Our Commitment to an Accurate Record
Research standards, sources, and how we use digital and AI tools
The Rainbow Story Hub Foundation uses artificial intelligence tools the way we use scanners, microphones and cameras: as equipment that helps a small volunteer organization do preservation work at a scale we could not otherwise reach. AI helps us transcribe oral histories, read damaged newsprint, index thousands of pages, and repair scratched photographs.
Our research rests on real collections — the City of Edmonton Archives, the Provincial Archives of Alberta, the Edmonton Journal through Edmonton Public Library’s ProQuest access, and the volunteer-digitized records that so much community history depends on. AI helps us search those collections and cross-check what we find in them. It is not one of them, and it never stands in for them.
It does not write our history for us, and it never invents it.
What this makes possible
Community publications are where much of Edmonton’s queer history survives, and they were never indexed. Finding a single name, venue or event across long runs of Times.10, Womonspace News and ISCWR Coronation programs has traditionally meant reading page by page. Oral history interviews sat untranscribed for years. Photographs too damaged to publish stayed in the box.
None of this is work that was being done and is now being taken away. It is work that would otherwise not get done at all at a volunteer organization our size.
Where we use these tools
Discover — Suggesting search terms, alternate spellings and period vocabulary; searching digitized text across our own collections; identifying which archives may hold what we’re looking for.
Document — Text recognition on scanned newspapers, newsletters and correspondence; extracting names, dates and venues from material too large to read by hand.
Organize — Indexing, cross-referencing and drafting collection descriptions for human editing.
Interpret & Write — Drafting, editing and structuring our own writing; plain-language versions; grant and administrative documents.
Visualize — Repairing damage to photographs we hold rights to; functional graphics where a project has no budget for design work.
Preserve — Transcribing and captioning oral histories; repairing degraded audio and film.
Verify — Cross-checking dates, names and sequences against our documented sources.
What we do not do
- We do not create images, audio, video or documents depicting real people, places or events.
- We do not put words in anyone’s mouth. We do not generate quotations, testimony or recollections, and we do not synthesize anyone’s voice or likeness — living or deceased.
- We do not alter imagery, events or facts to depict anything other than what occurred.
- We do not use material we do not own, license or have permission to use.
- We do not use AI in place of an artist we have the means to pay.
- We do not treat AI as a source. Every fact we publish is traceable to a record you can check yourself.
Where a gap exists in the historical record, we identify it as a gap. We do not fill it.
Artists
RSHF commissions and pays local 2SLGBTQ+ writers, researchers, artists, designers, illustrators, photographers and videographers, and we budget for that work in our grant applications.
We are volunteer-led and without core operating funding. Not every project is funded, and some are funded without a graphics allocation. In those cases we sometimes use AI tools for functional graphics — event promotion, social media, layout — and we would rather say so than pretend otherwise. Where there is money for creative work, it goes to an artist. Where a piece calls for artwork rather than a graphic, it is commissioned or it is not made.
Labelling
Restored images, AI-assisted transcripts, generated graphics and speculative artwork carry a label saying so.
Work that carries no AI label is human-authored. That is the default, and it applies to most of what we publish.
Ask us
If you want to know how something we published was made, ask. If you believe we have used these tools in a way our policy does not permit, or failed to label something we should have, bring it to us — we keep records, we will check, and we will answer. If we got it wrong, we will say so and correct it.
Effective date
This policy takes effect September 1, 2026 and applies to work produced on or after that date. Earlier work — produced before we had a written standard — is reviewed against the policy whenever it is republished or reused.
This page summarizes our full policy. The complete document sets out our commitments in detail, including our handling of copyright, privacy and consent, oral history transcription standards, and moving image work.
[Download the full policy (PDF)]
