In June, I set out to repair a badly damaged photograph of my grandmother, Myrtie Hairston Bryan. I uploaded the image to ChatGPT and asked it to restore the photo without altering the original coloring or her facial features. The result looked promising, and feeling satisfied, I labeled it “restored” and proudly shared it on my blog. My enthusiasm had been sparked after listening to Episode 25 of The Family History AI Show with Mark Thompson and Steve Little—though, in hindsight, I clearly missed an important point.
Restoration or Reconstruction?
When I first ran my grandmother’s photo through Artificial Intelligence (AI), I thought I was seeing a careful repair job. The scratches and spots vanished, and her face appeared smoother and clearer than before. Only later did I realize the AI hadn’t really fixed the original image at all—it had recreated it.
AI studied the damaged image and generated new pixels to replace the missing details. The end result certainly looked like my grandmother, but it wasn’t the same photograph she carried home from the studio in the early 1900s. It was a modern, computer-generated version of that moment.
This difference matters. Just as we distinguish between an exact transcription and a “cleaned up” version of a document, we should be clear about whether a picture shows original evidence or an interpretation.
When we use AI to “improve” old photographs—whether by sharpening, recoloring, or filling in missing areas—we’re creating a new version. Those images should be clearly labeled, and the unaltered scan should always remain the authoritative copy.
Moving Forward with AI
I have removed the June 20th post because I don’t want someone to see that recreated image and upload it as a true restoration to sites such as Ancestry, where it would be copied and shared repeatedly.
I still appreciate what AI can do. It’s a tool I use daily for a wide range of tasks, including genealogy. However, when it comes to photographs, I’ll be much more cautious and look forward to the day when AI can truly repair a historical image rather than rebuild it.
My thinking about AI and photos has been shaped in part by conversations in the genealogy community—especially discussions from The Family History AI Show with Mark Thompson and Steve Little.
If you want to know more about the families I research, click here to like my Facebook page, where you will see each post and other genealogical finds.
Diana
Sources
“Episode 25: AI Image Generation Advances,” The Family History AI Show, (https://podcasts.apple.com/us/podcast/ep25-chatgpt-4o-transforms-image-generation-jarrett/id1749873836?i=1000712214261accessed 21 October 2025).
“Episode 27: Restoration or Recreation?” The Family History AI Show, (https://podcasts.apple.com/us/podcast/ep27-ai-image-restoration-concerns-perplexitys-future/id1749873836?i=1000717145783: accessed 21 October 2025).
