Is An AI Picture Worth 1000 Words?

When AI systems describe and recreate images, they reveal their training biases and assumptions about what’s “normal” or expected. This activity makes visible how AI can transform, erase, or stereotype the realities we document. By watching our images get “translated” through AI’s lens, we can see whose perspectives dominate these technologies and whose stories get rewritten. This process often adds elements that weren’t there (like flags in meeting rooms or specific racial features) while removing cultural specificity and context.

Upload image → Get 1000-word AI description → Generate new image from description

Instructions:

  1. Choose a photo that’s meaningful to you (your classroom, community space, family gathering, local landmark)
  2. Upload it to an AI chat (like ChatGPT, Claude, Gemini)
  3. Ask: “Please describe this image in 1000 words, including every detail you can see”
  4. Copy the entire 1000-word description
  5. Open a fresh AI chat session (important: don’t use the same conversation)
  6. Paste the description and ask: “Please generate an image based on this description”
  7. Compare your original photo with what the AI created

Conscientization

Reading the world through this activity

  • What details did the AI add that weren’t in your original image?
  • What aspects of the original disappeared or got transformed?
  • If the AI added people, objects, or contexts – what assumptions did it make?
  • Whose version of “normal” does the AI seem to understand?

Praxis

Reflection leading to transformation

  • How might these AI translations impact how people’s stories get told?
  • What happens when AI systems “fill in the blanks” based on biased training data?
  • How could communities ensure their realities aren’t erased or rewritten by AI?
  • What would you want AI developers to understand from this exercise?

Dialogue

Ongoing discussion

  • Keep exploring: Try with images from different cultures, communities, contexts
  • Share your original and AI-generated images with others
  • What patterns do you notice across different people’s results?
  • How do the AI’s assumptions connect to larger systems of power and representation?

Scroll to Top