Yes. Copy the manual template and restart prompt, fill them in yourself, and save them with your original research notes.
Free manual template
Carry your research evidence into the next AI chat
Carry your question, sources and open gaps into a fresh chat. Keep checked evidence separate from leads you still need to verify.
Use the free templateGet 1 free summary
No account needed for the template. Check sources yourself; a handover does not verify them.
Keep an evidence trail, not just a conclusion
A research restart needs more than the latest answer. Record the scope, the excerpt behind each claim, what you actually checked, and why other findings were dropped. Keep your original notes and saved source material alongside the handover.
This is a manual workflow. Neither a citation nor an AI-written handover proves accuracy, and a new chat may be unable to open your links. Provide the relevant excerpts and ask it to identify missing access before continuing.
State one research question and its boundaries.
Save source URLs, locations and short excerpts; label full-text access versus snippet-only leads.
Check each claim against the source and record the specific check. Keep corrections and rejected findings visible.
Carry unresolved questions and one next search into the new chat.
Copy the free research handover template
Copy this, then replace the prompts with the details from your conversation.
RESEARCH HANDOVER Question: [the specific question you are trying to answer] Scope: [population, geography, dates, definitions and exclusions] Decision this informs: [what you will do with the answer] Checkpoint: [date, notes version and person who checked it] Original notes: [file/link; keep a separate saved copy] SOURCE REGISTER S01 | [title, author/publisher, publication date] URL: [original source URL, not only a search-results link] Location: [page, section, table or paragraph] Exact excerpt: [short relevant quotation; preserve units and caveats] Access: [full text opened / excerpt only / inaccessible] Checked by / on: [person, date; or unchecked] CLAIM LEDGER — one claim per entry C01: [precise claim with scope, units and time period] Source: [S01 + URL + location + supporting excerpt] Check state: [checked against source / unchecked / contradicted] What was checked: [exact check and by whom; do not just write verified] Limit: [what this evidence does not establish] Keep source statements separate from your interpretation. REJECTED OR SUPERSEDED FINDINGS C00: [old claim] | [rejected / superseded by C01] Reason: [contradiction, scope mismatch, duplicate or correction] Evidence: [source URL + excerpt or correction note] Do not reuse this finding as a current conclusion. UNRESOLVED QUESTIONS [gap, conflicting source or missing evidence] Next search: [query + preferred source + what would resolve the gap] Stop condition: [what is enough to answer the research question] NEXT SESSION First action: [one bounded search or source check] Needs access: [missing files/pages; ask rather than assume] Do not treat unchecked claims as established facts. Preserve the scope and uncertainty; record new findings separately.
What checked and unchecked mean here
Checked means a named person compared a specific claim with the relevant source material and described the check. It does not mean every statement in the source is correct, current, representative or independently confirmed.
Unchecked means the claim still needs a source check. A search snippet may help you find a document, but it does not establish its methods or caveats. If a source is inaccessible, save the lead and ask for the excerpt rather than claiming verification.
Label a corrected claim superseded and point to its replacement. Reject a finding when it fails your question or scope, while keeping the reason. Separate what a source says from the interpretation you draw from it.
Original fictional research notes
T01: Researcher: Notebook r2 asks whether the fictional Northbank Library should extend Saturday opening for a three-month pilot. Scope: Saturday visits at Northbank only; exclude weekday traffic and other branches.
T02: Researcher: S01 is the invented Northbank pilot report, https://example.org/northbank/pilot-r1, table 2. Exact excerpt: "Saturday visits rose from 80 to 120 per day during the four-week pilot." I compared the r1 excerpt with C00: Saturday visits rose by 50%. This is a before/after count, not evidence that the new hours caused the change.
T03: Researcher: S02 is an invented correction, https://example.org/northbank/correction-r2, table 2 note. Exact excerpt: "Corrected Saturday baseline: 100 visits per day; pilot count unchanged at 120." For r2 I checked the supplied correction and calculated (120 - 100) / 100 = 20%. C01 replaces C00; the old 50% finding is superseded.
T04: Researcher: S03 is an invented search snippet, https://example.org/northbank/survey, summary. Exact excerpt: "Most respondents prefer later Saturday opening." For r2 the full survey was not opened; sample size and question wording remain unchecked. C02 is only a snippet lead, not a finding about all library users.
