Jump to a section
Scan or photograph the original at the best quality you can and keep that file safe. Use AI to repair or colourise a copy, then compare it with the original: AI can invent details, especially in faces. Label restored versions as edited so future generations know what's real.
What AI can do with old photos
Depending on the tool, you can ask AI to:
- Remove scratches, creases and dust spots.
- Sharpen a blurry or faded image.
- Add colour to a black-and-white photo.
- Fix a tear or a missing corner.
Why it can quietly change a photo
AI doesn't recover lost detail. It guesses what was probably there. Just as chatbots sometimes state wrong facts confidently, as researchers at OpenAI have explained, image tools can fill in a blurred face with features that look convincing but were never there.
Colourising is a guess too. The AI doesn't know your grandad's jumper was green; it picks a colour that seems likely.
How to tell if a photo or video is AI-made, and what to do when you can't
OpenAI: Why language models hallucinate (5 September 2025) Kalai and colleagues: Why Language Models Hallucinate (research paper, 4 September 2025)
How to do it well
A few habits keep the history honest.
- Scan or photograph the original in good, even light. Keep that file, untouched, somewhere safe.
- Work on a copy, never the only version.
- Ask for gentle repairs first, for example: "Remove the scratches and dust but don't change anyone's face."
- Compare the result with the original side by side, especially the faces.
- Name the file clearly, for example 'Grandma 1962 – AI restored', so nobody later mistakes it for the original.
A word on privacy
Family photos show real people, some of whom can't give their permission. Uploads to AI tools are covered by the company's privacy rules. Google, for example, says a subset of Gemini chats is reviewed by human reviewers. Check your settings before uploading, and think twice about photos of children or anything private.
AI chatbots and your privacy: what they keep, who sees it, and how to take control
Sources
These sources support the guide. Tool settings, prices and policies change, so dated source boxes show when those details were checked.
- OpenAI: Why language models hallucinate (5 September 2025)
- Kalai and colleagues: Why Language Models Hallucinate (research paper, 4 September 2025)
- Google: Gemini Apps Privacy Hub
- The Register: OpenAI says models are programmed to make stuff up instead of admitting ignorance (17 September 2025)
- Anthropic: updates to consumer terms and privacy policy
AI helps with research and drafting. Dan checks the sources and is responsible for what is published.
How we check