Literature review protocol template for AI-assisted research

A literature review protocol defines how the review will be done before the team starts making source-selection decisions. When AI tools are involved, the protocol should also explain where AI can assist, what humans will verify, and how tool-assisted steps will be recorded.

The protocol does not need to be complicated. It needs to make the review question, search plan, screening criteria, extraction process, and AI-use boundaries visible.

This guide gives a practical literature review protocol template for AI-assisted research.

A literature review protocol is a plan for conducting a review. It explains the question, scope, sources, search strategy, screening process, extraction fields, synthesis approach, and reporting plan.

For AI-assisted research, the protocol should also define:

• Which AI tools may be used.
• Which tasks they may support.
• Which outputs require verification.
• How AI-assisted decisions are recorded.
• How citations and source claims are checked.

The protocol protects the review from decisions that change silently halfway through the project.

AI-assisted reviews need a protocol because AI can quickly expand, summarize, rank, or reshape a source set. That speed is helpful, but it can also make the workflow harder to explain.

A protocol helps prevent:

• Unclear search boundaries.
• Inconsistent inclusion decisions.
• Unreported AI-prioritized screening.
• Unchecked extraction fields.
• Citation errors.
• Overbroad source sets.
• Unsupported synthesis claims.

For team-level AI rules, see AI literature review policy template for research teams.

The protocol should start with a title and review question that define the project's scope. A vague title leads to a vague review.

Include:

• Working title.
• Review type.
• Research question.
• Population, concept, context, intervention, method, or outcome as relevant.
• Time period if needed.
• Source types.
• Main purpose of the review.

Example:

"This review examines how AI-assisted screening affects source selection in systematic reviews, focusing on published studies and method papers from 2020 onward."

The wording can change later, but the protocol should record changes.

Inclusion and exclusion criteria should turn the review question into screening decisions. They should be specific enough that reviewers can apply them consistently.

Include criteria for:

• Topic.
• Population or material.
• Method or study design.
• Outcome or concept.
• Publication type.
• Language.
• Date range.
• Peer-review status.
• Grey literature, if included.
• Full-text availability.

For criteria development, see inclusion and exclusion criteria for literature reviews.

The search plan should explain where and how sources will be found. If AI search is used, it should be treated as one part of the search workflow.

Include:

• Databases or search tools.
• Search strings.
• Natural-language AI search queries.
• Seed papers.
• Citation chasing plan.
• Grey literature sources, if any.
• Search dates.
• Filters.
• Search log format.
• Recall checks.

AI can help generate search terms, but the final search plan should be testable.

For query design, see AI literature review search strings.

Describe AI use by task. Do not write only that "AI will be used."

Specify whether AI may help with:

• Search-term brainstorming.
• Academic search.
• Abstract summarization.
• Screening prioritization.
• Full-text inspection.
• Data extraction suggestions.
• Theme development.
• Citation checking.
• Draft organization.

For each task, state what a human will verify. This keeps AI assistance inside the research method rather than floating outside it.

For responsible boundaries, see responsible AI automation checklist for research teams.

The screening plan should explain how records move from search results to included sources.

Include:

• Deduplication method.
• Title and abstract screening rules.
• Full-text screening rules.
• Reviewer roles.
• Conflict resolution.
• Exclusion reason categories.
• AI prioritization, if used.
• Stopping rule, if not screening all records.
• Audit sample plan, if applicable.

If AI changes screening order, record that clearly.

For stopping rules, see when to stop screening in an AI-assisted review.

The extraction plan should define what information will be taken from included sources and how it will be checked.

Include fields such as:

• Study design.
• Population or dataset.
• Method.
• Outcome or concept.
• Main finding.
• Limitation.
• Source location.
• Reviewer status.
• Notes for synthesis.

If AI suggests extracted fields, require source-location checks for high-value fields.

For quality control, see data extraction quality control for AI literature reviews.

The synthesis plan should explain how the review will move from extracted notes to claims. This is where many protocols are too thin.

Describe:

• Whether the review will be narrative, thematic, scoping, systematic, or mixed.
• How papers will be grouped.
• How conflicting evidence will be handled.
• How study quality will affect interpretation.
• How gaps will be identified.
• How AI-assisted summaries will be checked before use.

The synthesis plan does not need final answers. It needs a method for building answers from sources.

For argument planning, see how to turn paper notes into an argument outline.

Use this structure:

Review title: Working title.

Review question: Main question and subquestions.

Scope: Included topics, excluded topics, dates, language, and source types.

Search plan: Databases, AI search queries, seed papers, citation chasing, search dates, and log format.

Screening plan: Deduplication, title and abstract screening, full-text screening, reviewer roles, exclusion reasons, conflicts, and stopping rules.

AI use: Allowed tools, supported tasks, human verification points, and documentation.

Extraction plan: Fields, source locations, reviewer status, and quality checks.

Synthesis plan: Grouping method, evidence comparison, conflict handling, and gap framing.

Citation checks: Metadata verification, source support, and correction or retraction checks.

Reporting: Flow diagram, methods note, limitations, and AI-use statement.

Turn the advice into a repeatable workflow by defining the decision you need to make, the evidence required for that decision, and the record that will prove how the decision was made. In review planning, the problem is rarely one missing tool. The problem is usually that search, reading, checking, and writing happen in separate places without a shared rule.

Use a short operating routine:

• Name the review question or subquestion.
• Define the source set you are working from.
• Decide what counts as enough evidence for the next step.
• Apply the same criteria to every paper in that step.
• Mark uncertain cases instead of forcing a clean answer.
• Keep source locations for claims that may enter the final review.
• Review the workflow after each major search, screening, or writing session.

This routine keeps the work moving without making the review careless. It also gives supervisors, collaborators, and future you a way to understand why the source set changed.

Record the pieces that would be hard to reconstruct later. You do not need a diary of every click, but you do need enough detail to explain the path from question to source to claim.

For this topic, the most useful record usually includes scope, criteria, tool use, review type, human checks, and reporting notes. Add the date, tool or source used, reviewer status, and next action. If AI assisted the step, write down what it helped with and what a human checked.

The record should distinguish discovery from evidence. A tool may help find a paper, but the paper itself must support the claim. A summary may help triage a source, but the original source should support any statement that appears in the literature review.

Before writing, check whether the workflow has produced usable evidence or only useful notes. Notes help you think. Evidence supports a sentence.

Ask:

• Which claim will this source support?
• Is the claim narrower than the evidence?
• Have methods, sample, outcome, or concept details been checked?
• Are limitations visible?
• Are conflicting papers handled rather than ignored?
• Is the citation real, current, and relevant?
• Can another reader understand how this source entered the review?

If the answer is unclear, keep the point in notes rather than moving it into the draft. This is the small pause that prevents AI-assisted research from becoming polished but weak writing.

WisPaper can support the search, triage, library, and source-question parts of an AI-assisted protocol. Deep Search, Scholar Agent, and Inspiration Discovery can help researchers find candidate academic papers from natural-language questions.

Paper cards show source labels, summaries, authors, publication details, and preview images, which can support first-pass triage. Selected papers can be saved or uploaded into My Library, and Library QA can answer questions based on the user's own paper set.

For citation verification, TrueCite, powered by WisPaper, checks BibTeX files against real academic databases to flag hallucinated references. The protocol should still require researchers to verify source support for any final claim.

Try WisPaper