Conversational artificial intelligence is entering a new phase of maturity. Until now, assistants like Claude, ChatGPT, and Gemini have primarily distinguished themselves through their ability to generate text, code, or complex summaries. But a more fundamental shift is taking shape: the transition from AI that describes to AI that demonstrates. With its latest update, Anthropic introduces a key capability: Claude can now generate graphs, diagrams, and interactive visualizations directly within the conversation—and, most importantly, decide for itself when these representations are relevant.
This feature has been rolling out gradually since early 2026 within the Claude interface, accessible via the web (claude.ai) as well as in certain business integrations. It is enabled by default for users with recent versions of the model, notably Claude 3.5 and its updates, primarily on paid plans such as Claude Pro and Team. The initial rollouts took place in the United States, with a gradual expansion to Europe, including France, following a similar approach to that of the model’s other advanced features. This rollout method, now standard in the generative AI ecosystem, allows for adjusting usage patterns and observing user behavior before full-scale deployment.
According to Gartner, more than 75% of interactions with AI systems could become multimodal by 2027, integrating text, images, audio, and visualizations into a single interface1. Claude fits perfectly into this trend, bridging the gap between conversational AI and analytics and data visualization tools.
A major development in conversational interfaces
Before this update, Claude was already capable of generating charts on demand. Users could request a chart or visualization, which would then appear in a separate area, often as a side panel. This approach remained similar to a traditional model, where the user took the initiative and controlled the structure.
The new feature introduced by Anthropic fundamentally changes this approach. Now, visualizations appear directly within the chat thread, on the same level as the text. Users no longer need to switch between different interfaces; the analysis unfolds as part of the ongoing conversation.
This trend is particularly evident in everyday use. A simple question about market trends, a statistical distribution, or a scientific structure can trigger the appearance of an interactive graph without any specific action on the user’s part. The interface thus becomes a hybrid space where explanation and visual representation coexist seamlessly.
An AI capable of choosing the right format for an explanation
One of the most significant aspects of this development is Claude’s ability to choose the most appropriate format on its own. In some situations, text alone is sufficient. In others, a visual representation becomes essential.
Anthropic demonstrates this capability through several real-world examples. During a discussion about the periodic table, Claude can generate an interactive version where each element is clickable and provides additional contextual information. In another example, a question about a building’s structure leads to the creation of a diagram explaining the load distribution.
These visualizations are not static. They evolve as the conversation progresses. A new question can alter the graph, refine the data, or change the way it is displayed. This allows the user to engage in a dynamic exploration process.
This approach gives Claude access to advanced data visualization tools while maintaining the simplicity of a conversational interface. According to McKinsey, companies that incorporate data visualization into their decision-making processes see a 20% to 30% improvement in decision-making speed2.
From content creation to interactive analysis
This feature transforms the very nature of the tool. Claude is no longer limited to producing content; it becomes an analytical environment.
In practice, the user can:
- ask a question and instantly get a visual representation
- refine a graph by rephrasing the query
- explore different scenarios without leaving the interface
- turn an abstract explanation into a concrete visualization
This trend is part of the growing prominence of data storytelling, where understanding relies on the visual presentation of information. According to MIT Sloan, decision-makers are significantly more effective when information is presented visually rather than in text form3.
In this context, Claude serves as an accessible analytical tool that requires no technical expertise in data visualization or programming.
A fundamental difference from existing approaches
Anthropic already offers a system called Artifacts that allows users to generate documents, interfaces, or persistent applications. These elements are designed to be saved, shared, or exported.
The new visualizations follow a different approach. They are contextual, tied to the conversation, and evolve alongside it. They are not intended as deliverables, but rather as tools for explanation.
This distinction is essential. It reflects a distinction between two types of use:
- sustainable productions intended for use or distribution
- fleeting visualizations, designed to support reasoning
As for other players, the capabilities exist but are generally user-initiated. ChatGPT can generate graphics, particularly through its advanced tools or data analysis, but most often requires an explicit prompt or a dataset. Google Gemini also offers visualizations, but in contexts that are still relatively limited.
