AI tutor in a first-year statistics course
A 2023 controlled study of 180 university students reports achievement and engagement outcomes.
Literature review expert guideMaster’s · PhD · Research
Find the review method your question needs.
Start with what you need to understand—not a method name. LitWise turns your research intent, discipline, and evidence into a defensible review pathway.
- 14
- review methods explained
- 09
- discipline families mapped
- 05
- decisions to a starting method
Choose your entry point
Start from where you are now.
You do not need to read this guide from top to bottom. Choose the situation that is closest to yours and move to the next useful decision.
01
I have not started yet
I have a topic or a rough idea, but I do not know which review method fits it.
Choose a review method→02I already have papers
My PDFs and notes are growing, and I need a traceable way to organise the work.
Follow the workflow→03I need field-specific guidance
I need to know which sources, standards, journals, and methods matter in my discipline.
Explore my discipline→04I need practical tools
I am ready to search, screen, appraise, extract, analyse, or start writing.
Open the toolkit→01 · Guided pathway
Make the method decision in five clear steps.
Use this as a starting hypothesis. Confirm the final design with your supervisor, advisory committee, or review team.
Step 1 of 5 · Research intent
What are you trying to understand?
Choose the closest intent. A useful method begins with the decision your review must support.
04 · Method library
Compare the logic—not only the labels.
Compare before choosing
Put up to three review methods side by side.
A familiar name is not enough. Compare the claim, workload, team, output, and reporting expectations before committing.
Select methods to compare0/3 selected
Choose at least two methods to reveal the comparison table.
03 · Discipline atlas
Enter through your field. Then follow the question.
Every discipline has its own evidence culture. Select a field to see its common questions, sources, methods, and caution points.
02 · Execute the review
Execute your review in six traceable phases.
Each phase should leave an artefact that another researcher—or your future self—can inspect.
- 02
Search
Translate the question into concepts, synonyms, sources, and a reproducible search log.
Leave behind- Concept blocks
- Piloted strategies
- Search log
Decision gateDo known sentinel papers appear, and is every search reproducible?
- 03
Screen
Apply eligibility rules consistently and preserve the reason for every exclusion.
Leave behind- Deduplicated library
- Screening decisions
- Flow counts
Decision gateWould two reviewers interpret the criteria in the same way?
- 04
Appraise
Judge the kind and strength of claim each source can support.
Leave behind- Appraisal records
- Risk-of-bias judgements
- Evidence limitations
Decision gateIs the tool appropriate to the study design and review purpose?
- 05
Extract
Capture comparable evidence without stripping away context needed for synthesis.
Leave behind- Extraction form
- Evidence matrix
- Audit notes
Decision gateCan every synthesized claim be traced back to a source location?
- 06
Write
Build an argument that connects findings, uncertainty, limitations, and contribution.
Leave behind- Synthesis structure
- Tables or maps
- Reporting checklist
Decision gateDoes the conclusion stay inside the evidence and protocol boundaries?
03 · Guided practice
Build the artefacts, not just the vocabulary.
Use these client-side tools to rehearse decisions and create copy-ready working notes. Your checklist stays in this browser; no research data is uploaded.
Saved only in this browser
My review project
A progress checklist for the minimum traceable workflow. Completion is a reminder—not a quality score.
0%0/12 completed
01
Question framework builder
What do you want this review to answer? Choose the goal closest to your project. We will suggest a question framework, show a complete example, and give you fields for your own topic. Adapt the wording and scope to your discipline.
What kind of answer are you looking for?
View a completed examplePICOPopulation · Intervention · Comparison · Outcome
- Population or problem
- First-year university students
- Method, programme, or option
- Flipped-classroom teaching
- What it is compared with
- Lecture-based teaching
- Outcome of interest
- Achievement and classroom engagement
Question produced from this example
Among first-year university students, does flipped-classroom teaching improve achievement and classroom engagement compared with lecture-based teaching?
Suggested starting frameworkPICOPopulation · Intervention · Comparison · OutcomeA focused question comparing effects or outcomes
Question draft
Among [population], how does [intervention], compared with [comparison], affect [outcome]?
02
Screening calibration lab
Apply one example protocol to six records. At title/abstract stage, uncertainty normally moves forward instead of becoming an early exclusion.
0/6correct decisions
Why universities should embrace AI tutors
A 2024 editorial argues for adoption but reports no empirical study or learner outcome.
