ChatGPT's image tools have changed a lot since this guide first launched. The creative fundamentals here — composition, lighting, camera language, the ad templates, and the campaign workflow — still hold. What changed is how the images are made, edited, and shipped. Quick timeline:
gpt-image-2): visual reasoning, an optional "thinking" mode, flexible aspect ratios (3:1 → 1:3), up to 2K resolution, much stronger multilingual text, and up to 8 coherent images in a single request.The five biggest things to relearn: (1) it's no longer DALL·E; (2) you can edit with a Select tool right inside ChatGPT; (3) there's an optional "thinking" mode for complex, factual, or multi‑image visuals; (4) text and multilingual rendering are dramatically better; (5) every generated image now carries provenance signals (C2PA + SynthID). The new and rewritten 2026 sections below are marked 🆕.