A poorly thought pattern here.
Single AI output presentation in ChatGPT can have:
- preamble text
- an image or multiple parallel images (now seemingly not offered)
- multiple images created with tool iteration
- final text, the AI welcomely disobeying “don’t say anything”.
The copy button, when wanting to get the textual output, however only copies the last image to a clipboard.
This is wrong, deviates from expectation. It is also not needed if purposeful, as every image has a download button, besides being at full resolution in the UI and suitable for copying right out.
Why is there text to be copied? Because image generation works much better on a passed context when the chat AI has made a robust description instead of an internal tool field “prompt”, and it is unlikely to write well or extensively into an optional prompt field.