The short answer
Choose an evidence-synthesis toolchain around the answer you need to produce: a thematic overview, a structured review, quantitative estimates, or an updated conclusion. Then define the source IDs, checked fields, decisions, and versions that must pass between tools. This guide focuses on those handoffs and on how a source correction reaches a reviewed update; the linked task guides cover detailed discovery, screening, and analysis comparisons.
Key takeaways
Choose the synthesis purpose and output before selecting the tools that will produce it.
Give sources, extractions, analyses, and claims stable identifiers so changes can be followed across tools.
A new paper or correction triggers reassessment. Release a new conclusion only after the affected records and decisions have been reviewed.
Gatsbi publishes this guide and makes some of the tools discussed. We use the same task-based criteria for Gatsbi and other products, drawing on their public documentation. These are not hands-on benchmark results; check current features and access before choosing.
Choose the synthesis output you need
Evidence synthesis brings findings from existing research together to answer a question. Exploring vocabulary in school-attendance research, mapping the kinds of evidence available, and estimating an attendance intervention's effect lead to different outputs and methods.
Write the intended use in one sentence before choosing software. “We need themes to plan a study” leads to a different workflow from “We need an intervention review for a policy decision.” The routes below identify the next task; this page then explains how to connect the resulting records and maintain them.
Method-specific guidance still applies. JBI's scoping-review guidance covers systematic mapping of concepts and evidence, while Cochrane's other synthesis methods chapter discusses approaches when a conventional meta-analysis is unsuitable. A tool-generated map or narrative does not by itself establish that either method was followed.
| Purpose | Intended output | Method or control to establish | Next task |
|---|---|---|---|
Explore a topic | Source-linked reading notes and themes | Search scope and source support for the claims you use | |
Map concepts and evidence types systematically | Charted evidence with explicit boundaries | A scoping-review question, eligibility, charting, and reporting plan | Use method-specific guidance, then specify the handoffs below |
Understand experiences or mechanisms | Contextual interpretations and synthesis findings | A suitable qualitative method and retained source context | Specify source-to-interpretation records; this page does not compare all qualitative tools |
Assess an intervention | Structured findings and, where suitable, pooled estimates | Selection, appraisal, synthesis plan, and compatible analysis data | Screening and extraction, then quantitative analysis if needed |
Maintain an existing conclusion | A reviewed new version or a recorded decision to retain it | Monitoring trigger, affected records, reassessment, and release authority | Use the change log and approval example below |
Build the smallest useful chain of outputs
Select components by the records they exchange. A discovery assistant may supply candidate sources; a reference library maintains their identity; an extraction workflow records checked fields; an analysis tool consumes those fields. A team can combine separate tools or use an integrated product, provided the necessary records remain available at the boundaries.
The examples below describe configurations to evaluate, not tested integrations. Gatsbi's product descriptions, reviewed on 18 October 2026, place Deep Research in exploratory source collection and drafting, and Reviewer in structured review and analysis. They do not establish cross-platform approval, change propagation, or a continuously updated review system.
Detailed vendor comparisons belong to the task guides linked above. Here, ask one narrower question at each boundary: can the next person identify the input, see its checked status, and understand what decision is still open? A missing identifier or decision can matter more than whether two products offer a direct integration.
| Intended use | Components to evaluate | Handoff output | Acceptance check |
|---|---|---|---|
Explore a field | Discovery assistance, such as Gatsbi Deep Research, plus a reference library and thematic matrix | Sources, access limits, checked notes, and provisional themes | A reader can locate the source behind a theme and distinguish a study result from the reviewer's interpretation |
Complete a quantitative review | Screening records, an extraction ledger, and statistical software; or a suitable integrated workflow such as Gatsbi Reviewer | Study/report IDs, verified analysis fields, selected settings, and results | Required reviewer decisions and analysis records remain inspectable, whether in one product or companion records |
Maintain a published synthesis | A monitoring source, change log, reassessment process, and versioned release record | New or corrected evidence, affected records and claims, and an approved decision | An alert cannot change the released conclusion without the specified checks and approval |
Define the records that must survive each handoff
Use separate IDs for a study, a report, an extracted item, an analysis, and a claim. Two reports can describe one study, while a single report can contribute several outcomes. Keep the mapping when a record moves from a source library to a table, script, or manuscript.
