I compared ChatGPT and Gemini's AI image generation - and a single prompt tweak made a big difference
A ZDNET prompt tip shows how asking a chatbot to write the image prompt can improve results in ChatGPT and Gemini.
Intelligence analysis by GPT-5.4 Mini
ZDNET tests a simple trick for AI image generation: give the chatbot a basic idea, then ask it to write a fuller prompt for the image model. In examples with Gemini and ChatGPT, the generated prompts added detail and improved the resulting images.
Instead of telling a robot artist every tiny detail yourself, one person can ask the robot to write the instructions first. It is like having a helper turn a simple sketch into a full recipe, which can make the picture come out better.
Analysis
The prompt trick
ZDNET contributor Lance Whitney recommends starting with a simple description of the image you want, then asking the chatbot to turn that into a full prompt for the image generator. In the example used here, the request is for “a sunflower made of sheet metal in a pencil drawing style,” followed by a request for a prompt that can be used with Nano Banana or ChatGPT Images.
What the tests showed
Radhika Rajkumar says the method produced useful results in both Gemini and ChatGPT. Gemini returned a prompt that added details about the metal texture, visible welds, shading, and even a blurry garden background. ChatGPT produced an even longer prompt with details like thin curved metal plates, seams, rivets, etched vein details, graphite linework, cross-hatching, and high-contrast shading.
The article’s point is not that one model is universally better. It is that asking the chatbot to draft the prompt can surface useful visual details the user might not have thought to include, while also helping avoid wording that a generator might flag or refuse.
Practical limits
Whitney notes that if the prompt becomes too long, users can ask for a shorter version. The piece also stresses that prompt results vary across tools, because companies keep updating models. ZDNET also reminds readers to consider company policy, copyright uncertainty around AI images, and privacy settings before using these tools for work.
Bottom line
The article frames this as a beginner-friendly way to get stronger image outputs from ChatGPT, Gemini, or similar tools without needing expert prompt-writing skills.
Key points
- ZDNET recommends asking a chatbot to write the image-generation prompt for you.
- In tests, Gemini and ChatGPT both added useful detail to a simple image idea.
- The method can help avoid wording that a generator might flag or refuse.
- If the generated prompt is too long, the user can ask for a shorter version.
- The article warns that model updates, policy rules, and privacy concerns still matter.
If this approach works well, it can make AI image tools easier for beginners and save time for people who do not know how to write detailed prompts. It may also reduce refusals by helping the model phrase requests in a way the image generator accepts more easily.
The article says results vary by tool and model, so the same prompt may not work the same way everywhere. It also notes concerns around company policy, copyright uncertainty, and data sharing, which can limit how safely people use AI image generators.
