Write the question and screening rules
Define the concepts, population or context and the type of evidence you need. Choose inclusion and exclusion rules before reading results. For example, specify language, date range and study type where relevant. Keep a search log with the terms, source locations and dates so another reader can understand the coverage.
Build a synthesis matrix before drafting
Record each source’s question, method, sample, finding and limitations. Group papers by the issue they address rather than simply listing one abstract after another. Two papers can appear to disagree because they use different definitions or settings. Preserve those distinctions instead of forcing a single conclusion.
- Give each source a stable identifier.
- Record what the study measured.
- Separate author conclusions from your interpretation.
- Note missing full text or incomplete metadata.
Use the agent for a bounded literature review
Choose a literature-review output in Humanizily Research Agent and approve an outline organized around your research question. Supply relevant papers where available and describe the source scope. Ask for theme-based synthesis with citations. Review the evidence table and open the original sources before accepting statements about consensus or gaps.
Describe the review’s limitations honestly
State which sources and languages were searched and which access limits affected selection. Do not describe a general web-assisted review as exhaustive. A systematic review requires a defined protocol and appropriate search, screening and reporting methods; choosing an output label does not complete those steps. Keep the review proportionate to what you actually did.
Worked example
Starting point
Summarize papers about feedback.
Suggested wording or approach
Compare how selected studies define timely feedback in online undergraduate courses. Group findings by timing and feedback type, record study limitations, and distinguish disagreements from differences in setting.
Example research scope. It does not claim complete database coverage or identify a real consensus.
Common questions
Is an AI literature review automatically systematic?
No. The review method, documented search and screening process determine that description.
Can I use abstracts only?
They may help initial screening, but disclose this limitation and avoid claims requiring unread full text.
How should I organize the draft?
Use themes, methods or debates that answer the question rather than a sequence of isolated paper summaries.
Start with a focused research brief
Define your question, review the outline and check the evidence before sharing the report.
Open the Research AgentHow the agent works