Disability representation in AI-generated images has improved significantly with recent models. When I tested generating images of people with visible disabilities using Midjourney in 2024, I was overall disappointed in the results. Even when I presented about inclusive images for the Accessible and Inclusive Design Conference last summer, I had significant problems with disability representation. But now, both the Nano Banana 2 and GPT Image 2 models can generate better images of people using assistive devices such as wheelchairs, forearm crutches, hearing aids, and prosthetic limbs. I still had a few issues, but we now seem to be at a point where we can generate much more realistic images of people with disabilities. Representing of people with disabilities has always been a challenge for elearning, but AI images can be one solution for more inclusive images. In this post, you’ll see my prompts and results for disability representation.
For each pair of images below, the left or first image (depending on your screen size and email program) is Nano Banana 2, and the right or second image is from GPT Image 2 (the newest model in ChatGPT). In general, I prefer the Nano Banana 2 images; they feel a bit less staged and less like stock photos to me.
If you’re viewing this on my site and not via email, you can click each image to expand it.
Wheelchairs
Wheelchairs have been a problem for AI-generated images in the past, partly because so many stock images show able-bodied actors using temporary wheelchairs incorrectly. I was overall pleased with the results for both a “sporty wheelchair” and an “electric wheelchair.” There are still some flaws though—check the spokes on the wheels and the foot rests.
Style: Editorial photoSubject: A Puerto Rican man, mid 20s, in a sporty wheelchairAction: Working at a computerSetting; A modern open office in a large city, high rises visible through the windows, other employees in casual clothes working nearbyLight: Natural light from large windows
Style: Editorial photoPerson 1: A Black woman, mid 40s, average weight, in an electric wheelchair. Wearing A pink t-shirt and jeansPerson 2: A Korean woman, mid 40s, wearing a long gray t-shirt and black leggingsPose: Talking to each other, gesturing with their hands, friendly conversationSetting: A trail in a suburban park with a playground in the backgroundLight: dappled afternoon light, partial shade from trees
Forearm crutch
Forearm crutches are one of the objects I’ve never been able to generate accurately in Midjourney. Even a year ago, ChatGPT struggled to show them accurately while also fulfilling the rest of the prompt. It’s much closer now. We often default to wheelchairs to represent disabilities in images, but mobility aids like leg braces and forearm crutches should be shown too.
Style: Editorial photoSubject: Native American man, mid 50s, graying hair in a long braidClothes: A red button down shirt and gray slacksDetails: Using a forearm crutchAction: Presenting at a small team meeting, explaining and gesturing with one handSetting: A small modern office conference room, a projector screen with a column chart behind himLight: overhead florescent lights
Hearing aids
When I wanted an illustration of a woman wearing a hearing aid in January 2025 for a blog post, it took me over a dozen variations and edits to get it right. I’ve also had issues in the past where the AI models insisted that anyone wearing a hearing aid must be older and have gray hair, no matter what age I prompted for. This time, I got two plausible behind-the-ear hearing aids on my first attempt.
Style: Editorial photoSubject: White woman, mid 20s, hair pulled back into a messy bunClothes: A dark blue business casual blouseDetails: Wearing a behind-the-ear hearing aid Pose: Typing, working at a computer with two monitorsSetting: An office cubicle with gray wallsShot: Medium shot, head and shoulders
Cochlear implants
I think the cochlear implants here aren’t quite accurate; I don’t think they’d meet the approval of someone really familiar with these devices, especially at higher resolution. But this is closer than I’ve been able to get in the past. Because the implant is black in the Nano Banana image, it’s hard to detect.
Style: Editorial photoSubject: Indian woman, mid 30s, braided hairClothes: A red tunicDetails: Wearing a cochlear implantPose: Sitting behind a desk, working on paperwork, focused expressionSetting: A private office in a bank, dark wood deskShot: Close shot, focus on the side of the head
Prosthetic leg
I set up my prompt for this image to put the subject in shorts so the prosthetic leg is visible without any weird artifacts like hiking up someone’s pant leg to show the prosthesis. I was also curious how the models would handle the setting of the YMCA; both models included the logo.
Style: Editorial photoSubject: White man, mid 40s with a prosthetic legClothes: A black athletic shirt and gray athletic shortsPose: Playing basketball, passing the ball to a friendSetting: A gymnasium in the YMCALighting: Overhead lightsShot: Wide shot
Upcoming webinar on inclusive images
Disability representation is one of the topics I’m going to address in my upcoming Training Mag Network webinar, “Better Prompts, Better Representation: Inclusive AI Images for Learning.” This free webinar will focus less on the tools for AI images and more how to explicitly prompt for images that are more inclusive of gender, race, disability, and more. AI image models can reinforce stereotypes if you don’t prompt to counter their natural tendencies. That makes it important to know how to prompt well and what problems to watch out for.
Register for my free webinar on May 20, 2026 at 12 PM PT/3 PM ET.
What are your challenges with disability representation?
What challenges have you found with representing disabilities using AI images? Comment on this post or reply to this email. I’d love to have some more specific examples to test to see if it’s possible to generate with the most current models. What are the inclusive images you wish you had for your training?
