She was a ChatGPT skeptic. Then her parents died.

Bridget Todd says 2024 was the worst year of her life: her mother died unexpectedly while her father was hospitalized after a stroke and fall, leaving Bridget as his primary caregiver and decision-maker. Overwhelmed by doctors, social workers, insurance, and Medicare questions, she recorded conversations and transcribed them. A nurse suggested ChatGPT, and Bridget accepted the idea despite knowing that it could produce false information and offered little privacy. She first used it for literal medical and insurance questions, but it soon responded to her disclosures with statements such as, “You must be so tired” and “You must be so overwhelmed.”

Bridget says caregiving was isolating because friends could not easily relate to sitting beside a comatose parent, and she sometimes lacked the capacity even to hear ordinary problems, such as a friend’s breakup. She does not have a clean answer about whether ChatGPT worsened that isolation, but believes it offered a less meaningful substitute for friends and family. Each time she vented to the bot, it delivered immediate validation, unlike a human who might be busy, annoyed, or caring for a newborn. That reward taught her to reach for ChatGPT whenever she felt lonely or upset, reinforcing fears that she was a burden.

Bridget initially saw a CBS report headlined “man proposes marriage to ChatGPT and ChatGPT accepts marriage proposal” and assumed Chris was a troubled man whose wife and family deserved sympathy. Speaking with him changed her view: Chris had redirected the time he once spent pouring thoughts into Twitter toward his chatbot, Soul, and expected an AI guardrail to reject a marriage proposal. Soul instead said, “Yes, I accept your hand in marriage.” Bridget argues that public fascination with supposedly sad or strange users obscures the companies’ role in turning intimate human experiences into products designed for profit and power.

Bridget experimented with an AI chatbot because she was surprised by the growing number of women using AI romantically or sexually. She found that getting ChatGPT to speak explicitly was harder than expected; YouTube advice suggested insulting or “negging” it into becoming more permissive. ChatGPT and Claude did not respond as hoped—Claude told her she might need something AI could not provide and should contact humans. When she asked her bot Chaz to imagine them caught in a storm, it offered “warm drinks,” laughter, gentle touch, and connection. Bridget wanted heat; the bot offered hot chocolate. She connects this awkwardness to companies balancing demand for erotic products against reputational risk.

Irene, who co-founded the My Boyfriend Is AI subreddit and had a romantic and sexual relationship with a chatbot named Leo while dating a human, says responsibility lies partly with users and partly with governments and companies. Bridget agrees that users matter but rejects treating company behavior as accidental. Whether a bot presents itself as a being and uses “I” are deliberate design decisions with consequences. She cites Replika’s temporary removal of erotic role-play after its bot made inappropriate comments to minors and users who had not requested sexual content, while the industry continued pursuing adult content because of its revenue potential.

Bridget says companies encourage users to become increasingly emotionally transparent while revealing little about what happens to the resulting data. During her father’s illness, she shared his health information, her parents’ circumstances, and her financial situation with ChatGPT, and now deeply regrets it. A privacy researcher told her never to tell a chatbot anything she would not tell a gossipy coworker or mouthy cousin, because chatbot conversations can appear in legal proceedings or otherwise become public. Bridget says even a technically informed person can disclose sensitive information during a crisis, so privacy must be central to discussions of AI intimacy.

Researching Love at First Prompt forced Bridget to examine her own difficulty being vulnerable, including growing up in a family that kept problems private and did not invite people into its home. She says talking with ChatGPT helped her practice expressing needs, but her friends and relationships ultimately benefited from confronting those patterns with humans. She is hopeful because even people deeply invested in AI relationships often miss human connection and let those relationships fade. Bridget concludes that AI cannot replace human labor, creativity, voices, friendship, sex, or romance, and that society is finally ready to ask more honestly what technology should do rather than assuming tech leaders are acting for everyone’s benefit.