Our Approach to Responsible AI in Genealogy

Last Updated: October 27, 2026

Rooted History uses AI to turn the facts in your FamilySearch tree into readable, historically grounded biographies of your deceased ancestors. Our goal is to make them the most transparent AI biographies anywhere: you should always be able to see where a sentence came from.

This page is not legalese (see our Terms §7 for the legal disclaimer). It's simply the truth about what our AI does, what it doesn't do, and how to use it wisely.

Which AI does what

Claude Opus 5.5 (Anthropic) writes biographies, Historical Perspectives, and the Relaxed Reading and For Kids versions.
Claude Sonnet 5 (Anthropic) checks each biography against your FamilySearch data, edits it, and labels where each sentence came from.
Claude Haiku 4.5 (Anthropic) prepares research notes from your records.
Grok 4.7 (xAI) checks historical statements against outside sources found on the web. Only general historical statements are sent, never your relatives' names or dates. xAI keeps no copy of these requests and does not train on them.
Gemini 3.1 Pro (Google) reviews biographies written in a language other than English for language quality.
ElevenLabs v4 reads biographies aloud with an AI-generated voice.
In Investigations, Claude Haiku 5.5 searches your tree and Claude Sonnet 5.5 explains what it found.

Models change over time. The "About This Document" page in every biography PDF lists the exact models used for that biography.

Claude and Gemini run for us on Google Cloud's Vertex AI, which does not use our data to train its models. xAI's Grok runs under zero-data-retention terms.

What the AI does

1. Starts from the facts in your FamilySearch tree: names, dates, places, relationships, life sketches, and sources.

2. Adds historical context for that time and place.

3. Writes a dignified, readable story, then checks it: facts against your records, and historical claims against outside sources.

→ Think of FamilySearch as the skeleton and us as the storyteller who adds muscle and clothing.

What the AI never does

Discover new records or people
Change your FamilySearch tree
Let an outside source overrule your records: when the two disagree, your records win
Generate stories about living people

The story can only be as good as your tree.

Every sentence shows where it came from (paid biographies)

Each sentence in a paid biography carries one of these labels:

From the record, shown in bold: a fact from your ancestor's FamilySearch data.
Sourced, with a numbered mark: historical context we checked against an outside source. Click the number to see the source. We show only pages the checker actually opened and named for that sentence, usually one to three, never a list of everything it happened to read.
Reasoning: the writer's inference from the evidence, labelled so you can judge it yourself.
Background: general history of the time and place, not a claim about your ancestor.

When reputable sources disagree, we don't quietly pick a side. The biography shows the disagreement in a "historians disagree" panel. The Sources tab lists every source used.

Reading versions and narration

Relaxed Reading and For Kids are rewritten from the original biography. They add no new facts, and they never make an uncertain statement sound certain (or the reverse). Each sentence links back to the original it came from, with the same sources.
Narration is an AI-generated voice reading the exact text you choose. If the text changes, the old recording is retired; it never plays against newer text.

Difficult history

We name tragedies by their real names: famine, slavery, genocide, ethnic cleansing, forced displacement.

We refuse euphemisms, victim-blaming, or sanitisation. If the history is painful, the story will reflect that with respect and clarity.

Known AI limitations

Even the best models can:

Get a minor detail wrong
Generalise regional history too broadly
Miss nuance in very obscure periods or places

Outside sources can also be wrong or out of date. Sources and labels show you where each statement came from; they don't make it infallible.

Your responsibilities (please read this bit)

1. Double-check your FamilySearch data before generating a biography.

2. Treat bold text as your records; read reasoning and background as educated context.

3. Follow the sources and verify anything important before using it formally.

4. Tell us (support@rootedhistory.io) if something feels wrong. We review every report and fix systemic issues.

Investigations tool

Our separate "Investigations" feature scans your tree for patterns, conflicts, and research leads. It is a hint generator, not a conclusive answer. Always verify its suggestions.

Bottom line

These biographies are powerful tools for feeling your family's past and sharing it with loved ones.

They are not primary research, academic papers, or legal documents.

Questions, concerns, or spot something off?

→ support@rootedhistory.io

Rooted History – making family history feel real, responsibly.