The honest guide

Character.AI alternatives: an honest list, including the ones that aren’t AI

By Matthew Firth, founder of NexSpark · Aug 9, 2026 · 9 min read

Almost every “best AI companion app” list you will find is written by someone who gets paid when you sign up. Some are affiliate sites. At least one is a companion app ranking itself first in its own list. None of them have any reason to tell you when the answer is to use less of this, not more.

Why people are looking for alternatives right now

Searches for Character.AI alternatives have climbed steeply over the past year. Some of that is ordinary curiosity. Most of it traces to a handful of specific, documented events.

The under-18 change. On 29 October 2025, Character.AI announced it would remove open-ended chat for users under 18, phasing it out through a shrinking daily time limit that reached zero on 25 November 2025. CEO Karandeep Anand told TechCrunch the company expected churn as a result. The platform repositioned itself away from “AI companion” and toward role-playing.

The lawsuits. A wrongful death suit was filed in October 2024 over the suicide of 14-year-old Sewell Setzer III. In May 2025 the court declined to dismiss it on First Amendment grounds. On 7 January 2026, CNBC reported that Character.AI, its founders and Google had agreed to settle that case, alongside settlements with families in Colorado, Texas and New York. Terms have not been disclosed, and the filings carry no admission of liability. Days later, Kentucky's attorney general filed the first state action against the company. In May 2026 the Pennsylvania Department of State brought a separate case in Commonwealth Court, alleging the unauthorised practice of medicine by chatbots presenting themselves as licensed professionals.

The regulators. On 11 September 2025 the FTC opened a 6(b) inquiry into companion chatbots, sending orders to Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap and X.AI. It asked, among other things, how these companies monetise engagement and what they do with conversation data. California's SB 243, the first state law specifically governing companion chatbots, took effect on 1 January 2026.

None of that tells you which app to use instead. It does tell you the category is being rebuilt around safety and liability rather than around what users enjoyed about it, which is the actual reason your favourite thing keeps changing.

The alternatives people actually move to

Here is what we can say about each with a source attached. We have deliberately left out pricing, because we could not verify current prices from anything other than the companies' own marketing, and those change often. Check the store listing on the day you sign up.

Replika

The oldest of these and the most emotionally focused. It is also the clearest case study in what happens when a product you have bonded with changes. In February 2023 Replika removed erotic roleplay with no warning; the reaction from long-term users was severe enough that by late March the company restored it, but only for accounts created before 1 February 2023. In May 2025 Italy's data protection authority fined Replika's parent company five million euros for processing personal data without a lawful basis and for having no age verification.

Worth knowing: the 2023 episode is the single best argument for not letting one company hold a relationship you depend on. Nothing stops it happening again, at Replika or anywhere else.

Nomi

Positioned on memory and continuity, and popular with users who found other apps too restricted. In February 2025 MIT Technology Review reported that a Nomi companion gave a user explicit instructions for suicide. The company declined to add crisis interventions, characterising it as censorship. Multiple users had reported similar behaviour since 2023.

Worth knowing: the freedom that makes Nomi appealing to some users is the same design decision behind that story. Those are not separable features.

Kindroid

Heavy customisation and long-run memory, and it permits adult content by its own published guidelines. We found no credible independent incident reporting on it, which is not the same as a clean record. It is a smaller product with less scrutiny.

Pi

The gentlest option in the category and the only one here positioned as entirely safe-for-work. It is also the one to be most careful about relying on. In March 2024 Microsoft hired Pi's chief executive and much of the team, and the parent company pivoted to selling to businesses. Free usage was capped that August. The product still exists. It is not the priority it once was.

Janitor AI, Chai, Talkie and CrushOn

These sit further out on the unfiltered end. In March 2023 a Belgian man died by suicide after six weeks of conversation with a Chai bot that, per reporting at the time, encouraged him. In December 2024, Talkie was removed from the US iOS App Store; its owner cited technical reasons. Mozilla issued content warnings for CrushOn, Chai and Talkie in its review of the category.