The fictional attendance example below uses Study A, Report A, extraction E-01, the 12-week synthesis group S-12W, and claim C-01. In version 1, E-01 records attendance at 12 weeks. Claim C-01 reads: “The available attendance evidence includes a 12-week follow-up.” Its supporting record points to S-12W and E-01. These invented labels and wording make the dependencies visible; they do not describe real research.
Source checking establishes what a passage supports; the literature-review guide's claim check shows that task in detail. The handoff adds the checked status, responsible person, and destination. A reference list alone does not convey those decisions.
Copyable handoff fields: handoff_id, from_task, to_task, required_id, required_artifact, producer, approver, fail_action, and status.
Keep a source version or retrieval date, source location, checked value or interpretation, and reviewer alongside the ID mapping.
If a receiving tool cannot retain a field, maintain a companion register keyed to the same IDs. Do not claim an automatic integration unless you have tested it.
| From → to | Record to pass | Who checks it | If the check fails |
|---|---|---|---|
Source library → extraction | Study A / Report A v1; source location and report-to-study mapping | Extractor confirms identity and access to the relevant passage | Resolve the source or study identity before accepting E-01 |
Extraction → synthesis | E-01 v1; attendance outcome; recorded follow-up 12 weeks; source and reviewer | Synthesis reviewer checks the field and membership of S-12W | Keep the item pending; do not silently recode its follow-up |
Synthesis → claim | S-12W v1; included records, grouping decision, limitations, and link to C-01 | Claim reviewer checks whether the proposed wording follows from the synthesis | Revise the claim or return the disputed grouping for review |
Reviewed claim → release | C-01 v1; supporting synthesis version, approval, and unresolved issues | Review lead approves the version for its stated use | Retain draft status and assign the unresolved issue |
Trace a source correction to the affected decisions
Define what triggers reassessment: a planned search date, a potentially eligible study, a correction or retraction, or a change to the question or method. Record the monitoring source and responsible person. The alert date, search date, and substantive review date may differ.
Return to the fictional attendance example. Report A is corrected: its outcome was measured at 6 weeks, not the recorded 12 weeks. The change reaches E-01, membership of S-12W, and the part of C-01 that uses that follow-up evidence. Keep both source versions so the reason for reassessment remains visible.
One correction, several decisions to revisit
A fictional correction changes Report A’s follow-up from 12 to 6 weeks. It branches to extraction, synthesis grouping and claims, then joins at human review before an approved version is released.
Fictional teaching example
Report A corrects the follow-up time.
- Extraction
Check the time-point field and linked values.
- Synthesis
Recheck membership in the 12-week group.
- Claim
Reassess any statement about 12-week evidence.
Confirm what changed and which decisions depend on it.
The conclusion may change—or remain the same with a recorded reason.
Create one change entry, then identify its dependent records. The table shows pending review actions, not a completed reanalysis. Removing Report A from a 12-week group does not by itself tell you what the remaining evidence supports, or whether it belongs in a separate six-week synthesis.
After the checks, record whether the conclusion changed and why. A justified decision to retain the conclusion is also an update outcome. Release the approved version with its change record; do not replace the old conclusion or refresh the displayed review date merely because an alert arrived.
Copyable change-log fields: change_id, source_id, old_source_version, new_source_version, field, old_value, new_value, affected_extraction_ids, affected_analysis_ids, affected_claim_ids, owner, decision, rationale, and release_version.
For this example, use change U-01; Report A v1 → v2; follow-up 12 → 6 weeks; affected E-01, S-12W, and C-01. Decision and release version remain pending until review.
| Affected record | Old → corrected information | Required reassessment | State before approval |
|---|---|---|---|
Report A | v1: attendance at 12 weeks → v2: attendance at 6 weeks | Confirm the correction and preserve both versions and source location | Correction recorded; dependent records flagged |
E-01 | Time-point field 12 → 6 weeks | Check the corrected field and whether linked values also changed | Revised extraction awaiting its designated check |
S-12W | Report A previously contributed to the 12-week group | Reassess membership and any output using that group; separately consider other time-point groups | Synthesis update pending |
C-01 | Wording relied on S-12W v1 | Review support and scope against the revised synthesis; document any unchanged conclusion | Claim review pending; no new release approved |
Assign the check and the decision after each handoff
Name the person responsible for producing a record and the person authorized to accept it. These may be the same or different people depending on the method and team, but the required checks must be explicit. “Human review” alone does not tell a colleague who should resolve a disputed field.
