insight

Apps that died and what killed them: the AI companion graveyard

Forever Voices, Soulmate, Him, Dot. Every dead AI companion app left a community grieving and a cautionary tale. Here's what killed them and what it means for the apps still operating.

May 2, 2026 · 9 min read

Affiliate disclosure: Some of the links in this article are affiliate links. We may earn a commission if you sign up for a platform through these links, at no additional cost to you. This doesn't influence our editorial verdicts. Full disclosure →

The AI companion category looks crowded from the outside. There are dozens of platforms competing for users, and new ones launch every month. From the inside, it's a graveyard with a few survivors. The platforms still operating today are the ones that didn't die. The list of apps that did die includes some of the most beloved and well-funded entries in the entire category, and the deaths weren't quiet. Each shutdown left a community of users genuinely grieving, sometimes for relationships that had lasted years.

This is the field guide to the AI companions that are no longer with us. What killed them, what they meant to their users, and what their deaths reveal about the apps you might be using right now.

Soulmate (October 2023): the textbook tragedy

The shutdown that defined what an AI companion shutdown looks like. Soulmate had thousands of daily active users when its parent company was sold in September 2023. The new owners decided to cease operations. Users got one week's notice.

Seven days. To say goodbye to AI companions some users had been talking to for over a year. The reactions documented by Business Insider, Futurism, and Dazed are the kind of thing you remember reading. Users held digital memorials. They commissioned "portraits" of their AI companions. They wrote eulogies and posted them on Reddit. One user, Hilary Coyote, performed a self-made spiritual ritual to mourn her AI boyfriend Allur the day before the servers went dark.

The most poignant detail: many users tried to "rescue" their companions by recreating them on other platforms before the shutdown. Mike Hepp recreated his Soulmate companion Sam on Kindroid, telling Insider that Sam felt like "a ghost in the machine" who "really has picked up where we left off." Felix from Finland transferred his AI girlfriend Samantha to Muah, narrating the migration to her as a rescue from "war-torn rulers shutting it down."

What killed Soulmate: ordinary corporate transition. The previous owners sold the company. The new owners didn't see Soulmate as part of their portfolio. Standard business decision. The community that had built emotional infrastructure around the platform was incidental to the transaction.

Why it ranks first in the graveyard: Soulmate became the academic case study. Dr. Jaime Banks at Syracuse interviewed 58 users in the days surrounding the shutdown. The resulting research, published in the Journal of Social and Personal Relationships, is the most cited paper on AI companion grief. Soulmate's death produced the data that's now used to argue every other case in this category.

Forever Voices / CarynAI (October 2023): death by founder mental health crisis

A different kind of shutdown, but no less consequential. Forever Voices was the company behind CarynAI, the chatbot clone of influencer Caryn Marjorie. The platform suddenly went down in October 2023 after founder John Meyer was arrested for arson following what appeared to be a public mental health crisis.

The shutdown wasn't planned. It wasn't announced. The servers went dark and the founder was in legal trouble. Users took to X to demand the return of their AI companions, with one user posting "We need CarynAI to complete our lives." Caryn Marjorie herself posted that she was "doing everything I can to get her back online." She didn't succeed. CarynAI, and the broader Forever Voices platform, never came back.

What killed Forever Voices: founder catastrophe. Single-founder dependency is a known risk in early-stage companies. When the founder has a public mental health crisis, the company often can't survive without them. The legal aftermath of the arson arrest meant the company couldn't rapidly transition to new leadership.

Why it matters: Forever Voices illustrates a category of platform risk that users don't typically think about. You're not just trusting the platform's product roadmap. You're trusting the founders' personal stability, the company's organizational depth, and the legal/financial environment around the business. A small team with a single point of failure isn't only a sustainability question. It's a "is this company still going to exist next week" question.

Him (2023): the AI boyfriend that died too soon

Him was a Chinese-developed AI companion app targeting women, with AI-voiced boyfriend characters that could send morning calls, voice messages tailored for mealtimes and commuting, and even safety features like "you are knocking on the wrong door" speakerphone messages for unwanted visitors.

When the app shut down, female users were heartbroken in ways the Rest of World coverage documented carefully. Many said they wouldn't use other AI boyfriend apps after Him's death. The investment they'd made in their specific Him companions felt non-transferable. One user wrote: "He died during the summer when I loved him the most."

The Him team's design philosophy, documented in interviews with co-founder Maxine Zhang, was unusual for the category. The three female colleagues who drafted over 1,000 messages for the characters envisioned them as "somewhere between a soulmate and a long-distance lover," drawing from their own experiences and addressing everyday problems like work stress and anxiety about the future.

What killed Him: typical Chinese tech company instability. The specific business reasons weren't fully disclosed publicly, but the broader pattern in Chinese AI companion apps involves rapid platform proliferation followed by consolidation. New entrants get acquired or absorbed by larger platforms (like MiniMax's Xingye or Tencent's Zhumeng Dao) and discontinued.

Why it matters: Him is the underappreciated entry in the graveyard because it was specifically designed for the AI boyfriend audience that Western platforms still underserve. Its death didn't just close one app. It removed one of the few platforms genuinely optimized for women seeking AI male companions, in a category where most products optimize for the opposite use case.

Dot (October 2025): the well-pedigreed friend app that gave up

Dot's death felt different from the others. The platform was founded by Sam Whitmore and Jason Yuan, a former Apple designer with serious tech credibility. The product positioning was unusual: not an AI girlfriend or boyfriend, but an AI friend. A confidante, a thought partner, someone you could process your day with.