T05: Researcher: S04 is an invented Eastbank report, https://example.org/eastbank/annual, visits section. Exact excerpt: "Weekday visits increased by 30%." C03 was rejected for this question: wrong branch and weekdays, outside the Northbank Saturday scope.
T06: Researcher: Notebook r2 has no staffing-cost figures or control comparison. A recommendation is unresolved. Next search: Northbank Saturday pilot staffing cost report; seek a dated primary budget or rota with the same pilot period. Then request the full S03 survey and check its sample and wording. Do not infer causation from attendance counts.
Worked example: a library opening-hours study
Northbank, Eastbank, the reports and every excerpt are invented for this example. The example.org URLs are placeholders. The checks described are actions by a fictional researcher against supplied excerpts, not automated verification or actual browsing.
Copy this, then replace the prompts with the details from your conversation.
NORTHBANK LIBRARY — fictional worked handover, notebook r2 Question: Should Northbank extend Saturday opening for a three-month pilot? Scope: Northbank Saturday visits only. Exclude weekdays and other branches. Decision: Still unresolved; attendance alone is insufficient. Source boundary: All reports, excerpts, notes and checks below are invented. example.org URLs are illustrative placeholders, not real reports to verify. The fictional researcher checked supplied excerpts, not live web pages. C01 | Checked against the supplied correction in this example Claim: The corrected before/after Saturday count rises by 20%, from 100 to 120 daily visits. S02: https://example.org/northbank/correction-r2 | table 2 note Excerpt: "Corrected Saturday baseline: 100 visits per day; pilot count unchanged at 120." Check: Researcher used the r2 correction, then calculated (120 - 100) / 100. Limit: No control comparison; this does not establish a causal effect of longer hours. C02 | Unchecked lead S03: https://example.org/northbank/survey | search snippet, summary Excerpt: "Most respondents prefer later Saturday opening." Access: Full survey not opened. Sample size and question wording unknown. Do not generalize to all library users or use it as a verified conclusion. C00 | Superseded by C01 Old finding: 50% rise, based on 80 to 120 visits in S01. S01: https://example.org/northbank/pilot-r1 | table 2 Excerpt: "Saturday visits rose from 80 to 120 per day during the four-week pilot." Reason: S02 corrects the baseline to 100. Keep the old calculation for history only. C03 | Rejected for scope mismatch S04: https://example.org/eastbank/annual | visits section Excerpt: "Weekday visits increased by 30%." Reason: Eastbank weekdays cannot answer the Northbank Saturday question. Unresolved: Staffing cost, survey quality and alternative explanations for attendance. First search: "Northbank Saturday pilot staffing cost report". Preferred evidence: Dated primary budget or rota for the same pilot period. Next: Obtain full S03 survey; inspect sample size, wording and respondent selection. Stop condition: Compare same-period staffing cost and attendance; assess survey limits. If evidence is missing, report the gap before recommending a pilot.
Check the example before you reuse the pattern
The current 20% calculation cites the corrected baseline of 100. The original 50% calculation remains superseded, even though it was correct for the earlier numbers.
The survey is still a snippet lead. Its sample and wording have not been checked.
The Eastbank weekday figure is rejected for scope mismatch, not declared false.
No cost figures or control comparison are supplied. The handover leaves the recommendation and causation unresolved.
Copy the research restart prompt
Copy this, then replace the prompts with the details from your conversation.
Continue the research using the handover below and the supplied source excerpts. First restate the question, scope and first action. List which sources you can actually access. Keep source statements, interpretations and unresolved questions separate. For every new factual claim, record its source URL, location and supporting excerpt. Describe what you checked; a URL or citation alone is not a source check. Do not claim to have opened or verified a source you cannot access. Ask for the relevant excerpt. Keep unchecked, contradicted, rejected and superseded findings labeled. Do not promote them into conclusions. Resolve the first gap before widening the scope, and report remaining uncertainty. [Paste the handover and relevant source excerpts here]
1 free summary, no card required
Sign up, paste your conversation into the web app, and create your first real summary. Review the handover before using it in a new chat. One free summary per account covers 5,000–500,000 characters of conversation text.