Most people use AI chatbots for tasks, not companionship: what Anthropic’s Research reveals

Study shows most people don’t use chatbots for companionship.

What Anthropic found:

When Anthropic, the company responsible for developing the Claude chatbot, decided to analyze more than 4.5 million user conversations, they aimed to discover whether people genuinely seek friendship or emotional support from their chatbot interactions, or if they use these tools for other reasons. Their analysis revealed that, in reality, almost no one relies on Claude primarily for building a sense of companionship or for having emotionally intimate conversations, even though the platform could, in theory, support such interactions. Instead, what they found is that users almost always turn to the chatbot in order to complete practical tasks, find specific information, or get help with various problems that arise in everyday life.

Why this is different from the news

Although many stories in the media often focus on unusual cases in which people form strong emotional attachments to chatbots or begin treating these systems almost like friends or even romantic partners, these stories do not reflect the majority of user experiences, according to Anthropic’s findings. Media reports that emphasize companionship or romance with AI tend to get widespread attention, which can create a misleading impression that these kinds of relationships are common or growing, even though the data shows they are very rare. In reality, the overwhelming majority of users interact with chatbots for reasons that have little or nothing to do with seeking companionship.

How people really use chatbots

In most of the conversations reviewed, users approached Claude with clear, practical goals, such as getting help with drafting emails that need to sound professional, understanding complicated subjects that might be difficult to research alone, troubleshooting coding issues that are slowing down a project, or asking for a summary of a lengthy document that would take too long to read in full. Many people also take advantage of the chatbot to brainstorm creative ideas, organize their schedules more efficiently, or get straightforward explanations of topics they are trying to learn. These interactions, which make up the bulk of chatbot use, tend to focus on getting clear answers, reliable support, or useful information, rather than on building an emotional relationship or engaging in social conversation.

What this means for chatbots

The results of this study suggest that, for most people, chatbots like Claude are valued first and foremost as intelligent, flexible tools that can help solve practical problems, speed up routine tasks, and offer guidance or information quickly, rather than as substitutes for human friends or emotional connections. While it is true that a small group of users may experiment with companionship or conversational intimacy, the vast majority of interactions are utilitarian and goal-oriented, which means that the public perception shaped by sensational media stories does not match the reality of how these systems are actually being used every day.

Why this study may reflect overall chatbot use

The way people use Claude could give a general idea of how many people use chatbots in everyday life. Claude is one of the most popular assistants, and its users come from all kinds of backgrounds and ask for help with many different topics. When millions of conversations all show that most people use it for practical reasons, it suggests that many chatbot users, in general, are looking for information, help with tasks, or quick answers.

How it could differ for other chatbots

Even with that, things might be different on other platforms. Some chatbots, like Replika or Character.ai, were made for chatting, role-play, or even emotional support. They invite users to have long, personal conversations or to form a connection, and some people go there for exactly that reason. Other chatbots—such as ChatGPT, Gemini, or Copilot—tend to be used more for productivity, work, learning, and quick problem solving, just like Claude.

The tools, style, and even how the company advertises a chatbot can shape why people come and how they use it. If a chatbot offers games, emotional conversations, or characters to talk with, it will likely have more users who want companionship or entertainment. If it focuses on work and answers, that’s what users will expect.

What is known about other popular chatbots

Some surveys and reports have tried to find out how people use ChatGPT, Gemini, and others. The results often look similar to what Anthropic found: most people ask for summaries, help with writing, explanations, coding tips, or information. There are always some users who have longer, more casual conversations, but they are a small group.

How people use ChatGPT

Recent Pew data show adoption among U.S. adults has doubled in a year, with 26 % using it for learning, 22 % for entertainment, and 28 % of employed adults tapping it at work. Industry trackers estimate roughly 800 million weekly and 122 million daily active users worldwide, who lean on the tool for writing drafts, coding snippets, quick research, and language practice—tasks that help them “finish the job faster” rather than seek company.

How people use Google Gemini

A joint Google–Kantar study of early adopters found that 72 % reach for Gemini to boost productivity, 77 % to spark creativity, and 73 % to communicate more clearly. Users say the assistant “supercharges” routine work and idea-generation, with 92 % reporting a confidence lift after trying it. Traffic data back that utilitarian tilt: the service logged about 284 million visits in a single month, dominated by desktop users who drop in for summaries, code help, and cross-app actions inside Gmail, Docs, and Maps.

How people use Anthropic’s Claude

Anthropic sifted through 4.5 million Claude conversations and found only 2.9 % were “affective” (companionship, role-play, or counseling). Nearly all the rest centred on content creation, study aids, and troubleshooting. A separate education snapshot shows that 39 % of student chats involve generating or refining coursework, while 33 % tackle coding or math problems—further evidence that Claude’s core appeal is productive help, not emotional bonding.

How people use Meta AI

Folded into Facebook, Instagram, Messenger, and WhatsApp, Meta AI has raced past one billion monthly active users. Most encounters happen inside group chats where the assistant fields quick fact checks, drafts replies, and even spits out memes or image riffs—features tuned for social flow rather than heads-down work. Meta’s own messaging pitches the bot as a handy sidekick that can “show you things from the people and places you care about,” underscoring its lean toward casual, in-feed utility.

How people use character.ai

Character.ai encourages role-play with custom personas, and the numbers reflect that design: 41 % of users seek emotional support, while fantasy and sci-fi scenarios make up 51 % of all interactions. The platform claims about 28 million monthly active users and more than 18 million user-created characters, so its community skews toward storytelling, escapism, and companionship rather than productivity.

How people use Replika

Replika was built expressly for one-on-one companionship, and an independent roundup finds over 85 % of surveyed users say they’ve formed an emotional bond with their bot, chatting for reassurance, daily check-ins, or lighthearted conversation in more than 20 languages. Practical tasks exist, but they sit in the background; the draw is relationship, not output.

How people use Xai Grok

Grok, embedded in X (formerly Twitter), ballooned from 1.2 million to 25.8 million monthly visits after it Grok-3 upgrade, with a user base that is 67 % male and heavily 25-to-34. Its “real-time attitude” attracts people chasing fresh news, witty summaries, or edgy commentary; even U.S. lawmakers dip in for speech soundbites and rapid fact checks.

What we can and can’t say about Stella ai and Constella

Public, third-party data on Stella AI and Constella App remain scarce. Both market themselves to enterprises as knowledge-retrieval and analytic copilots, so early anecdotal reports suggest usage patterns closer to ChatGPT or Claude—drafting reports, querying company data—than to Replika-style bonding. Until independent studies emerge, any comparison is tentative.

______

The broader picture

Across mainstream “assistant” bots—ChatGPT, Gemini, Claude, Meta AI, Grok—practical help dominates: writing, coding, research, summarizing, scheduling. Platforms that invite fictional personas or aim at wellbeing—Character.ai and Replika—show a clear shift toward emotional or immersive chat. Design choices, marketing, and in-app prompts shape these outcomes more than the underlying model. So while Claude’s study hints at a wider norm, companion-centric services prove that a different interface can flip the balance toward feelings instead of files.

_______

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org