Adaptive tutoring in formal education
The abstract reports learning gains but does not identify whether participants were school or university students.
Generative tutoring for grade 8 mathematics
A 2022 school study reports test performance among 13–14 year-old pupils.
Protocol for evaluating a university AI tutor
A 2025 protocol describes planned outcomes but contains no results.
AI feedback assistant in postgraduate writing
A mixed-methods university study reports engagement but it is unclear whether the system provides tutoring or only administrative feedback.
03
PRISMA flow planner
Enter counts to check arithmetic and produce a copy-ready selection summary. This planning aid does not replace the official PRISMA 2020 diagram or its instructions.
Selection counts
Derived flow
Records identified0↓
Removed before screening0↓
Records screened0↓
Reports sought0↓
Reports assessed0↓
Studies included0
Open official PRISMA flow diagrams04
Statistical test chooser
Answer a few questions to find a sensible starting point, then check the assumptions, study design, and reporting guidance before running the analysis.
What does your research question ask?
Browse all 14 techniquesYou can open a technique directly if you already know what you need.+
05 · Research toolkit
Move from advice to working research artefacts.
Copy a starter, adapt it to your discipline, and preserve the decisions that make your review auditable.
Boolean search canvas
Build one block per concept. Add synonyms with OR; connect concepts with AND; add controlled vocabulary inside each database.
("artificial intelligence" OR "machine learning" OR AI)
AND
(education OR teaching OR learning)
AND
(ethics OR bias OR fairness)Search the right fields
Use database field tags such as title/abstract or TITLE-ABS-KEY when an all-fields search returns too much noise.
Use truncation carefully
A wildcard can capture useful word endings, but test every truncated term so it does not retrieve unrelated words.
Test proximity operators
When a database supports NEAR, W/n, or an equivalent operator, use it to keep related concepts close without requiring an exact phrase.
AI-assisted research
AI prompt lab
Use AI to expand, compare, and challenge your thinking while keeping every scholarly decision and source check with the researcher.
A reliable research prompt names four things
- 01Context
Topic, discipline, purpose, and the material available.
- 02Task
One concrete job for the AI to help with.
- 03Evidence rules
What it may use, must verify, and must not invent.
- 04Output
A structure that makes gaps and uncertainty visible.
Choose a research task
Copy-ready prompt
Expand search vocabulary
Turning a topic into concept blocks, synonyms, spelling variants, and terms to test in databases.
Context:
I am reviewing [topic] in [discipline, population, or setting]. My current question is [question or decision].
Task:
Help me build vocabulary for a reproducible database search. Separate the question into concept blocks and propose synonyms, acronyms, spelling variants, broader terms, and narrower terms for each block.
Evidence rules:
- Do not present a term as an official subject heading unless it has been checked in the named database thesaurus.
- Separate natural-language keywords from controlled-vocabulary candidates that I still need to verify.
- Flag terms likely to retrieve irrelevant results and explain why.
- Do not claim that the list is complete.
Output:
Create a table with: concept block | keywords and variants | controlled terms to verify | likely noise. Then draft Boolean blocks, clearly marking any syntax that must be adapted for [database].Choose appraisal by evidence design
The design of the included study—not the prestige of its journal—determines the appraisal lens.
| Evidence design | Inspect | Possible starting tool |
|---|---|---|
| Randomized intervention | Randomization, deviations, missing outcomes, measurement, reporting | RoB 2 or a field-equivalent tool |
| Non-randomized intervention | Confounding, selection, classification, deviations, missingness | ROBINS-I or a field-equivalent tool |
| Qualitative study | Fit, recruitment, reflexivity, analysis, context, credibility | CASP, JBI, or a discipline-appropriate tool |
| Mixed-methods study | Quality of each component and the logic of integration | MMAT or design-specific tools plus integration appraisal |
| Systematic review | Protocol, search, selection, appraisal, synthesis, bias | AMSTAR 2 or a purpose-appropriate review tool |
Research tool directory
Choose tools by the job they must do. Start with the smallest workable stack, confirm institutional access, and record any automation used in screening or analysis.
Reference management
Collect, deduplicate, annotate, and cite without losing the source trail.
Zotero
Free core · paid storageA strong default for most students and research teams, with browser capture and an open ecosystem.
Agree on collections, tags, and attachment storage before collaborating.Mendeley Reference Manager
FreeResearchers who want a straightforward PDF library and citation workflow.
