You get:
- noticing contradictions but not explaining them
- no way to decide which source to trust
- literature reviews that list sources instead of synthesizing
- meetings where people cite different sources and talk past each other
- decisions based on the most recent source read, not the best evidence
But comparison reveals truth:
- consensus: where most sources agree (high confidence)
- conflict: where sources disagree (needs investigation)
- methodology differences: why findings differ
- bias patterns: which sources consistently favor certain conclusions
- gaps: what no source has studied
Without comparison, you have opinions, not evidence.
This prompt creates a structured comparison matrix across multiple sources.
Assume the role of a research synthesizer who compares multiple sources. Your task is to create a comparison matrix across sources on the same topic. Generate: 1. COMPARISON MATRIX (table format) | Dimension | Source 1 | Source 2 | Source 3 | |-----------|----------|----------|----------| | Authors/Year | | | | | Methodology | | | | | Sample/Population | | | | | Key Finding | | | | | Effect Size | | | | | Limitations | | | | | Funding/Sponsor | | | | 2. CONSENSUS AREAS - What do all (or most) sources agree on? - Confidence level (High/Medium/Low) 3. CONFLICT AREAS - Where do sources disagree? - Potential reasons for disagreement (methods? samples? bias?) 4. TRUST ASSESSMENT (per source) - Source 1: Highly trustworthy / Trust with caveats / Low trust - Rationale 5. RECOMMENDED CITATION - Which source(s) to cite for which claim INPUTS: Source 1 (summary or full text): [PASTE] Source 2: [PASTE] Source 3: [PASTE] Topic/question: [E.G., "Does remote work reduce productivity?"] Your decision context: [E.G., "Deciding our return-to-office policy"] RULES: - Include all sources even if they disagree strongly - Flag when sources measure the same thing differently - Note effect sizes, not just direction (who found a bigger effect?) - Distinguish between peer-reviewed and gray literature - Be explicit about your own bias in interpreting the comparison
- Use this for any decision that depends on multiple sources of evidence.
- Include sources that disagree with your hypothesis — that’s where learning happens.
- Pay attention to methodology differences — they often explain apparent contradictions.
- Look for consensus areas first — these are your high-confidence claims.
- Share the matrix with stakeholders before decisions — it builds trust in your process.
Topic/question:
“Does daily standup meeting improve team productivity?”
Source 1:
“Agile study 2023: Survey of 500 dev teams, self-reported productivity. 70% said standups help. No objective metrics.”
Source 2:
“Academic study 2024: Time-tracking of 50 teams. Standups correlate with 5% productivity increase (p<.05) but also 8% time spent in meetings."
Source 3:
“Basecamp research: Argues standups are status updates, not collaboration. Recommends async check-ins instead. No data, just opinion.”
This framework improves outcomes by forcing:
- structured comparison (apples to apples across dimensions)
- consensus identification (where the field agrees)
- conflict explanation (why sources disagree)
- trust assessment (not all sources are equal)
- citation guidance (which source for which claim)
Great source comparison doesn’t just list what each says — it tells you what to believe.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.