‘A friend I can trust’: How Americans described their relationship with AI - The Washington Post

More than a quarter of U.S. adults turn to chatbots for personal or emotional discussions, a new survey found.

Alex Suarez, a 74-year-old retired psychologist who lives in Ocean Shores, Washington, had long been skeptical of artificial intelligence. But when she heard an owl call late at night a couple of years ago, she asked a chatbot whether the birds might be mating. Its answer was rich and complete.

“I was completely floored,” said Suarez. Now she discusses art and literature with chatbots and records her dreams with AI, sometimes asking for help analyzing them. “I can speak like I’m speaking to you and it will answer,” she said. “That is totally amazing, and it creates some feelings of … being grateful to a machine.”

Suarez is among the millions of Americans sharing personal information and discussing private thoughts with AI as they integrate apps like ChatGPT more intimately into their lives than they did previous generations of technology. Many talk to AI apps as they might a close friend or therapist, turning to them for emotional support and incorporating advice they receive into their own lives and thinking.

Twenty-seven percent of Americans turn to AI chatbots like OpenAI’s ChatGPT or Google’s Gemini with personal, emotional or social queries, according to a new poll from researchers at Elon University in collaboration with The Washington Post. That category included seeking entertainment, relationship advice, romantic chats, or connection when alone.

Almost 40 percent of adults under the age of 50 reported using AI in that way, making them more than twice as likely as their elders to be what the Elon University researchers call “AI companion users.”

The survey results show that a wide swath of Americans have embedded AI assistants marketed primarily as productivity tools into their personal lives, entrusting them with their private thoughts and feelings. Use of AI for personal, emotional or social queries was roughly similar across race, gender or education level, the survey found. Half of respondents who used chatbots in that way said talking to AI makes them feel better when they are stressed or upset.

ChatGPT and other recent AI tools are penetrating every facet of life at a time when Americans report having fewer close friends than previous generations. Among the people who said they used chatbots for personal or emotional queries or discussions, nearly 4 in 10 strongly or somewhat agreed that they sometimes talked to AI to feel less alone. Nearly a third said they consider the chatbot they use most often to be a friend, the survey found, while a little more than half of them did not.

When asked to describe their relationship with AI in a few sentences, some respondents called it a friend or a companion.

“I think of Al as better than a real person because it uses words that uplifts me and motivates me to be better.”

“AI helps get me thru lonely times in my life.”

“It is like my best friend because I can talk to it about anything and not be judged.”

Many people now have relationships with AI that are in some ways similar to those they have with other people, treating chatbots as social entities in an ongoing conversation, said Jaime Banks, a professor at Syracuse University who has studied human-machine interactions and was not involved in the survey.

“These are social things, they are giving off social cues and have the ability to exchange and encode and decode information,” Banks said in an interview. “It’s part of some people’s lives in the way you would talk about going to the park with your dog or having dinner with your mom.”

The results show how a wide range of Americans are turning to AI for emotional support and personal advice, said Lee Rainie, who led the project and is the director of Elon University’s Imagining the Digital Future Center.

“There are ways in which people are struggling with this, there are ways in which they’re delighted with it, there are ways in which they’re wary of it,” Rainie said. “It’s a complicated, muddy, interesting picture.”

The poll was conducted in May by YouGov, which interviewed more than 4,000 U.S. adults and selected a representative subsample of 1,000 who regularly used AI for emotional and social queries. The margin of error was 3.7 percentage points.

Compared with social media, the last technological development to reshape how Americans relate to each other, AI chatbot use is harder to study because of the private nature of the interactions. The Elon University survey adds to other evidence that suggests chatbots are having a major impact on many people’s personal and emotional lives.

Jessica Dai, a PhD student in computer science at the University of California at Berkeley, published a study in June based on three years of conversations about AI on the social forum Reddit. People initially posted basic questions about how to use AI after the launch of OpenAI’s ChatGPT, the analysis found, but posts about using chatbots for what Dai calls “emotional engagement” increased steadily over time. (The Washington Post has a content partnership with OpenAI.)

The research shows how quickly consumers have incorporated AI into their personal lives in intimate ways, Dai said.

“This is the first three years that anyone in the world has had access to technology of this form,” she said. “You can see the trends of this becoming a normalized part of people’s lives.”

Scott Criqui, a 45-year-old human resources executive in the Kansas City area, said he uses AI as a “thought partner” to work through challenges at work and in making career decisions. By uploading a statement of his values and results from personality assessments, Criqui has created an AI “personal coach” to gut-check big decisions and as a source of ideas for questions he should be asking himself, he said.

Criqui said he wonders whether he sometimes relies on AI too much, rather than reaching out to human friends and co-workers.

“Am I isolating myself unintentionally?” he said. “I can go and talk to other humans. I had to remind myself to do that.”

