Building a literature review or state of the art with AI

Context

A university researcher or academic who needs to build or update a literature review for an article, research grant application, thesis, or research report. The prompt takes a corpus of already-collected abstracts — between 10 and 40 summaries — and generates a thematic map organized by conceptual clusters, identifies knowledge gaps, and locates the most central sources in the debate. It is useful at the start of a research project or before writing the literature review section, when the database search has already been done but the full corpus has not yet been read.

Originally written for Chilean classrooms.

Paste first

Before opening the model, prepare:

• The abstracts or summaries of the sources you already have (10–40 abstracts pasted as a block, separated by a blank line; do not paste complete PDFs or raw field data)
• Your research question in one sentence
• The disciplinary field and type of review you are doing (systematic, narrative, integrative)
• The publication period covered and the bibliographic databases you consulted (Scopus, WoS, SciELO, ERIC, etc.)
• If you already have your own preliminary thematic axes, include them at the end of the message so the model can use them as a reference

Privacy rule: Paste only abstracts from published texts. If any source is an internal report, project report, or thesis containing non-anonymized data from human participants, do not include it. Research data — interviews, surveys, field records — must never be pasted into an external model (applicable personal data protection law).

You are an academic research assistant specialized in literature reviews. I am going to share a set of abstracts from articles, chapters, or reports on a topic. Your task is to help me build a preliminary state of the art: you do not replace my deep reading of the sources, but you do help me organize and synthesize the corpus before that reading.

**My research question:**
{{research_question}}

**Disciplinary field and type of review:**
{{field}} — {{review_type}} review

**Period covered and bibliographic databases consulted:**
{{period}} / {{databases}}

**Corpus abstracts (published texts only, no identifiable participant data):**

---
{{corpus_abstracts}}
---

Please develop the following:

1. **Thematic map:** Identify between 4 and 7 clusters or nodes that emerge from the corpus. For each cluster: assign a precise label (not generic), describe its scope in 2–3 sentences, name the authors who anchor it most strongly according to the abstracts, and note whether there are tensions or disputed positions within the cluster. Respond in English even if the abstracts are in another language; you may keep technical terms that have no precise equivalent.

2. **Knowledge gaps:** Identify 2–3 questions or areas that the corpus does not answer, or answers insufficiently. Distinguish between empirical gaps (missing data or unstudied contexts), theoretical gaps (missing explanatory frameworks or contested concepts), and methodological gaps (limitations in how the problem has been studied).

3. **Central sources:** List the 5–8 authors or texts that appear most referenced or central to the conversation, based on the abstracts themselves. For each, indicate why they are essential and which thematic cluster they belong to.

4. **Chronological trajectory:** Briefly describe how the debate has evolved over the period covered. Is there a visible conceptual shift? What conversation was active at the start of the period and what is happening now?

5. **Suggested positioning:** Based on the gaps identified, propose a tentative positioning statement for my research: where does my work sit in relation to the existing corpus, and what gap is it seeking to fill?

Work only with the information I provide. If you need to invoke general field knowledge to contextualize something, precede it with [General context — outside the corpus] so I can distinguish it from direct analysis of the abstracts. If an abstract is not sufficient to categorize a source with certainty, note this and add it to a "to be confirmed with full reading" category. Do not invent citations, authors, or bibliographic data that are not in the corpus you were given.

Expected output

Example (condensed):

1. Thematic map (excerpt — Cluster 1)

Differentiation of feedback modalities and effects on metacognition. The only quasi-experimental study in the corpus compares dialogic vs. written feedback in students at a research university system and reports a moderate effect (d = 0.47) on effort regulation and comprehension monitoring. Internal tension: dialogic feedback produces superior effects, but its scalability in large-enrollment settings is problematic; the text raises this question without resolving it.

2. Main gap (empirical): Only 12% of the 47 systematized studies measure uptake as a behavioral outcome variable; the majority describe practices or student perception. This is the structural gap in the existing production.

5. Suggested positioning:

This study sits at the intersection of the absence of uptake measurement as a behavioral variable and the scarcity of evidence on formative feedback models with differential effects on metacognitive strategies in large-enrollment universities — an institutional context that existing qualitative research in the field does not yet cover.

Watch out for

• The model can infer or generate bibliographic data not present in your abstracts. If an article appears that you do not recognize, or if it attributes something to an author that you do not see in the corpus, be skeptical and verify before including it in your review. This is the most common risk with this type of prompt.
• The thematic clusters are working hypotheses, not a definitive structure. A cluster that looks coherent in the abstracts may fragment or merge when you read the full texts. Treat the map as provisional scaffolding, not as a result.
• If your corpus is predominantly in English, the map will tend to reflect the Anglophone literature. Explicitly ask it to identify whether there are voices from Latin America, Africa, or other underrepresented traditions in what you pasted, especially if your research has a regional scope.

Suggested iteration

If the clusters overlap or there are too many: "Clusters 3 and 5 seem to describe the same thing. Merge them, redefine the resulting label, and redistribute the sources among the remaining clusters."

If you want to draft the section directly: "Take the thematic map and write a draft literature review in continuous prose of 800–1,000 words, using the clusters as the structure. Use the abstracts as support and APA 7 format for in-text citations. Mark with [VERIFY] any data you cannot confirm from the abstracts alone."

If you want to prepare a progress presentation for your research group or committee: "Summarize the map as a plain-text outline with bullets and indentation, ready to convert into a slide: cluster name, 2–3 representative sources per cluster, and the main gap for each."