This is one space where I think generative AI would be helpful - generate pictures that look mostly right and fill a thumbnail or something on a webpage that doesn't necessarily provide indispensable information but nevertheless fills a void, and costs little. You could probably describe what was in the original photograph and/or draw a representation of it crappily in Paint and use generative AI to fill in the blanks, such that it wouldn't be considered a derivative work but rather just a representation of facts.