Here, Claude introduces an additional capability: the ability to anticipate the need for visualization, which represents a significant advancement in the design of AI interfaces.
A tool that combines productivity and learning
The implications are wide-ranging. In education, interactive visualization makes it easier to grasp complex concepts. A diagram or graph can make an abstract concept immediately understandable.
In professional settings, this capability allows users to quickly explore data, test scenarios, and generate visual analyses in a matter of seconds, without needing specialized tools.
According to Deloitte, nearly 49% of companies that have adopted advanced AI solutions report a significant improvement in their understanding of internal data4. Directly integrating visualization into conversational interfaces could amplify this trend.
AI thus acts as an intermediary between data and decision-making, facilitating interpretation and communication.
Ethical Issues and Limitations of Automated Visualization
This development, however, raises several questions. The first concerns reliability. A visualization can give the impression of accuracy, even if the data is incomplete or the assumptions are questionable.
The second point concerns bias. The choice of chart type, scale, or structure can influence the user’s perception. A visual representation is never entirely neutral.
The question of accountability also arises. When a decision is based on an automatically generated visualization, it becomes essential to understand the source of the data and the choices made regarding its presentation.
These issues fall within emerging regulatory frameworks, particularly in Europe with the AI Act, which emphasizes the transparency and explainability of artificial intelligence systems5.
In this context, the use of these tools must be accompanied by a critical analysis and an understanding of the underlying mechanisms.
Toward AI That Orchestrates Knowledge Formats
With this development, Claude exemplifies a broader transformation in artificial intelligence. Models are no longer limited to generating text; they are now capable of choosing the most appropriate format for conveying information.
This capability could be extended to other formats, such as audio, video, or simulation. The conversational interface would then become a single entry point for exploring, analyzing, and understanding complex systems.
The question remains open. How far can these systems go in automating explanations, and how can we preserve the user’s critical role in interpreting the results?
A gradual transformation of interfaces and usage patterns
By integrating interactive visualizations generated in real time, Anthropic positions Claude at the intersection of conversational AI and analytical tools. This evolution goes beyond a mere functional improvement; it reflects a more profound transformation in the way information is produced, structured, and understood.
AI no longer simply answers a question. It now chooses how to answer it, using the most appropriate format—whether text, a diagram, or a graph. This ability to combine different modes of representation paves the way for more intuitive interfaces that can guide the reasoning process rather than simply delivering a result.
In both educational and professional settings, this shift could redefine how technology is used. Access to visualization becomes immediate and seamless, and no longer requires specific skills. Analysis becomes more interactive, and understanding develops gradually, building as the exchange unfolds.
One key question remains. As these systems become increasingly autonomous in organizing information, how can we preserve users’ critical thinking skills when faced with representations that, due to their apparent clarity, may influence interpretation? The value of these tools will therefore depend as much on their performance as on how they are used, scrutinized, and contextualized.
Learn more
Claude’s evolution toward real-time visualization capabilities illustrates a broader transformation in AI interfaces, which are increasingly focused on interaction, exploration, and data understanding. On a related topic, check out our article “Claude Code Voice: Anthropic Finally Lets You Control Your Code With Your Voice, ” which analyzes how new ways of interacting with AI are redefining professional uses, from programming to data analysis.
References
1. Gartner. (2023). Top Strategic Technology Trends: Multimodal AI.
https://www.gartner.com
2. McKinsey & Company. (2023). The Data Visualization Advantage.
https://www.mckinsey.com
3. MIT Sloan. (2022). Data Visualization and Decision Making.
https://mitsloan.mit.edu
4. Deloitte. (2024). State of AI in the Enterprise.
https://www2.deloitte.com
5. European Commission. (2024). AI Act Overview.
https://digital-strategy.ec.europa.eu