A research synthesis connects findings across a defined source set while preserving each source’s method, audience, limitations and unanswered questions.
You review the facts and access. Nothing is published without your approval.
The method behind the research synthesisAsk your agent to compare evidence in context, keep incompatible counts separate and label proposed interpretations as proposals.
How can an AI agent help synthesize research responsibly?
Give your agent an authorized source set and a specific question. Ask it to map observations to findings while retaining each source's method, audience and limitations. A reviewed HTML synthesis can show agreement, disagreement and next questions without pretending a selected bundle is a complete evidence base.
What to give your agent
A bounded research question and the full authorized source set selected for it.
Source identifiers, dates, methods, participant or dataset context and supplied quality limitations.
Observations and quotations permitted for this audience; separate researcher interpretations.
Contradictory findings, missing evidence and the decision this synthesis may inform.
Define the boundary of the evidence
State which sources were supplied and why they are relevant. Distinguish a synthesis of selected notes from a systematic review. Do not imply your agent searched every study or accessed research it was not given permission to use.
Keep observations connected to their sources
Give each claim a source identifier and retain the method and context needed to interpret it. Repeated mentions can come from the same participant or event; do not count them as independent evidence without checking.
Explain differences before combining results
Different tasks, samples and measurements can produce apparently conflicting findings. Put those differences beside the comparison. Do not pool percentages, average incompatible scales or turn qualitative notes into prevalence estimates.
Separate what you learned from what to do next
Use HTML to show the finding, supporting and conflicting evidence, limitation and next question. Label any proposed action as interpretation, not a study result. Review privacy and every source link before sharing; keep raw participant data in the authorized research system.
Synthesize only the authorized research sources I provide around [question]. Ask for missing source context or scope. State the source-set boundary. Map each observation and finding to a source identifier, method, audience and limitation. Preserve disagreement, unknown participant overlap and missing evidence. Do not pool incompatible counts, invent prevalence, claim a systematic review or stronger causal conclusion, or access unapproved material. Separate source findings from your proposed interpretations and next questions. Exclude identifying participant data and use only approved excerpts. Create readable static HTML with no scripts or forms and my saved style if available. Show me the draft and source checks. Do not publish or change access without separate explicit approval.
Check before you shareAccuracy, clarity, and permission
Is the selected source-set boundary stated?
Can every finding be traced to its source and context?
Are disagreements and incompatible measures preserved?
Are proposed actions separate from evidence and raw participant information excluded?
When HTML fits
A cross-source reading page for a bounded product or service question, with evidence and limits visible beside the synthesis.
When another format fits better
Do not substitute this for a systematic review or qualified medical, legal or other high-stakes evidence assessment. Keep raw research, consent records and analysis tools in their authorized systems.
What we started withThe same facts, as a wiki page
Fictional team wiki
Birch Reports — what to investigate next
Question and scope
Snapshot September 7, 2026. Owner: Ash. Question: what should we investigate to make reports easier to use? This covers selected sources A and B only, not an exhaustive search or systematic review.
Source A — First-time interpretation
A: in a September 2 session, three of five first-time volunteers asked what a label meant while interpreting a synthetic sample report. This small convenience group is not representative of all users. Timing, completion and satisfaction were not measured in the supplied record.
Source B — Finding saved work
B: five of eight selected support notes from September 1–5 mention difficulty finding a saved report. These concern existing-work retrieval, not first-time interpretation. Selection was not random, and note counts are not unique-user counts.
Interpretation
Overlap in participants or events is unknown. Tasks, methods and audiences differ; combining counts would not establish a population rate. Ash proposes separate investigation of newcomer terminology and existing-user retrieval. This is a proposed direction, not a proven redesign or universal main problem.
Decision and privacy
No redesign, staffing or research date is approved. The product team decides in its research planning process. No participant identities, verbatim quotes or private links are supplied.
Inspect the HTMLOne self-contained file
The example is sanitized and isolated from this site. No scripts, forms, or live data.
