Start with what the picture has to do. A picture in teaching earns its place by helping students understand, remember or engage with something. Before you describe it, decide its job: setting the scene for a case, showing a process, prompting discussion, or making a page less forbidding. A picture with a clear job is easier to describe, and one without a job is decoration that competes with your content for students’ attention.
Describe the content, not just the subject. “A laboratory” gets you a generic laboratory. “A first-year student weighing a sample on an analytical balance, wearing safety glasses and a lab coat, with the draft shield open” gets you a picture you can teach from. Say who is in it, what they’re doing, where, and what detail matters most. Then use “Change this picture” to fix one thing at a time, rather than starting again.
Keep it simple. One idea per picture works best on a slide or a handout. Busy scenes are hard to read when projected, and students’ eyes go to the wrong detail. If you need to show several things, make several pictures. Leaving space around the subject also gives you room for your own labels and arrows, which you can add in your slide software where you control the spelling.
Keep words out of the picture. AI image tools still misspell, invent and blur text, especially longer labels. Put headings, labels and explanations on the slide or page around the picture instead, where they’re accurate, editable and readable by screen readers. If you do ask for words in the picture, keep them to a few short labels and check each one letter by letter.
Check it as carefully as anything else you teach with. AI images can look convincing and still be wrong: an extra finger, a stethoscope worn backwards, a wiring diagram that couldn’t work, a chemical structure that doesn’t exist, safety equipment missing in a lab. Look at the picture the way an expert in your field would, read the “Check before teaching with it” note, and never use an AI picture as evidence of how something actually looks when the detail matters. For anatomy, equipment and procedures that students must get right, use an authoritative source.
Make it accessible from the start. Every picture that carries meaning needs alt text that says what students need to know from it, not just what it shows. The Image Creator writes a first draft for you; edit it so it fits how you’re using the picture, and keep it to a sentence or two. A purely decorative picture should be marked as decorative in your LMS instead. Don’t rely on colour alone to carry meaning, and check that any important detail is large enough to see when projected.
Show people respectfully and representatively. Over a unit, the people in your pictures should reflect the range of your students and the communities they’ll work with, in age, ethnicity, body size, disability and gender, without turning anyone into a stereotype. Clinical and care scenes should keep people’s dignity. If a picture leans on a stereotype, change it or make another.
Use only original pictures, and say they’re AI-made. The Image Creator won’t draw real people, brand logos or a named artist’s style, and you shouldn’t try to get around that. Never use a picture to portray a real student, patient or colleague, even loosely. Label AI-generated images as such in your materials (a short note in the caption is enough), so students know what they’re looking at, and check your institution’s policy on AI-generated content.
Know when a real photo or diagram is better. An AI picture is ideal for scene-setting, scenarios, illustrations of everyday situations and visual variety. It’s the wrong choice where accuracy is the point: a real X-ray, a historical photograph, a published data chart, an authoritative anatomical diagram. Use AI pictures to support your teaching, and keep the real thing for when students need to see what is real.