New research shows how people talk to AI chatbots

New studies show people mainly use text to talk to AI. This is like asking a question and getting an answer, which is the most common way to use AI today.
By Newsroom | Technology, Artificial Intelligence |

New research, emerging from late 2024 and mid-2025, is attempting to map the complex territory of how humans engage with Large Language Models (LLMs). These studies, often drawing from broad surveys and specific taxonomies, highlight a foundational interaction mode: text-based conversational prompting. This basic input method, according to work presented at the CHI Conference on Human Factors in Computing Systems, underpins much of the current human-LLM dynamic.

Categorizing Engagement

Beyond simple prompting, researchers are building frameworks to understand the nuances of these interactions. One proposed taxonomy differentiates between various modes of interaction, including straightforward text-based exchanges and more structured input via user interfaces (UIs). The research points to the existence of "Mode 1.1. Text-based Conversational Prompting" and "Mode 1.2. Text-based Conversational Prompting with Reasoning," alongside UI-based approaches like "Mode 2.1. UI for Structured Prompts Input," "Mode 2.3. UI for Iteration of Interaction," and "Mode 2.4. UI for Testing of Interaction." This suggests an evolving relationship, moving from pure dialogue to more controlled and deliberate methods of engagement.

Read More: OpenClaw Uses Local AI Models Instead of Cloud for Better Privacy

Broader Contexts and Applications

The discourse around human-LLM interaction extends beyond purely technical classifications. Various publications, including those appearing in proceedings from 2023 and 2024, have explored its implications across different fields. This includes:

  • Responsible AI: Adapting user experience (UX) practices to address the challenges posed by responsible AI development.

  • Collaboration and Creativity: Investigating human patterns in interacting with LLMs to foster better collaboration and creative output.

  • Medical Applications: Reviewing human-AI interaction within machine learning, particularly for clinical decision support systems and the impact of chronic conditions on patient-AI engagement.

  • Education: Examining the integration of generative AI, conversational agents, and chatbots in educational settings.

One survey from June 2025 also touches upon the integration of LLMs with knowledge-based methods, indicating a growing interest in combining these powerful tools with existing data structures, though no specific data was utilized in that particular research.

Frequently Asked Questions

Q: How do people talk to AI chatbots based on new research?
Most people talk to AI chatbots using text, like typing a question and getting an answer. This is called 'text-based conversational prompting'.
Q: What are the different ways people can interact with AI?
Besides simple text chats, people can use special screens (UIs) to ask AI questions in a more organized way or to test how AI works. Some methods also let people ask AI to explain its thinking.
Q: What areas are being studied for human-AI interaction?
Studies are looking at how AI is used safely, how it helps people be creative, how it can help doctors, and how students use AI in school.
Q: How is AI being combined with other knowledge methods?
Some research from June 2025 shows people are interested in linking AI with existing information systems, but this specific study did not use any real data.