Framing & Literature

Act as a critical reviewer for a [FIELD] journal indexed in Scopus Q1.

My draft research question is:
"[PASTE YOUR QUESTION]"

Do the following, in order:
1. Tell me what is vague, unmeasurable, or already well-established in this question.
2. Identify the implicit assumptions I am making.
3. Rewrite it as 3 sharper alternatives: one descriptive, one explanatory, one comparative.
4. For each alternative, state the design and data needed to answer it, and the likely "so what?" objection a reviewer would raise.

Be blunt. Do not compliment the question.

Forcing the model into a reviewer role produces critique instead of agreement. The 'do not compliment' line removes the default flattery.

I am running a literature search on: [TOPIC].

Build a Boolean search strategy I can paste into Scopus and Web of Science.
- Break my topic into 3-4 concept blocks.
- For each block, list synonyms, British/American spellings, acronyms and related terms.
- Combine with AND between blocks, OR within blocks, using truncation (*) and quoted phrases.
- Give me the final string in Scopus TITLE-ABS-KEY() syntax and a plain-text version.
- Then suggest 3 terms that are likely to create false positives and how to exclude them.

Search strings are pure syntax and vocabulary work — a task with no fabrication risk and a huge time saving.

Below are [N] abstracts I retrieved from my database search.

My inclusion criteria: [LIST]
My exclusion criteria: [LIST]

For each abstract, output a table row: ID | Include / Exclude / Unclear | one-line reason | which criterion it maps to.
Mark "Unclear" whenever the abstract does not contain enough information — do not guess.
At the end, list the IDs you marked Unclear and tell me exactly what I need to check in the full text.

ABSTRACTS:
[PASTE]

You supply the abstracts, so nothing can be invented. 'Unclear' is an allowed answer, which stops the model from forcing a decision.

Using only the sources I uploaded, build a synthesis matrix as a table with these columns:
Author & year | Context/sample | Theoretical lens | Method & data | Key finding | Stated limitation.

Then, still using only these sources:
1. Group the studies into 3-5 thematic clusters and name each cluster.
2. Identify where findings conflict, and quote the specific sentences that conflict.
3. List what none of these sources address.
Cite the source for every row and every claim.

NotebookLM is grounded in your uploaded documents and cites back to them, so every line is checkable. This is the single highest-value AI use in a literature review.

Here are my synthesis notes: [PASTE MATRIX OR SUMMARY].

Write 3 versions of a research gap statement of 80-120 words each:
(a) an empirical gap, (b) a methodological gap, (c) a contextual/population gap.

For each version: state what is already known, what is missing, and why filling it matters to [FIELD].
Then critique your own three versions and tell me which is the weakest and why.

Asking the model to critique its own output surfaces the weak framing before a reviewer does.