A neat little project came my way recently. A colleague shared a screenshot of a flowchart outlining the Academic Integrity process and asked for my input on how to make it more accessible. Academic integrity, of course, has taken on renewed significance in the era of generative AI. As colleges grapple with questions about authorship, originality, and the appropriate use of AI, helping students and faculty understand academic integrity policies and procedures has become especially important.
The flowchart had been created in Canva, and while it worked well visually, getting it out of Canva in a useful, accessible format was another matter. In a sense, the diagram was a prisoner of the platform.

The original Academic Integrity flowchart in Canva. Exporting it as a static image limited its accessibility, navigation, and scalability.
Attempts to add alternative text didn’t carry over when exporting to PDF. Exporting as a PNG flattened everything into a static image, with no keyboard-accessible elements or functional links. And when you enlarged it to read the smaller text, it became increasingly blurry. The visual design was there, but much of the functionality needed to make it accessible wasn’t.
My first thought was to create a narrated screencast walking viewers through the process. That would have provided another way to access the information. But then I started wondering whether we could do something more interesting. What if, instead of trying to work around the limitations of the exported image, I could rebuild the flowchart as an interactive webpage? That seemed like a perfect opportunity to experiment with vibe coding and generative AI. As the project evolved, my colleague described it as a great example of creative accessibility, a phrase I think captures what we were trying to accomplish.
From a PNG to an Interactive Webpage
I decided to give ChatGPT Codex a try. I could have explored a similar approach with Claude or even BoodleBox, but Codex was the tool I happened to use for this project. I started with the original screenshot and described what I wanted to accomplish. Rather than trying to extract usable code from Canva, I worked with Codex to reconstruct the diagram as a scalable vector graphic (SVG), preserving its content and decision-making pathways while freeing it from the limitations of a static image.
Unlike a PNG, an SVG can be enlarged without becoming blurry. But I wanted to go further than simply making a sharper version of the diagram. Working back and forth with Codex, I began reconfiguring the flowchart and building an interactive experience using HTML, CSS, JavaScript, and SVG. The idea was to give visitors two ways to understand the process: a step-by-step decision guide and a visual flowchart they could zoom into and explore, with working links to relevant policy documents.
The HTML provided the page structure, CSS handled the layout and styling, JavaScript supported the interactions, and SVG made it possible to retain a crisp, scalable diagram. The process involved plenty of experimentation. I’d describe what I wanted, examine the results, identify something that wasn’t quite right, and work through another revision. Sometimes it was a visual adjustment. Other times it involved navigation, links, or making sure the relationships between different parts of the process were represented correctly.
The AI was doing much of the coding, but I was making the design decisions, testing the results, and deciding what needed to change. That’s an important distinction when we talk about vibe coding. It’s not simply asking AI to make something and accepting whatever comes back. At least in my experience, the interesting part is the iterative process of developing an idea, seeing what’s possible, and refining it.
Getting It Online
Once I had a working prototype, I uploaded the webpage package to my Apache web server, running in a LAMP hosting environment through Reclaim Hosting. Having access to my own web hosting environment made this part especially easy. I could upload the files, test the resource in a browser, make changes, and share a working link with colleagues without having to wait for an institutional website update. The prototype lived on Skidmore Domains, where others could try it and provide feedback. After a few rounds of revisions, I started thinking about how this could become something more permanent.
Although the prototype worked as a standalone site, the Academic Integrity process really belonged on the College’s official website. That’s where students and faculty would naturally expect to find it. I took a first pass at incorporating the College’s colors and visual identity to demonstrate how the resource might fit into the existing website. It wasn’t intended to be a finished institutional design, but it helped show what was possible. With support from colleagues, I then shared the complete HTML, CSS, JavaScript, and SVG package with the College’s web team. They reviewed the files, incorporated the resource into the College’s content management system, and adapted it to the newer website design. Once it was migrated, I tested the page on Mac and Windows computers and my iPhone to make sure the functionality had survived the move. A few final adjustments later, the resource was live and linked from the Academic Integrity section for both instructors and students.
Explore the finished Academic Integrity Process
What started as a screenshot and a question about accessibility had become an interactive resource on the College’s website.
Where Universal Design for Learning Comes In
One of the things I particularly liked about this project was how naturally it connected to Universal Design for Learning (UDL). I’ve been working with colleagues on UDL and thinking about how we can make accessibility a more intentional part of the design process, rather than something we address after a resource has already been created. CAST’s UDL Guidelines 3.0, particularly its guidance on supporting multiple ways to perceive information, provide a useful framework for thinking about this.
In this case, the goal wasn’t simply to make the original image bigger or sharper. It was to provide different ways for people to access and understand the same information. Some visitors might prefer to follow a decision guide one step at a time. Others might want to see the entire process visually, zoom in, and explore the relationships between its different parts. One of the most useful suggestions during our revisions was a small change to the introductory language. Instead of telling visitors to follow the decision guide and then explore the flowchart, we changed it to say they could follow the decision guide or explore the visual flowchart.
That little word, or, made a difference. It emphasized choice, which is really at the heart of UDL. We don’t all need to encounter information in exactly the same way. Of course, creating multiple pathways doesn’t automatically mean a resource meets every accessibility standard. Keyboard navigation, screen-reader compatibility, reading order, and other technical considerations still matter and need to be tested. But I think there’s something valuable about approaching accessibility as a design opportunity rather than simply a compliance exercise.
Creative Accessibility and Vibe Coding
I’ve been thinking more about my colleague’s description of this work as creative accessibility. The original request was straightforward: How can we make this flowchart more accessible? But once I started experimenting, the question became: What else could this resource be? That’s where generative AI became particularly useful. I didn’t have to sit down and write all the HTML, CSS, JavaScript, and SVG code from scratch. Instead, I could concentrate on what I wanted the experience to accomplish and work with Codex to develop it.
That doesn’t mean technical knowledge or critical judgment suddenly become irrelevant. Quite the opposite. You still need to understand the problem, recognize when something isn’t working, and test what the AI produces. And accessibility is an area where we need to be especially careful about assuming that generated code is correct simply because it appears to work. Still, the ability to move from an idea to a functioning prototype through a conversational process opens up some interesting possibilities.
For educators, instructional designers, and others who may not have extensive programming experience, it creates opportunities to experiment with interactive learning resources that might previously have required specialized development support. It also connects to something I’ve been exploring in my teaching and work with AI literacy: creative agency. The goal isn’t to surrender the creative process to AI. It’s to use these tools to extend what we can do, while remaining responsible for the decisions and outcomes.
AI Doing Some Good
There’s plenty to be concerned about when it comes to generative AI, including its environmental footprint, reliability, implications for intellectual labor, and effects on teaching and learning. I think those concerns deserve serious attention. At the same time, I don’t want us to overlook practical examples where these tools can help people solve real problems. This project didn’t begin with a desire to use AI. It began with a colleague who had created something useful and wanted to make it more accessible.
Generative AI helped me experiment with a solution. My web hosting environment gave me a place to develop and share it. Colleagues helped refine the experience, and our institutional web team took the finished package and integrated it into the College’s website. It was a relatively small project, but I think it illustrates something worth exploring. Accessibility doesn’t always have to mean adapting an existing resource to meet a minimum set of requirements. Sometimes it can lead us to rethink the resource altogether.
And perhaps one of the more useful applications of generative AI is helping us turn those ideas into something tangible. Not AI replacing the designer or educator, but AI helping us explore possibilities we might not otherwise have had the time or technical capacity to pursue.
For me, that’s a pretty good example of AI doing some good.

