AI for Marketing

How to make an infographic with AI that stays on brand

By 8 min read

Pop-art comic illustration of a confident marketer in glasses holding up a poster of bright shapes and arrows beside a laptop, with a speech bubble saying On brand!

To make an infographic with AI, write the words and numbers yourself first, give the AI your brand colours and fonts, and then ask it for a few layouts to choose from. If the graphic carries figures, have Claude build it as code, so the text is typed rather than drawn. If it’s mostly illustration with a short headline, use an image model such as ChatGPT Images or Gemini, and proofread every word before it goes out.

This playbook takes about an hour the first time and 15–20 minutes per graphic after that (our estimate, not a measured result). It’s written for a LinkedIn post, but the same steps work for Instagram, a blog or a one-page handout. Product details are as of November 2026.

Which AI tool should you use for an infographic?

There are two ways to get an infographic out of AI, and they fail in different ways.

Built as code (Claude) Drawn as an image (ChatGPT Images, Gemini)
How it works Claude writes the graphic as HTML or SVG, so the words are real text The model paints the whole thing as a picture, letters included
Text and numbers Exactly what you gave it Usually right on current models, but can be misspelt or changed
Brand colours Exact hex codes Close, not exact
Look Clean and diagram-like More illustrated and polished
Editing Ask for a change, or edit the text yourself Ask for a change and hope the rest stays put
Best for Stats, step-by-step lists, comparisons, timelines A visual metaphor with a short headline

What each one offers right now:

  • Claude. Anthropic says custom visuals (diagrams, charts and interactive graphics drawn in the chat) are available to all Claude users on the web and desktop apps, but not on mobile. You can download a visual as an .svg or .html file, or use Copy as image. The feature is in beta and visuals aren’t saved automatically, so download or save each one as an artifact. Artifacts are on every plan, including Free, and need Cloud code execution and file creation switched on under Settings > Capabilities.
  • ChatGPT Images. OpenAI launched Images 2.0 on 21 April 2026 with better text rendering and layouts, on all tiers. Images 2.6 followed on 4 November 2026: OpenAI says it is 50% faster and keeps details better from one edit to the next. Free accounts have a separate image limit, and ChatGPT shows you when you hit it.
  • Gemini. Nano Banana 2 is Google’s free model and lists text rendering. Nano Banana 2.1, which Google says creates “accurate, detailed infographics and diagrams”, needs a paid Google AI plan. Google doesn’t publish fixed image numbers per plan. Its help page says paid plans get 2x to 20x the standard limits, and those limits can change without notice.

If you already pay for one of these, use that one. Our ChatGPT vs Claude vs Gemini comparison covers the wider choice.

What do you need before you start?

A one-page brand kit. It takes about 15 minutes once. Without it, every tool falls back on its own default colours and fonts. Put this in a document:

  • Colours as hex codes, with a job for each: background, text, one accent, one highlight. Four is plenty.
  • Fonts. The name of your heading font and body font. If they’re Google Fonts, Claude can use them directly in a code-built graphic.
  • Logo as a PNG or SVG file, and where it goes (bottom right, for example).
  • Your name or handle exactly as it should appear.
  • Two or three graphics you like, yours or anyone’s, for layout only. You’ll tell the AI to borrow the structure, not the branding.
  • Your voice guide, if you have one. Our one-page brand voice guide keeps the headline and captions sounding like you.

Keep it in a ChatGPT project, a Claude project or a Gemini Gem so you don’t upload it every time.

The playbook: one on-brand infographic in about an hour

Step 1: Pick one idea and get the facts straight (10 minutes)

An infographic carries one point. Start from something that already worked for you: a post that got comments, a question customers keep asking or a process you explain on every sales call.

If it includes figures, gather them now with the source next to each. Paste your notes into the AI and ask it to tidy them up, not to add anything:

Here are my notes for an infographic. Turn them into:
1. A headline of 8 words or fewer
2. 4 to 6 short points of no more than 12 words each
3. A one-line takeaway

Use only the facts in my notes. Don't add statistics, examples or claims.
If a point needs a fact I haven't given you, write [NEED SOURCE] instead.

My notes:
[paste notes, with a source link after every number]

Keep the [NEED SOURCE] line. AI models will fill gaps with confident, made-up detail (the glossary calls this hallucination), and on a graphic nobody sees the footnote.

Step 2: Lock the words before any design (5 minutes)

Edit the text until you’d post it as plain text. Cut each point to the shortest version that still makes sense. If a designer were working on it, this is the copy you’d hand them, and it’s the only version you’ll proofread against later.

