Grounded in evidence
AI helps me transcribe, organise, analyse and draft from large collections of family history material including certificates, photographs, newspapers, letters, interviews and research notes.
That allows me to spend more time where it matters most - researching, evaluating evidence, resolving conflicting information, writing and producing your finished collection.
The historical judgement remains mine.
Every story is referenced back to the records you provide or that I discover during research. Wherever practical, you can trace important conclusions back to their original sources. If the evidence is uncertain, I say so. I never knowingly invent people, events or evidence simply because an AI system suggests them - in fact my workflow tries to eliminate the possibility of something like this in the first place.
Where AI is used, I choose services that provide the highest available privacy controls and, wherever possible, prevent client material being used for AI model training.
If you would prefer no AI involvement, simply let me know before work begins. In many cases I can complete the work another way, although this may affect cost, turnaround time and the range of digital products available.
Privacy, consent & living people
Family history often involves people who are still living, or stories that deserve particular care.
Protecting their privacy, dignity and wishes is one of my core professional responsibilities.
Every project begins with informed consent before AI-assisted tools are used. That consent allows me to use AI to help organise, analyse, draft and produce your family history from the material you provide. It does not automatically mean every document, photograph or story about every living person should enter an AI workflow or appear in the finished collection.
Depending on the circumstances, I may seek additional consent, anonymise names, remove identifying details, summarise sensitive material, process sections manually, or recommend excluding some material altogether.
To guide these decisions, I use a simple three-tier privacy model.
Tier 1 - Historical Material
This is material that can normally be processed using AI once you have approved the project. It generally relates to deceased people, historical records and publicly available information where there is no realistic risk of exposing the private information of living people.
Examples include:
- historical birth, marriage and death records
- census, electoral and parish records
- newspapers
- shipping and immigration records
- land records
- wills and probate
- military records
- historical photographs where no living person is identified
- maps and published historical sources.
As a practical guide, family groups whose youngest children were born more than 100 years ago are usually considered low risk. More recent generations receive a closer privacy review.
Tier 2 - Living People
This material may involve living people or more recent generations but is not necessarily highly sensitive.
Examples include:
- family trees containing living relatives
- recent family letters and notes
- photographs including living people
- family stories involving people who may still be alive
- first-person memoirs or life stories of living relatives, drawn from interviews or personal recollections
- ordinary biographical information about living relatives.
Where appropriate consent exists, Tier 2 material can often be used within my normal AI-assisted workflow. A first-person memoir or recorded life story usually sits here or higher, because it centres on a living person and their recollections - so it always begins with that person’s consent and a closer privacy review.
Where additional privacy is needed, I use an anonymise-first approach. Names are replaced with neutral labels such as Living Relative, Daughter 1 or Cousin A. Identifying details are removed before the material enters the AI workflow. The key linking those labels to real names is kept separately and never entered into the AI system. Real names are restored manually during final editing.
This approach provides greater privacy protection but requires more manual work. It may increase cost and may limit some products, such as interactive family websites, podcasts or AI-powered family assistants. If this is recommended, I will explain the implications before work begins.
Tier 3 - Sensitive Information
Some material requires a much higher level of care.
Examples include:
- information about children
- adoption and donor conception
- DNA discoveries
- health or disability information
- abuse or family violence
- criminal matters
- serious family conflict
- information shared in confidence
- memoirs or life stories touching on health, family conflict, or matters shared in confidence
- recent events involving living people who have not given consent.
For this material I may recommend that it is handled manually, anonymised, summarised, excluded from AI-assisted processing, or omitted from the finished work altogether.
Occasionally I may decide not to undertake part of a project where the ethical, privacy or consent issues cannot be managed appropriately.
That decision is never because the story is unimportant. It is because protecting living people is more important than producing another page of family history.