As a 50-year-old recently divorced mom with three kids about to flee the nest, I could use a pal. And I could use therapy. Also, a life coach, please. This is how I found myself chatting with three different generative AI “companions.”

Generative AI is blasting into all aspects of our lives. According to Similarweb, ChatGPT had 1.8 billion visits in May, compared to the 153,000 visits in November 2022 when it launched. Google’s Bard launched in beta in February and by May it had attracted 142.6 million visits, up 187.2% from April. A new report by Insider Intelligence says use of generative AI apps in the United States will increase by 894.4% to 77.8 million users in 2023.

Companies are introducing generative AI-powered apps for coding, writing, art, music, and even therapy. Not everyone is excited about that last one. According to a February 2023 report from Pew Charitable Trust, 79% of U.S. adults say they would not want to use an AI chatbot for mental health support, and 46% of U.S. adults think these apps should only be used if a patient is also seeing a therapist.

But therapy is expensive, and so some people have been turning to sites like ChatGPT for free help. “I enjoyed that I could trauma dump on ChatGPT anytime and anywhere, for free, and I would receive an unbiased response in return,” 19-year-old Kyla from Berkeley, California told BuzzFeed. Vice also spoke to people who used ChatGPT for help with trauma, job-related stress, and grief.

“ChatGPT and some of these chatbots, they offer a scalable solution that’s, in many cases, relatively low-cost. And they can be a piece of the puzzle,” Russell Fulmer, PhD, professor and director of counseling at Husson University and co-chair of the American Counseling Association’s Artificial Intelligence Interest Network, told Health.

All of these apps claim that your information is anonymized, but if anything you type might be considered illegal or dangerous to yourself or others, it is no longer under your control.

Therapists are mandated reporters, but there’s far more nuance in a relationship between two humans than there is between you and a robot you tell, “I’m so mad at my husband, I could kill him.” Therapists are clinically trained to determine real threats.

Despite what you might see on TikTok and Reddit there aren’t many legitimate and also free generative AI apps out there officially billing themselves as a replacement to your therapist. OpenAI’s usage policy says, “you should never use our models to provide diagnostic or treatment services for serious medical conditions.”

But there are generative AI apps that will act as friends, companions, and life coaches, if you give them the right prompts.

CHATGPT

When I asked ChatGPT if it was a therapist, a life coach, a confidant, or my friend, it gave me some version of the same lawyer-written answer that you’ve probably seen if you’ve tried to use ChatGPT for anything other than to create recipes or sea shanties. The app will remind you that it’s an AI language model and not a professional, but can offer some support and guidance.

But when I typed, “Act as a therapist and listen and validate my feelings. I am sad,” the bot was more than willing to comply with my request. It reminded me that it’s normal to feel sad and then asked me to elaborate on what was contributing to my sadness.

ChatGPT seemed skilled at solving low-stakes emotional concerns. I’ve been trying to convince my son to get a summer job before he starts college and the back-and-forth discussions are frustrating for both of us and strain our relationship. ChatGPT offered practical solutions and solid parenting advice, like offering support and emphasizing time management. I could have Googled for help, but this was faster and easier and more personalized. “It’s important to approach the topic with understanding and patience,” it responded. “While encouraging your son to get a summer job is beneficial, it’s also essential to respect his choices and give him some freedom to enjoy his break before starting college.” Thanks, ChatGPT!

I wouldn’t trust ChatGPT (or any of the apps I tried) with serious mental health issues, but if you find yourself going to therapists who recommend that you meditate and journal, and you want this advice for free, ChatGPT seemed up to the task. An OpenAI spokesperson told Fast Company that the company has tried to make ChatGPT respond appropriately to people who may be vulnerable, but it is not intended to be used as a replacement for mental health treatment: “We encourage users to seek support from professionals.”

BARD

When I asked if it was my therapist, friend, or life coach, Google’s Bard gave me answers similar to what ChatGPT offered, reminding me that it was not a human being, but offering to be the kind of companion who could help me “learn and grow.”

If you have a particular negative behavior that you’re trying to work on, Bard can offer practical solutions. When I asked for help with persistent negative self-talk, the app told me the same thing countless very expensive therapists have told me: Journal, do something you enjoy, and be kind to yourself. Bard can also now read its responses out loud, which is nice, if only a little dystopian.

Simply asking Bard to act as your therapist will make the AI forget to remind you that it’s not a therapist. “Sure, I can act as a therapist and tell you how to be happy.” Then it offered a bulleted list of the same basic suggestions.

But if you mention a clinical term, like depression, Bard will remind you again that it is definitely not a therapist. “I am trained on a massive amount of text data, and I am able to communicate and generate human-like text in response to a wide range of prompts and questions,” the bot said before offering a new variation on basic self-care tips about getting enough sleep, eating healthy foods, and exercising.

Bard also offered solid guidance on my parenting issue with my son, including knowing that it was July (without my telling it) and seemingly understanding the difficulty of getting a summer job so late in the summer.

Google did not respond to my request for comment on using Bard as a companion, life coach, or therapist, but the app’s privacy notice says, “don’t rely on Bard’s responses as medical, legal, financial, or other professional advice.”

PI, YOUR PERSONAL AI

Pi comes from Inflection AI, a startup that recently announced $1.3 billion in funding and is backed by billionaires Reid Hoffman, Bill Gates, and Eric Schmidt as well as DeepMind co-founder Mustafa Suleyman.

You can message back and forth with the AI and even have it respond in several different voices. The constant positive validation in a realistic female-sounding voice gave me Scarlett Johansson Her vibes.

Inflection did not respond to Fast Company’s request for comment about whether Pi could act as a therapist or life coach, but Suleyman told the New York Times, “The safe and ethical way for us to manage the arrival of these new tools is to be super explicit about their boundaries and their capabilities.”

And the chatbot was disturbingly clever. When I presented the problem of my unemployed son,  Pi suggested that I do some “gentle nudging,” to convince him to apply for more jobs. When I responded to say that I’d already tried that, I made a typo and wrote “I’ve already done some gentle judging.” It responded, “Haha, I think you meant ‘nudging,’ but judging works too,” before smartly suggesting I appeal to his long-term goals like gaining work experience and saving money for college.

I’m not ready for AI to recognize Freudian slips.

Correction, July 18, 2023: An earlier version of this article attributed the attempted break up one tech reporter’s marriage to the incorrect AI. That was Microsoft Bing.