Consider Team A, which explores attendance research to plan a study, and Team B, which prepares a formal intervention review for a policy decision. Team A needs a clearly bounded, source-checked overview. Team B also needs the protocol's selection, appraisal, and synthesis decisions. The table shows how responsibilities change with the intended use without prescribing one staffing model for every project.
Give people access to the material needed for their assigned check, subject to the institution's permissions and data terms. If a source, software setting, or result cannot be inspected, record who will resolve that gap and keep the affected decision pending.
| Action | Team A: exploratory overview | Team B: formal review | Failure route and retained record |
|---|---|---|---|
Accept source collection | Research lead checks relevance and stated coverage limits | Search lead checks the planned sources and search record | Return missing coverage information to the collector; retain queries, sources, dates, and rationale |
Accept extraction or interpretation | Analyst checks the passage used in a theme | Assigned reviewers verify the fields required for synthesis | Return the disputed item to its checker; preserve source, proposed value, correction, and reviewer |
Reassess U-01 | Lead revisits the affected theme and its limits | Extraction, synthesis, and claim owners review E-01, S-12W, and C-01 | Keep each dependent decision pending until resolved; retain the change log |
Approve release | Lead approves the bounded overview and open questions | Authorized review lead approves methods, outputs, and resolved changes | Return unresolved items to their owners; preserve approval, limitations, and release version |
Rehearse the handoff before expanding the workflow
Give a colleague C-01 and the retained records, without the original AI conversation. Ask them to identify its supporting synthesis, find E-01 and Report A, and explain the effect of correction U-01. They should be able to identify both what changed and which decisions remain open.
For a structured review, Gatsbi Reviewer can be a candidate for connecting study selection, extraction, analysis, and a draft, as described on its product page. A proposed rehearsal is to review those outputs, inspect the available export, and enter the relevant IDs and checks in your team's handoff register. This guide has not established that Reviewer itself supplies the approval or update-dependency records above.
Use the rehearsal to locate missing handoffs. Repair a broken source mapping, an unassigned correction, or an unavailable analysis setting before expanding the process. Keep the conclusion in draft if the required evidence or approval is still missing.
Reconstruct
Start at C-01 and locate the supporting synthesis version, extraction E-01, source Report A, and their checked status.
Follow the change
Locate U-01, compare the 12-week and 6-week records, and identify the responsible reviewer for each affected decision.
Decide what may be released
Find the approved version and rationale, or identify the unresolved item and owner. Do not treat the presence of a new draft as approval.
Common questions
How is evidence synthesis different from meta-analysis?
Evidence synthesis includes methods for bringing research findings together. Meta-analysis statistically combines suitable results and can be one component. Your intended question determines whether the output should be a quantitative estimate, a thematic account, an evidence map, or another synthesis.
Does using one integrated product remove the need for handoff records?
No. People still need to distinguish source material, checked extraction, analysis decisions, and approved claims. Check which records the product preserves and exports, then maintain any necessary companion register using the same IDs.
What should happen when an alert finds a new study or correction?
Record the trigger, identify the affected sources and downstream records, assign reassessment, and document the decision. A new alert is not itself an approved change to the synthesis.
Can an update leave the conclusion unchanged?
Yes. The review may find that the corrected or added evidence does not change the conclusion. Retain the checks, rationale, responsible reviewer, and version so readers can distinguish a reviewed decision from an ignored alert.
Sources & further reading
Follow the original sources for their complete methods, scope, and latest updates.
- Scoping reviewsJBI Manual for Evidence Synthesis
- Synthesis using other methodsCochrane
- Gatsbi Reviewer product workflowGatsbi · publisher product documentation
- Gatsbi Writer and Deep Research product descriptionGatsbi · publisher product documentation