Skip to content

Creating great infographics with ChatGPT

I’ve been playing around with creating infographics with ChatGPT. I’ve found that it can make surprisingly decent infographics, even for someone (like me) with little design experience. But there’s a downside: unless you give it enough direction, you can easily end up something that looks generic and lazy.

Here’s what I’ve learned:

1. Know what ChatGPT does well (and poorly)

What ChatGPT is good at when creating infographics:

  • Speed: An infographic it would have taken me 2 hours to design in Canva is ready in less than a minute.
  • Low barrier to entry: I’m creating decent infographics with limited design skills.
  • Idea generation: ChatGPT suggests examples and visual approaches I hadn’t considered.

Where ChatGPT is limited when creating infographics:

  • Generic design: You’ve probably seen the default ChatGPT infographic style by now. It’s polished but very generic, often using the same color palette and visual elements.
  • Visual clutter: ChatGPT tends to use a lot of visual elements, whether or not they support the content.
  • Hallucination: As with any AI-generated content, ChatGPT can make up facts, statistics, labels, or other information.
  • Poor object-level editing: I discovered that, unlike Canva, ChatGPT doesn’t reliably let you change one element while leaving everything else alone. Often when I asked for a specific revision, it unexpectedly changed other parts of the design.

2. Give ChatGPT the right ingredients

Giving ChatGPT descriptive, detailed instructions is a key to success for almost all content. For best results, share these elements with ChatGPT so it knows what you’re looking for:

Format:

  • Is an infographic the best format for this information, or would it work better as a flowchart, checklist, etc.?
  • If you’re working from existing content, which parts of the text would benefit most from being visualized?

Content:

  • Who is the infographic for, and what are their goals?
  • What are the main points and key takeaways you want to highlight?
  • If you’re working from existing copy: How could this content be optimized to work in an infographic format?

Design:

  • What format do you have in mind, or where do you plan to share the infographic? (E.g., a LinkedIn carousel, an Instagram post, a presentation slide)
  • What style elements should ChatGPT use? Specify things like the color palette, font, and overall style or vibe.
  • Are there any specific images, icons, or logos that need to be included?
  • Do you have a sample of an infographic you particularly like or find especially useful? Sharing examples often helps ChatGPT know what you’re after.

3. Evaluate the output

Once ChatGPT has rendered an image, compare the infographic against the requirements you started with. Did ChatGPT accomplish what you asked?

  • Is all of the information accurate?
  • Is it easy to scan and understand?
  • Do the images and icons reinforce the message?
  • Is anything unnecessary or distracting?
  • Would you be comfortable publishing it as-is?

You can also use ChatGPT or another AI tool as a second set of eyes: “Pretend you’re an experienced information designer. Critique this infographic. What would you change before publishing?”

4. Iterate until you get what you want (or at least close)

I was almost never satisfied with the first version ChatGPT gave me, even when I had provided detailed instructions. Here are some tips that helped me revise with ChatGPT:

  • Start with the content, not the design: Make sure the main message, sections, and wording are right. If not, change those first before tackling the visuals. Trust  me—when I started with the visuals, I had to re-do design elements if I changed the underlying content.
  • Change one thing at a time: I found that the results were better when I revised one element at a time versus trying to batch my edits together. (When I instructed it to change this AND that, it often ended up changing one but not the other, or changing a third thing I didn’t ask for.)
  • Tell ChatGPT both what to change and what to leave alone: For example, “Change only the title. Keep the layout, colors, icons, fonts, and all other text exactly the same.” Double check after each revision to make sure nothing shifted unexpectedly.

5. Add human touches to keep your infographic from feeling like AI

If you provide enough information, your infographic is less likely to look generic and AI-rendered. Here are a few additional ideas for making your design deliberate:

  • Start with your own insight: The best infographics communicate an idea, not just information. If ChatGPT is deciding both what to say and how to present it, the result will often feel generic.
  • Add something only you know: Include an original analogy, an observation from your experience, a quote, or an example from your own work. AI can summarize common knowledge, but it can’t access your perspective.
  • Don’t fill every inch: AI tends to maximize the canvas, cramming visual elements into every possible space. I found that it particularly helpful to instruct ChatGPT to provide white space around the edges.
  • Avoid visual clichés: If every point is accompanied by a light bulb, rocket ship, target, or handshake emoji, the infographic feels like corporate clip art.

6. But don’t sweat it

The goal isn’t to hide the fact that you used AI. You want to create something that’s accurate, useful, and easy to understand. Not everything AI generated is AI slop – if you put thought, originality, and judgment into it, it’s not slop. Your readers (for the most part) will care less about whether something was AI generated than whether it was helpful and effective.  

An infographic created with ChatGPT about creating infographics with ChatGPT

I followed my own tips to create an infographic based on this blog post:

Is it perfect? Eh, maybe not. We could find things to criticize about this infographic. However, it’s way better than anything I could have designed myself in Canva. It took me about 20 minutes to draft the creative brief and go through about four iterations until I liked the output. I think that makes it at least worth a try!

Leave a Reply

Your email address will not be published. Required fields are marked *