Alt text helps make visual content accessible to all students, including those using screen readers. Generative AI can support this by helping create and improve image descriptions.
For this activity you will use ChatGPT-4o to create alternative text for an image and assess its quality.
- Log into ChatGPT-4o. If you prefer not to create an account, feel free to use the sample image and alt text for your evaluation.
- Upload an image and use the following prompt: “Act as an expert in web accessibility and write concise and meaningful alternative text for this image so that it can be described via a screen reader.”
- Use the Writing Alt Text from the BCcampus Accessibility Handbook for Teaching and Learning and review the AI-generated alt text for accuracy, clarity, and relevance.
Complete this Activity
Share your assessment, including any refinements or improvements you made.
Example
Alternative text generated by ChatGPT: A small building stands beside two large leafless trees in the middle of a vineyard, with rows of pruned grapevines stretching into the distance under a clear blue sky.
Generative AI (GenAI) is rapidly transforming how we conduct research, scholarship, and teaching and learning. Permitted use of GenAI in these areas is constantly shifting and dependent on the organizations or instituions for which you are engaging. Before sharing any outputs of your GenAI use, it is important to consult policies and guidelines on accepted use. To learn more, review the resources on GenerativeAI and Open Scholarship.
Image Credit: Featured image Ie7 alt as tooltip by Sameboat (CC BY SA), Example image Mazet in vineyards, Mèze cf02 by Christian Ferrer (CC BY)
I uploaded an image of my experimental strawberry setup within UBCO’s Plant Growth Facility and received the following output from ChatGPT: “Rows of strawberry plants growing in white Grodan rockwool slabs inside a controlled indoor environment. Each plant is connected to a drip irrigation system with black tubing”.
Following the BCcampus Accessibility Handbook for Teaching and Learning guidelines, this output was not missing anything.
I uploaded a photo of my cat and received the following: “Close-up of a black and white cat with bright green eyes, resting on a colorful, patterned blanket. The cat is looking slightly upward with a calm and alert expression.” Based on the BCcampus guidelines, it seems to work well for alt text. It even included the term “close-up” which is a recommendation of the guidelines.
I uploaded a very old photo of myself as a naval cadet with two peers. The description came back as “Three young women in matching navy uniforms with berets sit on the edge of a boat, smiling and adjusting their socks—two wearing mismatched blue and white socks. A man stands behind them, and another is visible inside the boat’s cabin.” While we were sitting on a hatch cover, the description was fairly accurate and not missing any key information per BCcampus Accessibility Handbook for Teaching and Learning
I uploaded a figure that I use in a course to help show the risk associated with different earth systems from climate change. Over all the description was decent and matched the information required per the Accessibility Handbook.
While the description was good, it did struggle to explain the structure of the scientific diagram and called one part of it “planetary Boundaries” which felt like an odd use of what I would call concentric rings or circle showing different risk levels. However, I would say that it mostly did a good job and with a few simple edits I would gladly include this text to help students who need screen readers.