Ghosts of DC has been digging up Washington’s buried history since 2012. AI has changed how that work gets done. Here’s the honest version.
How the work gets done
The work has changed shape. We used to read the old newspapers ourselves, hunt down the photos ourselves, and piece the stories together ourselves. Now we spend our hours pointing AI at the right archives, reading what it brings back, checking it against the source, and turning it into a post. The hours are still long. We just get more done in them. A story that used to take a weekend takes an afternoon. Stories we’d never have gotten to at all now make it onto the site.
What stays with us is the judgment. We pick which stories to tell, choose the angle, decide what’s worth your time, and read every post before it goes live. The decisions are ours. The mistakes are ours.
Research
The archives we lean on most:
- Library of Congress, especially Chronicling America (digitized newspapers back to the 1800s), HABS/HAER (the historic buildings survey), and the Geography and Map Division.
- The National Archives.
- Smithsonian Open Access and the Smithsonian Institution Archives.
- DC Public Library’s Washingtoniana / People’s Archive.
- New York Public Library Digital Collections.
- Boston Public Library (Flickr Commons).
- The Washington Post historical archive.
- The Internet Archive for digitized books, journals, and city directories.
- Wikimedia Commons and other public-domain image collections.
AI pulls from these. It doesn’t get to decide what’s true. Every historical claim is checked against its source, and the sources are listed at the bottom of every research-driven post so you can check our work.
Writing
AI drafts and edits. We direct: which story to chase, which angle to take, which 1886 quote to pull, which lead opens the post.
House rule, every post: SEO title, search snippet, headline, and first paragraph all promise the same thing. No bait-and-switch.
Images
We colorize black-and-white photos, generate images when no historical one is available, and sometimes animate stills or turn them into live photos. The animations are experimental. They’re our attempt to imagine what a moment might have looked like in motion, not a recreation of what was actually there.
We label these in the caption. We never pass off a generated or altered image as the real thing.
Accountability
AI makes mistakes. So do we. We’ve been making them in public since 2012 and apologizing for them ever since. We’re not an academic history blog. But credibility and trust are the only currency we have, and AI doesn’t change that.
If you spot something wrong, email hello@ghostsofdc.org. We’ll fix it where appropriate.
Last updated: June 2026. We’ll revisit this as the tools change.