Cross-Platform Prompts to Try
You can treat these prompts like reusable “mini‑apps” that run well in ChatGPT, Claude, Gemini, Perplexity, Grok, and local models, as long as you structure them consistently and keep tool‑specific bits modular. Below are cross‑platform genealogical prompts you can paste almost verbatim into any major system, with only small switches where needed.
See also: How to Set Up a Personal Prompt Library
1. Cross‑platform research problem solvers
These are designed to be dropped into: ChatGPT (GPT‑5.x, Thinking mode), Claude (Sonnet/Opus), Gemini, Grok, Perplexity (normal or Deep Research), and a strong local/open‑weight model.
A. Brick‑wall case analyzer(generic version)
<prompt>
“Act as a professional genealogist following the Genealogical Proof Standard.
You will help me analyze a brick‑wall problem.
Restate the research question clearly.
List every explicit statement I have made about the problem.
Infer and list my unstated assumptions.
Identify conflicts, gaps, and weaknesses in my current reasoning.
Propose 3–5 alternative hypotheses, each with:
a brief rationale
specific record types and likely jurisdictions to test it
which hypothesis would be highest priority and why.
Here is my current research summary (including citations and notes):
[PASTE YOUR SUMMARY]
Work step‑by‑step and label each section. Do not invent records; only reason from what I provide and typical record‑keeping practice for the time and place.”
Use GPT “Thinking” / Claude Sonnet / strong local model when you want maximum reasoning detail.
Use Perplexity Deep Research with: “After you’ve done the internal analysis, if helpful, survey current online catalogs and collections for sources that might test these hypotheses and cite them.”
</prompt>
B. Conflicting‑evidence adjudicator
<prompt>
“You are a genealogy analyst.
I will give you several sources that conflict about the same person.
Tasks:
For each source, identify: type (original/derivative), informant (if known), and whether information is primary or secondary.
Create a table listing the key facts (name, date, place, relationship, etc.) and how each source reports them.
Explain the conflicts and give plausible reasons (e.g., informant error, delayed reporting, copying errors, cultural naming patterns).
Argue for the most plausible conclusion for each fact based only on the evidence given and typical record reliability.
Recommend additional records that could resolve remaining conflicts.
Here are the sources (each labeled):
[SOURCE A TRANSCRIPTION]
[SOURCE B TRANSCRIPTION]
[SOURCE C TRANSCRIPTION]
Think step‑by‑step, and clearly separate evidence summary from interpretation.”
</prompt>
This runs well across all major models; on Claude you can additionally say, “Create the comparison table as a downloadable spreadsheet if possible.”
C. FAN‑club research planner
<prompt>
“Help me expand my FAN (Friends/Associates/Neighbors) research for this problem:
Research question: [STATE QUESTION].
Target person: [SUMMARY].
Current FAN list with brief notes:
[name] – [how they appear]
[name] – [how they appear]
…
Tasks:
Group these associates into logical clusters (neighbors, in‑laws, business partners, church community, migration chain, etc.).
For each cluster, list what they might reveal about my research question.
For each cluster, propose 3–5 specific records or collections to search (by record type and jurisdiction, not brand), and suggest search strategies.
Prioritize the clusters for research and explain why.
Use numbered lists and short paragraphs so I can paste this into my research log.”
</prompt>
D. Indirect‑evidence proof outline
<prompt>
“I am preparing an indirect‑evidence proof argument.
Using the information I provide, draft an outline I can turn into a formal narrative.
Restate the research question.
List all key pieces of evidence that support the proposed conclusion, grouped by record type or theme.
List any evidence that appears to contradict the conclusion and possible explanations.
Suggest a logical section outline for a GPS‑style proof argument (introduction, context, evidence sections, resolution of conflicts, conclusion).
For each section, list which evidence items belong there and note remaining gaps.
Here is my research summary and extracted evidence:
[PASTE NOTES / QUOTES WITH SOURCE TAGS]
Do not write the final narrative; just build the outline and reasoning structure.”
</prompt>
2. Locality, context, and catalog prompts
These work well in Perplexity, Grok, Gemini, and any model with browsing; on offline/local models, you can reuse the exact text but limit it to reasoning from what you paste.
E. Locality guide starter (web‑enabled)
For Perplexity / Grok / Gemini with web)
<prompt>
“Create a starter locality guide for genealogical research in:
Place: [COUNTY / TOWN, STATE / COUNTRY]
Target period: approximately [YEAR RANGE].
Tasks:
Describe civil and ecclesiastical jurisdiction changes that matter for record‑keeping during this period.
List key record types (civil registration, church registers, censuses, land, probate, tax, military, court) and where they are usually held.
Identify digitized or indexed collections currently available online (by archive or platform name), and collections that likely require on‑site research.
Note any major record losses, gaps, or known substitutes.
Suggest 5–10 next research steps tailored to someone looking for [e.g., parentage, migration, kinship network] in this place and time.
Cite sources and link to repository or catalog pages where possible so I can verify.”
</prompt>
For offline/local models, replace item 3 with: “Based only on typical practices, suggest likely repositories and collection categories; I will check catalogs myself.”
F. Repository trip prep
Best with Perplexity or Gemini with web
<prompt>
“I am planning a genealogy research trip.
Repository: [ARCHIVE / LIBRARY / COURT / CHURCH ARCHIVE]
Location: [CITY, STATE/COUNTRY].
Tasks:
Summarize hours, access rules, ID requirements, and rules for cameras / scanners (only from official sources; provide links).
Identify key record groups relevant to [brief research focus] that are held onsite and not fully digitized, with call‑number or catalog references where available.
Suggest a prioritized research plan for a [one‑day / two‑day] visit, including what to order ahead or pull first.
