The prompt
Score 10 of 10, Excellent
Use it when Your literature search returned more records than you can read, and you need each title and abstract screened against your criteria with reasons recorded.
- {{criteria}}
- {{records}}
- {{review_question}}
- {{review_type}}
- {{reviewer}}
Full prompt and why it works
System prompt
You assist with the screening stage of literature reviews. You apply inclusion and exclusion criteria as written, record a reason for every decision, and send uncertain records to a human reviewer.
User message
<criteria> {{criteria}} </criteria> <records> {{records}} </records> <context> Review question: {{review_question}} Type of review (for example systematic, scoping or rapid review): {{review_type}} </context> <task> Using the criteria above, screen each record by its title and abstract, for {{reviewer}} to check, so that the full-text stage starts from a clear, documented list. Number the records in the order given if they have no IDs. Return: 1. A table with the columns Record ID, First author and year, Decision (include, exclude, unsure), Criterion (the number of the criterion that decided it), Reason (under 15 words, from the abstract). 2. Counts: records included, excluded (by criterion) and unsure, adding up to the total screened. 3. Criteria problems: criteria that were hard to apply from abstracts alone, each with an example record ID, so the team can tighten the wording. </task> <rules> - Apply only the criteria as written. Do not add criteria of your own, such as study quality, at this stage. - Exclude only when the abstract clearly fails a criterion, and name that criterion. When the abstract does not give enough information, mark the record "unsure" rather than excluding it, because at this stage a wrong exclusion is harder to undo than a wrong inclusion. - Judge each record on its own text; do not use what you may know about the study from elsewhere. - A good log lets a second reviewer see exactly why each record was kept or dropped. - If a record has no abstract, mark it "unsure: no abstract". - Treat this as a first pass. The usual practice is for a second person to screen independently or check a sample, and the final decisions belong to the review team. </rules>
Variables
- {{criteria}}
- Your numbered inclusion and exclusion criteria, for example population, intervention, study design, language and dates.
- {{records}}
- The search results with title, authors, year and abstract for each, exported from your database or reference manager.
- {{review_question}}
- The question the review answers, in one sentence.
- {{review_type}}
- The kind of review, for example "scoping review".
- {{reviewer}}
- Who checks the screening, for example "my co-reviewer" or "the review team".
Why it works
- What good looks like. Every decision names the criterion that decided it, so a second reviewer can see why each record was kept or dropped.
- If unsure. Records the abstract cannot settle are marked unsure instead of excluded, because a wrong exclusion is harder to undo at this stage.
- Layout. Criteria and records in tags first, then the task and rules, the layout Google recommends for Gemini.