In this 2026 guide, I walk you through the exact click-by-click process to generate images in Claude AI. I start from the very beginning: opening your browser and going to claude.new so you land in a fresh conversation immediately. From there, I show you the simple layout (left sidebar + main chat area) and the small plus icon on the left that I use to spin up a clean, dedicated chat for each new image idea—think of it like opening a new sketchbook page so your experiments stay organized. Then I get into the part that actually determines your results: writing a solid prompt. I explain how to be specific about subject, style, and mood, and I give concrete examples like “a misty medieval stone castle on a cliff at sunrise” with cinematic lighting, or a “whimsical watercolor” scene with a soft pastel palette. I also cover composition cues (wide angle, portrait close-up, top-down view), atmosphere words (moody, vibrant, serene), and quick negatives like “no text in the image.” Finally, I show the simple submit flow (click submit or press Enter), and how small wording tweaks and re-submits are the fastest way to iterate toward the exact look you want.