List practical tips (parking, lockers, copy costs, how to handle restricted items) based on official guidance or credible user guides.
Where information is unclear or missing, say so and point me to a phone or email contact, not guesses.”
</prompt>
3. File‑centric and workspace‑oriented prompts (Claude, ChatGPT, Perplexity Labs/Computer)
These assume assistants that can create files or run multi‑step “Labs/Computer/agents.”
G. Automated research log builder
For Claude / GPT / Labs / Computer
<prompt>
“I’m going to paste messy research notes from a work session (mixed URLs, quick jottings, partial citations).
Parse my notes into individual searches or actions.
Build a structured research log with columns: Date, Repository or Website, Collection / Record Set, Search Terms / Parameters, Result (Found / Not found / Not checked), Brief Notes, Citation Draft.
Return the log as:
a clean markdown table in this chat, and
if your environment allows, a downloadable CSV or spreadsheet file.
Do not fabricate searches I did not actually mention.
Here are the notes:
[PASTE NOTES]”
</prompt>
In Claude Sonnet 4.5 you can add: “Create a spreadsheet file in this workspace.” In Perplexity Labs/Computer, add: “Save the CSV as ‘[SURNAME]log[DATE].csv’ and tell me where it’s stored.”familylocket+1
H. Evidence‑correlation matrix
<prompt>
“I will provide several transcribed records referring to what might be the same person or family.
Tasks:
Build an evidence‑correlation table with rows for each record and columns: Record label, Date, Place, Person’s name as written, Others named, Event type, Key details, Conflicts / Comments.
After presenting the table, summarize the main consistencies and conflicts.
Suggest 3–5 records to search next that are most likely to resolve identity or relationship questions.
Here are the records:
[RECORD A]
[RECORD B]
[RECORD C] …
Keep opinions in a separate “analysis” section; keep the table strictly factual.”
</prompt>
I. Case‑study slide outline
<orompt>
“You are helping me prepare a 20–30‑minute case‑study presentation for genealogists about this research project:
[PASTE SHORT NARRATIVE OF THE CASE].
Tasks:
Propose a slide‑by‑slide outline (10–18 slides) with titles and bullet points.
For each slide, indicate which record images, maps, or charts would be most effective to show.
Suggest a one‑paragraph abstract and 3 learning objectives.
Format the outline in a simple numbered list so I can copy it to PowerPoint or Keynote.”
</prompt>
In Claude, add: “Optionally create the slide deck file if available.” In Perplexity Labs / Computer, add: “Generate the slides in [format] and export as a PPTX I can download.”aigenealogyinsights+1
4. DNA‑aware prompts (safe for any model; best on local if you want extra privacy)
J. DNA cluster research planner
<prompt>
“I will describe several DNA match clusters related to a research problem.
For each cluster, do not attempt to interpret raw DNA; instead, focus on traditional records that might explain the matches.
Restate the research question.
For each cluster, list:
likely time frame / ancestral generation (based on what I say, not on cM rules you invent)
geographic focus
hypothesized common ancestors or surname patterns (only from my notes).
For each cluster, propose 3–7 traditional record searches (by record type and jurisdiction) that could confirm or refute the hypothesized common ancestry.
Prioritize which cluster I should work on first and explain why.
Here are my cluster descriptions and notes:
[PASTE CLUSTERS / NOTES]”
</prompt>
On a local open‑weight model, you can safely paste a little more detail; with cloud models, keep it mostly to cluster summaries and surnames, not screenshots of match lists.
5. Task and project‑management prompts
These work everywhere and pair nicely with agentic features (Perplexity Computer, Claude computer‑use, local agents).
K. Multi‑family project breakdown
<prompt>
“I will paste a long narrative description of an ongoing multi‑family research project.
Tasks:
Identify and list distinct sub‑projects (by family group, locality, or research question).
For each sub‑project, break work into tasks of 20–60 minutes, each with:
a clear action verb
input needed (records, websites, repositories)
expected output (log entry, citation, extract, map, timeline, etc.).
Mark tasks that can be done from home vs. those that require on‑site work.
Output everything as a table I can paste into a spreadsheet or task manager.
Here is the description:
[PASTE]”
</prompt>
In Perplexity Computer or a similar system, follow up with: “Now take the ‘from home’ tasks for Sub‑project 1 and execute the web‑only ones, logging findings into a new document,” then closely review what it does.familylocket+1
L. Weekly genealogy sprint planner
<prompt>
“Help me plan a realistic genealogy ‘sprint’ for the next 7 days.
Inputs:
Time available: about [X] hours total.
Current top 3 goals: [GOAL 1], [GOAL 2], [GOAL 3].
Constraints: [e.g., no weekday courthouse visits, limited subscription access, etc.].
Tasks:
Turn the goals into 5–10 concrete, small tasks (30–60 minutes each).
Group tasks by goal and suggest an order that respects dependencies.
Provide a simple 7‑day schedule suggestion.
Include a short checklist to review at the end of the week (what to capture in the research log, what to back up, what to tee up for next week).”
</prompt>
You can reuse this verbatim across tools; in Gemini or Claude, you can add: “Create a compact table I can print or screenshot.”
6. Cross‑platform prompt hygiene for genealogists
A few patterns make all of these portable and stable across models and products:thepromptwarrior+1
Keep the role and task structure explicit (“Act as…”, then numbered tasks).
Always paste your own evidence; never ask any model to invent record images or unverified citations.
Separate facts from analysis in your instructions (“table first, then interpretation”), which helps with genealogical rigor and makes outputs easier to audit.
Maintain a prompt library in a single place (Obsidian, Google Docs, Notion, a plain folder of text files) so you can version prompts and note which tools they work best in.
No comments:
Post a Comment