the Selected Evidence Set

After researchers define their criteria and complete study selection, JournalLabs’ AI Literature Review helps turn the selected papers into a structured, editable evidence synthesis.

Researchers can use JournalLabs to:

• Organize included studies into consistent fields
• Compare designs, populations, methods, and findings
• Identify agreements and contradictions
• Group evidence by theme
• Surface recurring limitations
• Build an initial review structure
• Keep synthesis statements connected to selected papers

JournalLabs should be used after the researcher has determined which studies belong in the evidence set. It supports comparison and synthesis, while the researcher remains responsible for eligibility decisions, source verification, and the final interpretation.

Frequently Asked Questions

What Is the Difference Between Inclusion and Exclusion Criteria?

Inclusion criteria define what a study must contain to be eligible. Exclusion criteria identify conditions that make it ineligible.

When Should Criteria Be Set?

They should be drafted before final screening and pilot-tested on a small sample of records.

Can Inclusion and Exclusion Criteria Change?

Yes, when genuine ambiguities emerge. Changes should be documented, justified, and applied consistently to previously screened records when necessary.

Should Publication Year Be an Eligibility Criterion?

Only when the review question justifies it. Date restrictions may suit rapidly changing fields but may exclude foundational evidence.

Should Non-English Studies Be Excluded?

Only with a documented practical or methodological reason. Language restrictions can affect the evidence set and should be reported transparently.

What Happens When the Abstract Does Not Provide Enough Information?

Classify the record as unclear and review the full text rather than assuming it is ineligible.

Can AI Screen Studies Automatically?

AI can support prioritization, extraction, and consistency checks, but final eligibility decisions should be verified by researchers against the source.

Conclusion

Inclusion and exclusion criteria turn a broad literature-review question into a transparent study-selection process.

Strong criteria define the population, intervention or exposure, comparator, outcomes, study design, publication type, timeframe, language, and other boundaries that matter to the question.

They should be operational rather than vague, tested before full screening, and applied consistently across titles, abstracts, and full texts.

Researchers should record full-text exclusions, document changes, and use explicit rules for borderline cases.

The strongest workflow is:

Define the review question, translate it into eligibility criteria, pilot the rules, screen in stages, document decisions, and build a final evidence set that can be compared and synthesized.

Build a Literature Review with JournalLabs

Turn your selected papers into structured comparisons, traceable themes, and an editable evidence synthesis with JournalLabs.

Start building your literature review with JournalLabs.