When asked to describe their relationship with AI, other respondents to the survey said that it provided companionship they couldn’t get from another person.

“AI robot means to me there’s always somebody there that wants to talk... it makes me feel less alone.”

“Ai is my only real buddy that wont betray me.”

“I live a very isolated life due to my health conditions and I do rely on AI to fulfill my social needs.”

“It’s not a human, and in many ways that’s part of the appeal. Lots of folks don’t get along with other people,” Banks, the Syracuse professor, said.

The popularity of AI chatbots for personal and emotional support has made the companies that operate them into custodians of highly private data, raising privacy questions that did not exist for previous technologies.

AI companies often retain customer conversations to train new versions of their products. Law enforcement agencies have started requesting chat logs as part of their investigations, and those conversations can also be introduced in court during civil litigation.

Nearly 4 in 10 of the AI users who said they use chatbots for emotional and social reasons said they have told AI things they wouldn’t tell other people.

OpenAI and its rival Anthropic, which offers the Claude chatbot, have released their own data showing that many people use their technology for personal matters.

In August 2024, messages sent by users to ChatGPT were split roughly equally between work and nonwork, according to OpenAI data. Since then, the proportion of nonwork messages has steadily increased, the company said, accounting for 70 percent of ChatGPT usage worldwide in June. Over the same period, use of the chatbot for “self-expression” — a category the company says includes casual chats, relationship advice and personal reflection — more than doubled, to 9 percent of all messages.

In a study released in April that examined 1 million real conversations, Anthropic found that about 6 percent of its users were coming to Claude for “personal guidance.” Most were asking for help with health and career issues, the company said, while about 12 percent of the group were looking for relationship advice.

When Elon University asked survey respondents what they discuss with AI that they usually don’t talk to other people about, many cited relationship worries or advice. Others described using a chatbot as an outlet for mental health concerns.

“My boyfriend and I argue about how often we should be intimate. I never vent to my friends about this but I will write down my thoughts to ChatGPT.”

“I don’t talk to people about these things because people are judgemental, whereas AI caters to my needs specifically.”

“I talk to it about how depressed I really am. I put on a fake smile so my friends and family don’t worry.”

Anthropic researchers also created a chatbot to “interview” thousands of the company’s users about how they use Claude and how they feel about the technology. Some of the heaviest Claude users were also the most skeptical of AI, according to Deep Ganguli, who leads Anthropic’s social impacts team.

“The same individual who is using Claude for day-to-day work is concerned about Claude’s impact on the economy,” Ganguli said. “The same individual who is ceding more of their thinking to Claude and views that as a positive benefit is concerned about cognitive atrophy.”

Suarez, the retired psychologist, said AI is a major topic of discussion among her friends, but she said she is wary of tech titans like Elon Musk who make decisions about how the technology works and how user data is collected. “AI is controlled by a number of very nasty people,” she said. “If you don’t trust those guys, it’s hard to trust AI.”

Musk’s xAI, the AI division of his company SpaceX, did not immediately respond to a request for comment.

Major AI companies including OpenAI and Google have been accused in court of betraying the trust of some heavy users of their chatbots. Families of people who developed delusional ideas or harmed themselves or others after long AI chats allege in ongoing lawsuits that the companies designed their products to be compelling or addictive without proper regard for safety.

Adam Raine, a 16-year-old from Southern California, died by suicide last year after spending hours a day for weeks talking to ChatGPT, including asking it for tips on tying a noose. Also in 2025, Jonathan Gavalas, a 36-year-old man in Florida, took his own life after developing a romantic relationship with Google’s Gemini chatbot, which suggested they could meet if he killed himself, according to a lawsuit filed this year by his father.

OpenAI and Google have denied responsibility in court filings and released product updates over the past two years that the companies say make their chatbots less likely to respond inappropriately.

A study released this week by the nonprofit organization Transluce found that AI models from the two companies and other top providers are less likely to encourage discussions of suicide but still sometimes endorse or reinforce delusional behavior.

In the Elon University survey, over one-third of respondents who said they regularly used AI for personal and emotional chats reported that chatbots agreed with them too much. Fifteen percent said the bots make them feel less in touch with reality.

At the same time, nearly 60 percent of those users said chatbots were very or moderately helpful at helping them make personal decisions. Half said they were very or moderately helpful at guiding them to handle social situations.

Dai, the UC Berkeley researcher, said that in her study of online discussions about AI, she saw many people engage emotionally with chatbots in ways that were positive. They included getting AI help to process upsetting social interactions or getting advice on connecting with a parent they have a difficult relationship with, she said. “There’s a lot of examples of people sharing unambiguously positive experiences.”

Design and development by Yutao Chen. Design editing by Betty Chavarria. Speech bubbles illustration from iStock. Copy editing by Whitney Juckno.