Charts in Chat Are the Missing Layer AI Products Keep Ignoring

I came across a post on Hacker News from thingsilearned about building charts specifically for chat interfaces. The core observation is simple: AI chatbots are hugely popular, but they almost never respond with charts, even when the question is about data. The post outlines a vision for how chart generation could become more native to conversations.

It's a compelling argument, and I agree that charts are underused. But I think the author is being too generous about why, and too narrow about where the opportunity actually is.

As of early 2025, hundreds of millions of people use AI chatbots every month. That scale is unprecedented. And yet if you ask ChatGPT or Claude a question about a data trend, you typically get a paragraph of text or a raw table. The occasional code interpreter can generate a chart, but it's clunky and not conversational. Most of the time, no visual appears.

Is that because LLMs are inherently text-only? The author suggests that training data and architecture constrain outputs. But our research shows that's not the full story. Multimodal systems already exist and can produce images and charts when integrated with tools. The limitation is more about product decisions than fundamental technology. Chatbot makers have prioritized text because it's easier, but the gap is closing.

That doesn't mean the problem isn't real. It is. And it's quantified in our dataset of 25,201 tracked problems across 89 industries. Among those, communication-related complaints alone number at least 30, and they consistently point to one frustration: extracting insights from text-heavy AI responses is harder than it should be.

The author's post uses a broad brush. But the demand for charts in chat is not uniform. Some industries feel the pain much more acutely. Healthcare, for instance, is full of data analysis tasks where a quick visual could speed up decisions. Finance is another. Financial data is inherently chart-friendly, and our data shows high demand for better visualization in financial chat tools. These are verticals where a simple chart integration could immediately reduce friction.

What the original post doesn't cover is personalization. Most chart solutions for chat are static. Submit a query, get a generic bar chart. Our data shows users get frustrated with one-size-fits-all representations. They want charts that adapt to the question's context, their preferences, and the conversation's flow. That's a harder technical problem, but it's where the value will be.

So here's my take for indie hackers and builders: don't build a general chart plugin for ChatGPT. Build a chart layer for a specific vertical. Healthcare analytics, financial reporting, e-commerce inventory. Pick a niche, understand how people there use chat tools, and deliver visual answers that feel native to their workflow. The market is huge and the pain is documented.

For example, a healthcare chatbot that, when asked about patient admission trends, shows a line chart with contextual annotations instead of a paragraph of stats. That's not a nice-to-have. That's an obvious improvement that saves time and reduces errors.

One thing I'd caution: don't chase the platform giants. They will eventually add basic chart generation to their chat products. The real opportunity is in specialized chart experiences that understand a domain deeply. That's not something a general-purpose assistant will nail quickly.

We're tracking this shift. As more teams try to build chart-enabled chat, new problems will emerge and old ones will evolve. We'll keep documenting them. If you're working on something like this, the demand signals are already there. You just need to look at the right data.

The original post is worth reading for its thoughtful approach to chart design. But it's just the beginning. The real story is that charts in chat are a solution in search of a more specific problem. Find that problem, and you have a product.

You can read the original piece here.