Check export and collaboration needs before committing a large shared library.EndNote
Paid · institutional access may applyLabs and institutions that already use EndNote styles, libraries, and support.
Check whether your university already provides a licence before purchasing.Screening & review management
Remove duplicates, apply eligibility criteria, resolve conflicts, and preserve an audit trail.
Rayyan
Free + paid plansTitle and abstract screening, especially when several reviewers need blinding and conflict resolution.
Free and paid capabilities differ; agree on reviewer roles before inviting the team.ASReview
Free · open sourceLarge screening sets where active learning can prioritise likely-relevant records while humans retain decisions.
Predefine stopping rules and report the model, version, and human decisions transparently.EPPI-Reviewer
Subscription · partner access may applyComplex evidence reviews and maps that need collaborative coding, synthesis, and review management.
It can be more platform than a small thesis needs; check training and access first.Notes, writing & collaboration
Turn reading notes into a shared synthesis and a manuscript that remains maintainable.
Notion / Obsidian
Free core · optional paid servicesLinked notes, concept development, and a durable synthesis workspace before drafting.
Choose Notion for shared work or Obsidian for local files; avoid maintaining the same notes in both.Overleaf
Free + paid plansLaTeX manuscripts, technical theses, and real-time collaboration around journal templates.
Confirm that your supervisor and coauthors are comfortable reviewing LaTeX projects.Google Docs + Paperpile
Docs free · Paperpile paidTeams that already review drafts in Google Docs and want citations inside that workflow.
Decide which reference library is authoritative to prevent duplicate or drifting records.Analysis & synthesis
Code qualitative findings, calculate pooled estimates, or manage an end-to-end review.
NVivo / ATLAS.ti
Paid · student and institutional optionsSystematic coding of qualitative findings, memos, relationships, and team-based analysis.
Software does not choose a coding logic; define the analytical approach before importing everything.R + metafor
Free · open sourceReproducible meta-analysis, sensitivity checks, and publication-quality statistical outputs.
Use it with statistical supervision when effect measures, dependence, or heterogeneity are complex.Covidence
Paid · often institutionalTeams that want screening, extraction, risk-of-bias work, and exports in one review platform.
Check university access and confirm that its workflow matches your review method before setup.Copy-ready template
Search log template
Make every database search reproducible.
Database: [name] Platform: [provider] Date searched: [YYYY-MM-DD] Coverage: [start–end] Exact search string: [paste] Filters: [list or none] Results returned: [n] Export file: [name] Notes and changes: [why]
Copy-ready template
Eligibility decision form
Keep screening decisions consistent and explainable.
Study ID: [author-year] Reviewer: [name] Stage: [title/abstract | full text] Decision: [include | exclude | discuss] Criterion triggered: [criterion] Exclusion reason: [one specific reason] Notes: [context or uncertainty]
Copy-ready template
Evidence extraction matrix
Connect study context, design, finding, and quality.
Citation | Country/setting | Aim | Design | Sample/data | Intervention or phenomenon | Outcome/finding | Context | Limitation | Appraisal | Reviewer note
Copy-ready template
One-page protocol starter
Align the team before the search expands.
Decision this review supports: Review question: Review type and rationale: Population/phenomenon: Concept/intervention: Context/comparator: Eligible evidence designs: Sources to search: Appraisal approach: Synthesis approach: Registration/reporting plan: Known constraints:
Six failure modes to catch early
- 01Method-first thinking
Choosing a fashionable label before defining the claim and evidence.
- 02One-database confidence
Treating one index or Google Scholar as complete coverage.
- 03Moving criteria
Changing eligibility after seeing results without documenting the decision.
- 04Tool mismatch
Using one appraisal checklist for unlike study designs.
- 05Spreadsheet amnesia
Keeping findings without page, table, or quotation traceability.
- 06Conclusion inflation
Making causal or universal claims that the included evidence cannot support.
Official reporting resources
Guidelines evolve. Confirm the current version and the requirements of your target journal or institution.
06 · Before the search
Write these four sentences before opening a database.
- 01My review must help someone decide…
Name the decision, audience, or knowledge gap—not just the topic.
- 02The review includes evidence about…
Define people, phenomena, concepts, settings, dates, and boundaries.
- 03I will judge evidence by…
Choose quality criteria that match each study design and knowledge claim.
- 04The final contribution will be…
State whether you need an estimate, map, explanation, theory, or agenda.
“The method is justified by the question, evidence, and intended claim.”LitWise decision principle
Ready to begin?