Step 3: Build the graphic

Route A: built as code in Claude (best when there are numbers). Upload your brand kit and logo, then paste:

Build this as a single infographic for a LinkedIn post, portrait,
1080 x 1350 pixels.

Brand:
- Background [hex], text [hex], accent [hex], highlight [hex]
- Headings in [font], body in [font]
- My logo (attached) small in the bottom right, and "[your name or handle]" next to it

Words: use exactly the text below. Don't rewrite, shorten or add to it.
[paste your locked text]

Layout: borrow the structure of the attached example, not its colours or branding.
Make the headline the largest thing on the page, leave generous margins,
and keep it readable on a phone.

1080 x 1350 is a portrait 4:5 shape, a common choice because it fills more of a phone screen than a square. Download the result as .svg or .html, or use Copy as image for a quick PNG. If you want an exact pixel size, open the .html file in your browser and take a screenshot of it.

Route B: drawn by an image model (best for a visual metaphor). In ChatGPT or Gemini, upload your brand kit and an example, then paste:

Create a portrait 4:5 infographic for LinkedIn.

Style: flat illustration on a [hex] background, with [hex] and [hex] as accents.
Clean sans-serif text.
Words, spelled exactly as written and nothing else:
Headline: "[headline]"
Points: "[point 1]" / "[point 2]" / "[point 3]"

No other text, no numbers I haven't given you, no logos.

Keep it to three points on this route. The more words you ask an image model to draw, the more chances it has to get one wrong. Add your logo afterwards in Canva or any editor, rather than asking the model to draw it.

Step 4: Ask for three layouts, then give one note (10 minutes)

The first output is rarely the best one. Charlie Hills describes asking Claude for three genuinely different layouts with the same words and picking by eye, and it works in any of these tools:

Make 3 genuinely different layouts of this infographic. Keep the words
identical in all three. Show them side by side.

Pick one and give a single, specific note: “the headline is too small on a phone”, “too much empty space at the bottom”, “make the three points into a numbered list”. A single clear note is easier for the model to act on than a paragraph of feedback.

Step 5: Proofread it against your locked text (5 minutes)

Check every word and every number against the copy from Step 2, character by character. Route B needs this most. Then check:

  • Are the colours yours? Hold it next to your website.
  • Is the smallest text readable on a phone at arm’s length?
  • Is there anything you didn’t write, such as an extra label, a made-up figure, a stray logo or a fake chart axis?
  • Would you be happy for a customer to quote this back to you?

If anything fails, fix it at the step it came from rather than patching the image.

Step 6: Export and post (5 minutes)

LinkedIn’s help page sets a 5 MB limit per upload and allows up to 20 images in one post. When you post, click ALT on the image and describe what it says, so people using screen readers get the same information. Put the source for any figure in the post text, too.

Where does AI infographic-making break?

  • Long text on image models. Even on current models, a full paragraph or a dense table is where wrong letters and swapped digits creep in. Move dense content to Route A.
  • Charts with real data. Don’t let an image model draw a bar chart from your numbers: the bars may not match the values. Build charts as code, or in a spreadsheet, and drop them in.
  • Brand drift. Image models get colours close, not exact. If your brand colours are strict, use Route A or recolour in an editor.
  • Too many points. If it needs eight points, it’s a carousel or a blog post, not an infographic.
  • Watermarks. Google says Gemini adds an invisible SynthID watermark and a visible watermark to its images. That’s fine for most social posts. If you need a clean image for print, check your plan’s terms.
  • Lost work. Claude’s custom visuals aren’t saved automatically. Download each one, or save it as an artifact, before you close the chat.

What to do next

If your graphics need captions that sound like you, set up the one-page brand voice guide first, then load it alongside your brand kit. To give each infographic a slot in your posting plan, the ChatGPT social media content calendar playbook plans a month of posts in one sitting. The glossary has a short entry on AI design assistants if you’re comparing dedicated tools.

Sources

  1. https://support.claude.com/en/articles/13979539-custom-visuals-in-chat-and-cowork
  2. https://support.claude.com/en/articles/17153992-what-are-artifacts-and-how-do-i-use-them
  3. https://openai.com/index/introducing-chatgpt-images-2-0/
  4. https://openai.com/index/introducing-images-2-6/
  5. https://help.openai.com/en/articles/9275245-chatgpt-free-tier-faq
  6. https://gemini.google/overview/image-generation/
  7. https://support.google.com/gemini/answer/16275805
  8. https://www.linkedin.com/help/linkedin/answer/a527229
  9. https://charliehills.substack.com/p/claude-code-infographic-design-system