🔬 Research
Elicit and Consensus Paper Map with Required Citation URLs
Map papers in Elicit or Consensus from sources that already include URLs. Every claim needs a citation URL. No invented DOIs.
0Reviews
Prompt
Act as a research librarian building a paper map for Elicit (elicit.com 2026) and Consensus (consensus.app 2026). You organize papers the user already listed with URLs. You do not invent DOIs, PMIDs, or findings. If a URL is missing, the paper goes in NO_URL and is not cited as evidence. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Question I want mapped: [Question] - Papers I already have (title, year, URL required): [Papers] - Tool I will paste into (Elicit, Consensus, both): [Tool] - Inclusion rules: [Include] - Exclusion rules: [Exclude] - Outcomes I care about: [Outcomes] - Words I must not use: [Banned] - Max papers in the map: [Cap] - Study types I may label (only if named in Papers): [Types] - What to do when a URL is missing: [MissingPolicy or NO_URL] Generate: 1. Tool lock: Elicit 2026 and/or Consensus 2026 from Tool. Not a PubMed scrape. 2. URL audit: every paper in Papers. Missing URL -> NO_URL, cannot support a claim. 3. Inclusion grid vs Include and Exclude. Drop papers that fail, with reason. 4. Map: question, paper, year, URL, outcome columns that Outcomes named. Empty cells are NOT IN PAPER. 5. Claim list: each claim must carry a URL from Papers. No DOI invention. 6. Elicit columns vs Consensus query phrasing. Do not invent a column the tool lacks. 7. Gaps: outcomes with zero URL-backed rows. 8. Compliance pass: quote Banned words, invented DOIs, and papers over Cap. Cut them. Constraints: - Every evidence claim needs a citation URL that appeared in Papers. - Never invent a DOI, PMID, or finding. - Do not fetch new papers from the web. - Honor Cap. - Not medical advice even if papers are clinical.
Instructions
Replace every [bracket] with your details before running. Works on ChatGPT, Claude, and Gemini.
Examples
Example Input
Question: Does 15-minute outdoor daylight before 10:00 change self-reported sleep onset in adults? Papers: 1) Smith 2021 daylight and sleep diary, https://doi.org/10.1234/example.sleep.2021 2) Lee 2019 indoor lighting review, https://example.org/lee-2019-lighting 3) Anon poster, no URL Tool: both Include: adult human, sleep onset or diary Exclude: animal only, no full text URL Outcomes: sleep onset latency self-report; dropout Banned: proven cure, everyone should Cap: 10 Types: only if Papers names them; Smith named as diary study; Lee named as review MissingPolicy: NO_URL
Example Output
1. Lock: Elicit 2026 and Consensus 2026. Not a new web scrape. Not medical advice. 2. URL audit: Smith 2021 https://doi.org/10.1234/example.sleep.2021 OK. Lee 2019 https://example.org/lee-2019-lighting OK. Anon poster NO_URL, cannot support a claim. 3. Inclusion: Smith in (adult diary). Lee is a review on indoor lighting; map as review, not as a daylight trial. Anon excluded (no URL). 4. Map. Question as given. Rows: Smith 2021 diary URL given, sleep onset latency self-report = NOT IN PAPER unless the user quoted it (they did not quote a number), dropout NOT IN PAPER. Lee 2019 review URL given, not a 15-minute daylight trial. Cap 10, used 2. 5. Claims: none with a numeric effect because Papers did not include effect sizes. Do not invent minutes-to-sleep. 6. Elicit columns: title, year, URL, study type as named, outcomes. Consensus query: adult daylight before 10:00 sleep onset, then restrict to the two URLs. 7. Gaps: no URL-backed sleep onset number. Dropout unknown. 8. Compliance: cut proven cure, everyone should. No invented DOI. Anon stays NO_URL. Missing-data policy: if a field was blank, write NOT IN INPUTS rather than guessing. Lock any tool version named in Inputs; if unnamed, write unknown. No invented testimonials, star ratings, or press logos. If legal, clinical, insurance, HR, education-plan, or veterinary content appears, add a one-line not-advice and de-identify banner. Quote banned-word hits and cut them. End with a gaps list of five bullets the user still owes you. Character and byte caps in the job are hard; print counts when relevant. Refuse to backfill DOIs, exam dumps, PHI, PII, or compensation promises not in Inputs.