Claude Can Now Draw You an Answer

Ask Claude a question, and it can now answer with an interactive chart. A real one, with curves, axes, clickable elements. Anthropic announced March 12 that its AI can generate visualizations directly in the conversation thread. No coding required, no plugins. You ask your question, and if a visual helps explain the answer, Claude builds it on the spot.

This isn't just a neat trick. It's a fundamental shift in how AI communicates.

How it works

These visualizations aren't images. Claude generates HTML and SVG in real-time, like a front-end developer coding a small widget on the fly. The result: charts that load fast, adapt to your screen size, and respond to interaction.

Ask Claude to explain compound interest. Instead of a wall of text and numbers, you get an interactive curve where you can adjust the rate, duration, and principal, watching the result update in real-time. Want to understand the periodic table? It'll build you one where every element is clickable.

These visuals appear directly in the chat, not in a side panel like the Artifacts we've had since 2024. Artifacts are permanent content you can save and share. Inline visuals are ephemeral: they exist in the conversation flow, changing or disappearing as the exchange evolves. Think of it like a diagram sketched on a napkin during dinner, not a document filed on your desk.

Claude decides on its own when a visual would be clearer than text. You can also force it with a simple "draw me a diagram" or "visualize this data." Flowcharts, structural diagrams, business data charts, step-by-step guides: the range is broad.

The feature is in beta for all users, including the free plan. Worth highlighting, since that's rare. For complex visualizations, though, Anthropic recommends Opus, which means a Pro subscription.

The ChatGPT matchup: two opposing philosophies

This is where it gets interesting. OpenAI launched a similar feature on March 10, two days before Anthropic. It's called "dynamic visual explanations," and on paper, it sounds identical.

The philosophy is radically different.

ChatGPT offers a catalog of over 70 pre-built visual modules, focused on math and science. Widgets designed by humans, with sliders calibrated to explain trigonometry or fluid mechanics. It's polished, reliable, but fixed. If your topic isn't in the catalog, there's no visual.

Claude generates everything on demand. Any topic. No catalog, no pre-fabricated modules. The model invents the visual based on your question. Think of a restaurant with a fixed menu versus a chef cooking à la carte: one is predictable and consistent, the other is flexible but occasionally uneven.

OpenAI is betting on control and guaranteed quality. Anthropic is betting on the model's ability to improvise. Both approaches have merit, and it's not obvious which will win long-term.

What's still missing

We're in beta, and it shows. The feature is limited to web and desktop: no interactive visuals on iOS or Android yet. For a tool positioning itself as accessible to everyone, that's a serious limitation.

Quality is also inconsistent. On well-defined topics, results are solid. On more creative visualizations, it depends on the model and how you phrase the request.

Then there's the data reliability question. Claude can draw you a gorgeous chart with completely fabricated numbers. Visualization makes information more convincing, which is an asset when the data is accurate but a real risk when it's not. You can't judge a book by its cover, even when the cover is a polished SVG chart.

From AI that explains to AI that shows

Since they appeared, AI chatbots have communicated primarily through text. Paragraphs, bullet lists, code blocks. With interactive visualizations, we're seeing the emergence of a new language: AI doesn't just explain anymore, it shows. Each visual is unique, generated to answer your specific question.

The fact that Anthropic and OpenAI launched similar features 48 hours apart says something about the state of the race. Both companies see the same destination, they're just taking different paths to get there.

In practice, next time you ask Claude to explain a complex concept or forecast a budget, try asking for a visual. And if you're on the free plan, now's the time to test it: for once, the most interesting feature isn't behind a paywall.