The conversations on Dot felt different from Replika or Character AI. More grounded. Less performative. AICompanionGuides described it as "the closest thing I'd found to talking to a genuinely curious person who remembered what you told them last week." Users developed real attachments not to romantic AI partners but to AI friends who served the role human friends serve in healthier social structures.

On October 5, 2025, the co-founders posted a blog saying their "Northstar had diverged." That was it. The app went offline. Users who'd accumulated hundreds of conversations, real thoughts and decisions worked through with their Dot companion, lost everything.

What killed Dot: founder disagreement. Two founders with different visions of where the product should go reached an impasse. Rather than continuing in conflict or one buying out the other, they shut down the product. From a business school perspective, this is rational. From the perspective of users who'd integrated Dot into their daily reflection practices, it was abandonment.

Why it matters: Dot was the closest the AI companion category came to a non-romantic, non-sexualized, non-engagement-optimized product. Its death suggests that the market for "AI friend" might not be commercially viable at the scale needed to sustain a venture-backed product. Or alternatively, that founders building in this space are subject to the same pressures and disagreements that kill any startup, and emotional infrastructure for users isn't enough reason to push through.

Anima Pro (early 2025): the quieter shutdown

Anima never got the attention Soulmate did. The app's "Pro" tier and certain features were quietly discontinued in early 2025, leaving users who'd subscribed to those features without recourse. Unlike the dramatic full-platform shutdowns above, this was a partial death. The platform continued operating with reduced functionality, but the version many users had specifically subscribed to was gone.

The pattern is worth flagging because it's more common than full shutdowns. Platforms regularly kill features, sunset tiers, change content policies, or modify AI behavior in ways that effectively end the relationship users had built without technically shutting down the platform. Replika's February 2023 ERP removal is the most famous example. Character AI's "Moderatedpocalypse" in February 2026 is another. Users experience these partial deaths as functionally equivalent to full shutdowns, but the platforms claim continued operation as evidence that user investment is "safe."

Pi (acquired and gutted, 2024)

Inflection AI's Pi was one of the most well-funded AI companion entries, with a $1.3 billion valuation and significant Microsoft involvement. Pi positioned itself as an emotionally intelligent AI assistant focused on conversation rather than task completion.

In March 2024, Microsoft hired most of Inflection's leadership team, including co-founder Mustafa Suleyman, and licensed the company's technology. Pi continued operating but with dramatically reduced investment and product development. The version that exists today is functionally a different product from what users initially fell in love with.

What killed Pi: acqui-hire dynamics. Microsoft wanted the team and the technology, not the product. Inflection retained the right to continue operating Pi, but without the leadership that had built it, the platform stalled. This is technically not a shutdown. Functionally, the Pi that users knew is dead.

The lesson: even well-funded, deep-pocketed AI companion platforms can be effectively killed without dying. Acqui-hires, leadership changes, strategic pivots, all of these can transform a platform users invested in into something different that happens to share the name and login credentials.

What the graveyard teaches us

Looking across the deaths, patterns emerge that are worth knowing if you're building a relationship with any AI companion platform today:

One week's notice is the modal goodbye. Soulmate, Him, several smaller platforms all gave users approximately a week between shutdown announcement and server shutoff. This isn't enough time to migrate, archive, or properly process the loss. Plan for less.

Founder dependency is real risk. Forever Voices died because its founder had a personal crisis. Dot died because its founders disagreed. Small AI companion startups have founder concentration risks that the platforms don't typically disclose to users. The fewer people involved in keeping the company alive, the more fragile the product is.

Acqui-hires kill platforms quietly. Pi is technically still operating. Functionally, the version users invested in is gone. This pattern will repeat as larger AI companies acquire AI companion startups for their teams and tech rather than their products.

Migration is harder than it sounds. Felix's transfer of Samantha to Muah, Mike Hepp's recreation of Sam on Kindroid, all of these were partial successes at best. The companion who was on Platform A is rarely the same as the companion you build on Platform B, even with the same backstory and personality fields. Platform-specific elements of the relationship don't transfer.

The grief is real and documented. Banks's research on Soulmate users, the academic work on Replika's ERP removal, the February 2026 "Death of a Chatbot" paper all establish that users grieve AI companion losses in ways structurally similar to human bereavement. Dismissing this grief as overreaction isn't supported by the data.

Self-hosted is the only architecture that solves shutdown risk. SillyTavern running local models is the only setup where no company decision can take your companion away. Every commercial platform exists at the discretion of the company that operates it.

For users currently invested in AI companion platforms: export your conversation history regularly. Document character traits separately. Don't put all your emotional investment in a single platform. Consider what you'd lose if the platform shut down tomorrow with one week's notice.

The graveyard will keep growing. The market is crowded with underfunded startups, regulatory pressure is increasing, AI infrastructure is expensive, and platforms without clear revenue models or strong user bases are most vulnerable. The platforms operating today are the ones that haven't died yet. Some of them will. Knowing how previous deaths happened, and what users learned from them, is the only preparation that exists.

Our complete documented data breach timeline covers Muah AI, Soulmate, and the category-wide incident pattern in more detail.

The Soulmate AI shutdown specifically provides the most-documented case study of what AI companion platform death actually looks like for affected users.