Join for free

Synthesize research findings with explicit limitations

Author: AILesson9 min setupTested with:ChatGPTReviewed: 2026-08-28

Quick answer

Combine findings into calibrated conclusions without erasing uncertainty, heterogeneity, or missing evidence. Provide: Structured findings and sources, Research and decision question, Synthesis standard. Expected result: A decision-oriented synthesis with confidence boundaries, exceptions, and next evidence needs.

1

Add your context

Your text stays in this browser. AILesson Prompts does not send it to a model or server.

2

Your prompt

Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt

Synthesize only the supplied findings for the stated decision.

Finding IDs, sources, methods, samples, estimates, uncertainty, and limitations:
[findings]

Population, issue/intervention, comparator, outcome, period, context, and decision:
[question]

Evidence hierarchy, materiality, acceptable inference, audience, format, and length:
[standard]

Check whether findings answer the same question before combining them. Preserve source IDs and distinguish direct evidence, indirect evidence, descriptive context, expert interpretation, and hypothesis. Do not average estimates unless measures, populations, comparisons, and methods justify it. Do not convert lack of statistical significance into no effect, association into causation, absence of evidence into evidence of absence, or repeated sources into independent confirmation.

Write: scoped answer; convergent findings; heterogeneity and exceptions; quality/transfer limits; evidence gaps; plausible alternative explanations; and calibrated decision implications. For every conclusion state confidence qualitatively with reasons, evidence that supports and weakens it, population/time boundary, and what new observation would change it. Separate “can decide now,” “can test now,” and “cannot conclude.” Include a source-to-conclusion ledger and a short plain-language version that retains the central limitation.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.

From input to outcome

A worked example

See how concrete context turns this recipe into a usable result

Actual input

Structured findings and sources
F1/C randomized working paper: 1,204 adults at four urban libraries; reminders 63%, control 58%, difference 5 percentage points, 95% CI 1-9; excludes no-mobile patrons; peer review unknown. F2/A local non-randomized pilot: 214 opted-in adults, 61% versus 54% prior period; seasonal mix differs; 11 opt-outs, 3 late messages, ~25 staff-min/day. F3/E interviews: 12 local patrons selected from pilot volunteers; 8 said texts were useful, 4 wanted channel choice; recruitment favors engaged users. F4/D rural descriptive dashboard: 47% collection among all ages, no intervention. F5/C secondary outcome: no clear difference in late fees, estimate +0.2 dollars, 95% CI -0.4 to +0.8.
Research and decision question
Should the city library run a four-week randomized reminder test among consenting adults, measuring collection before expiry and operational guardrails? Decision is testing, not rollout. Context is this library in late 2026.
Synthesis standard
Direct randomized evidence preferred; local feasibility evidence relevant. A 3-point collection difference is operationally material for planning, but uncertainty must remain. Executive memo, plain language, ≤450 words. Qualitative confidence only.

Example output

Scoped answer: the supplied evidence supports running a bounded local randomized test among consenting adults; it does not support permanent rollout or a promised effect size. Confidence is moderate because one sizable randomized urban study estimates a positive collection effect, while local evidence establishes feasibility but not causation.

Convergence: F1/C estimates a 5-percentage-point difference, with interval 1-9. Its point estimate exceeds the 3-point planning threshold, but the interval includes smaller effects. F2/A has a compatible 7-point descriptive difference and supplies local operational signals. Do not average 5 and 7: designs and comparisons differ. F3/E suggests acceptability and channel choice matter, but the engaged volunteer sample cannot estimate prevalence. F4/D is context only, not effect evidence. F5/C does not show a clear late-fee effect; its interval includes modest harm and benefit, so “no effect” would be wrong.

Limits and alternatives: transfer from four urban libraries is uncertain; both intervention findings omit or underrepresent adults without mobile access; local period/season changes could explain F2; message timing or participant engagement could contribute. Peer-review status is unknown. The evidence says little about children, rural patrons, permanent behavior, or equitable channel access.

Can decide now: authorize design of a four-week test with consent, random allocation, pre-defined collection outcome, channel/accessibility review, opt-out/late-message/staff-time guardrails, and stop thresholds. Can test now: whether the local effect clears 3 points without unacceptable guardrail cost and whether channel choice is needed. Cannot conclude: reminders caused the local pilot difference, reduce late fees, work for patrons without mobile numbers, or merit rollout. Change condition: a corrected F1 null estimate, infeasible local operations, or unacceptable opt-out/access disparity would weaken the recommendation; a well-run local test clearing threshold with stable guardrails would strengthen it. Ledger: conclusion effect→F1/F2; feasibility→F2; acceptability→F2/F3; limits→F1-F5.

Why this works

  1. 1

    Question alignment prevents a synthesis from combining estimates that only look similar

  2. 2

    Change conditions make confidence actionable by showing what evidence should alter the decision

Check the result

  • Are conclusions scoped to matching populations, measures, periods, and methods

  • Does every conclusion include support, weakening evidence, limits, and a change condition

  • Are current decisions, testable hypotheses, and unsupported conclusions separated

Evidence behind the method

Sources and quotations

Primary sources that support specific design choices in this recipe. They do not guarantee a particular AI result.

  1. The prompt separates findings, certainty, applicability, limitations, and decision implications because Cochrane guidance requires conclusions to reflect uncertainty and avoid going beyond the available evidence.

    Chapter 15: Interpreting results and drawing conclusions — Cochrane Handbook for Systematic Reviews of InterventionsSource reviewed:

Use it with confidence

Frequently asked questions

Practical answers about when to use this recipe, what to provide, and where human review still matters

What should I prepare before using “Synthesize research findings with explicit limitations”?

For “Synthesize research findings with explicit limitations,” prepare Structured findings and sources, Research and decision question, and Synthesis standard. Replace placeholders only with information you can verify. If a detail is unknown, preserve that uncertainty explicitly instead of asking the model to infer it.

When is the “Synthesize research findings with explicit limitations” result not ready to use?

The result is not ready if it does not yet deliver the stated outcome—A decision-oriented synthesis with confidence boundaries, exceptions, and next evidence needs—from the supplied evidence, or if it relies on unresolved assumptions, missing approvals, or invented details. Use the checks as release gates: revise the source inputs or assign a named, authorized reviewer instead of polishing an unsupported output.

Which AI tools have recorded tests for “Synthesize research findings with explicit limitations”?

The published test record for “Synthesize research findings with explicit limitations” lists ChatGPT as of 2026-08-28. This confirms recorded runs, not guaranteed compatibility or identical results in later product versions. For another tool or version, keep every constraint visible and repeat the result checks before use.

More ways to explore

Where this recipe fits

Text promptsResearch & analysis

Keep the work moving