<!doctype html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>Birch Reports · Research synthesis</title><style>
*{box-sizing:border-box}html{color-scheme:light}body{margin:0;background:#f5f8fc;color:#182e46;font:17px/1.65 system-ui,sans-serif}main{max-width:1000px;margin:auto;padding:clamp(20px,5vw,64px)}h1,h2,h3,p{margin:0 0 18px}h1{font:500 clamp(32px,5vw,58px)/1.14 Georgia,serif;letter-spacing:-.025em;overflow-wrap:anywhere}h2{font:500 29px/1.3 Georgia,serif;overflow-wrap:anywhere}h3{font-size:19px;line-height:1.4}.label{font:700 12px/1.6 system-ui,sans-serif;letter-spacing:.08em;text-transform:uppercase}.intro{font-size:20px}.lead{background:#23486b;color:white;padding:24px;margin:28px 0}.lead p:last-child{margin:0}.grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(min(100%,230px),1fr));gap:20px;margin:24px 0}.card{background:white;border:1px solid #bbc5c4;padding:24px}.metric{font:500 38px/1.2 Georgia,serif;display:block;margin-bottom:12px}section{margin:36px 0}ul,ol{padding-left:22px}li{margin-bottom:12px}.note{font-size:14px}.warning{border-left:5px solid #956118;padding-left:20px}a{color:inherit;text-underline-offset:4px}a:focus-visible,summary:focus-visible{outline:3px solid #956118;outline-offset:4px}table{width:100%;border-collapse:collapse;table-layout:fixed;font-size:15px}th,td{text-align:left;vertical-align:top;border-bottom:1px solid #bbc5c4;padding:12px 8px;overflow-wrap:anywhere}caption{text-align:left;font-weight:700;padding:12px 0}pre,code{white-space:pre-wrap;overflow-wrap:anywhere}footer{border-top:1px solid #bbc5c4;padding-top:24px;margin-top:40px;font-size:14px}@media(max-width:480px){th,td{padding:10px 4px;font-size:14px}}@media print{body{background:white}.card,.lead{break-inside:avoid}h2,h3{break-after:avoid}}
</style></head><body><main><header><p class="label">Fictional two-source synthesis / September 7, 2026 / Owner Ash</p><h1>Two moments of confusion. Two questions to test.</h1><p class="intro">What should Birch Reports investigate next? The supplied evidence points to different tasks—not one proven universal problem.</p></header><section class="lead"><h2>Proposed direction, not an approved redesign</h2><p>Investigate terminology with newcomers and report retrieval with existing users separately. No staffing or research date is approved.</p></section><section><h2>Keep each finding beside its context</h2><div class="grid"><article class="card"><p class="label">Source A / September 2</p><h3>Interpreting a first report</h3><p>Three of five first-time volunteers asked what a report label meant while using a synthetic sample.</p><p class="note">Small convenience group; not representative. No timing, completion-rate or satisfaction measure supplied.</p></article><article class="card"><p class="label">Source B / September 1–5</p><h3>Finding saved work</h3><p>Five of eight selected support notes from existing users mention finding a saved report.</p><p class="note">Selected notes, not random sampling or unique-user counts. Different task from Source A.</p></article></div></section><section class="warning"><h2>Do not combine these into a rate.</h2><p>Methods, audiences and tasks differ. Participant or event overlap is unknown. Adding the counts would not establish how common either problem is across the customer population.</p></section><section><h2>What this synthesis leaves open</h2><p>Whether clearer terminology or changed retrieval would help needs further investigation. The product team must decide the next research step in its existing planning process.</p><p class="note">Sources A and B are the entire supplied set. This is not a systematic review. No participant identities or private source links are included.</p></section><footer>Fictional worked example by share/artifacts. Names, dates, figures and records are invented for demonstration. Static snapshot; no live data or automatic updates.</footer></main></body></html>
The original worked scenarioThe source scenario, step by step
Two sources, two different moments of confusion
Fictional non-sensitive usability notes, not population research.
The original notes
Source A: five first-time volunteers tried a sample report; three asked what a label meant. Source B: eight selected support notes from existing users; five mention finding a saved report. Participants may overlap and this is unknown. Methods, tasks and audiences differ. Neither source is representative. Proposed interpretation: test terminology with newcomers and retrieval with existing users rather than claim one universal main problem. No combined rate or approved redesign.
Source A
Five first-time volunteers used a sample report; three asked about a label.
Source B
Eight selected existing-user support notes; five mention finding a saved report.
Comparison limit
Methods, tasks and audiences differ; participant overlap is unknown and neither source is representative. No combined rate.
Next question
Proposed: test newcomer terminology and existing-user retrieval separately. No redesign is approved.
Preserving the differences produces a better next question than forcing one universal conclusion.
Questions, adaptations & sourcesGo deeper when you need to
How is this different from explaining one paper?
A paper explainer preserves one study's argument and boundaries. A synthesis compares a selected set and must also explain source selection, conflicting evidence and differences in methods.
Can I count mentions across interviews?
Only with a verified definition and deduplication method appropriate to your data. A mention count is not automatically a count of people or a population percentage.
What if sources disagree?
Show the disagreement and relevant context. It may identify separate audiences or a question to investigate; do not hide it to produce a confident headline.