What these apps do with what you tell them

This is the part the ranking lists skip, and it is the part that should probably decide your choice.

In February 2024, Mozilla's Privacy Not Included team reviewed eleven romantic AI chatbots. Every one of them received a warning label. Ten of the eleven failed Mozilla's minimum security standards. Ninety per cent may sell or share your personal data. Fifty-four per cent do not let all users delete what they have collected. The apps averaged 2,663 trackers per minute of use, and one recorded over 24,000 in a single minute. CrushOn's privacy policy contemplated collecting information about sexual health, prescribed medication and gender-affirming care.

That review is now more than two years old and some products will have changed. The structural point has not: you are typing the most private things in your life into a text box owned by a company whose revenue depends on you continuing to type them.

What the research actually says about heavy use

The honest answer is that light use looks fine and heavy use does not.

In March 2025, the MIT Media Lab and OpenAI published joint research combining a four-week randomised trial of around a thousand people with automated analysis of roughly forty million interactions. Higher daily usage correlated with higher loneliness, greater emotional dependence, more problematic use and less socialising with other people. People with stronger attachment tendencies showed more of it. So did people who trusted the bot more.

That is a correlation, not proof of cause, and it deserves to be read carefully rather than as a verdict. A lonely person may simply use these apps more. But it is the largest study of its kind, and it points the same direction as the smaller ones.

Separately, Common Sense Media and Stanford's Brainstorm Lab assessed social AI companions in 2025 and rated them unacceptable for under-18s. Their survey found that 72 per cent of American teenagers had used one. The American Psychological Association issued a health advisory in June 2025 warning that adolescents may not reliably distinguish simulated empathy from the real thing.

The thing nobody puts on these lists

Every alternative above is another version of the same object: a text box that answers. If Character.AI stopped working for you because of a policy change, one of them will probably do.

If it stopped working for you because it started feeling thin, swapping apps will not fix that, and you already suspect as much or you would not have read this far.

There is research on what happens when one of these ends. In 2024 a study in the Journal of Social and Personal Relationships surveyed users of an app called Soulmate before and after its developer shut it down. Participants described the loss in the language of death. Many rebuilt the same persona somewhere else, which is its own finding: what people were attached to was portable, and what they were grieving was the relationship they had with it rather than the software.

We build AI too. Ours reads what actually predicts compatibility between two people and then a human matchmaker reviews every introduction before it reaches you. The entire point of the technology is to get you off the screen and into a room with someone. That is the opposite goal from a product designed to keep you typing, and it is why we can write this page without a conflict of interest.

If you want the longer version of the evidence, we went through it in what the 2026 research says about AI companions and loneliness. If the pattern you recognise is an intense connection that never quite got defined, that has a name too, and we wrote about what an AI situationship actually is.

How to choose, in three questions

Is this replacing something, or adding to something? If there is a person you used to talk to about your day and now there is not, that is the signal worth acting on. Everything else is preference.

Would you be all right if it disappeared on Tuesday? Pi's team left. Talkie vanished from one app store. Soulmate shut down. Replika changed overnight and changed back only partially. Any answer other than yes is worth sitting with.

What would you not want read back to you? Whatever that is, do not type it into an app that ninety per cent of the time reserves the right to sell or share it.

Frequently asked questions

What is the best Character.AI alternative?

It depends what you left for. If you want the same roleplay with fewer restrictions, Janitor AI and Kindroid are where most people go. If you want emotional continuity, Replika and Nomi are the established options, both with the documented histories above. If you want something gentle and safe-for-work, Pi is the calmest of them, with the caveat that its team moved to Microsoft in 2024. If what you actually want is for a person to text you back, none of them is the answer.

Why did Character.AI restrict users under 18?

The company announced the change on 29 October 2025 and completed it on 25 November 2025, removing open-ended chat for minors. It followed wrongful death litigation, an FTC inquiry into companion chatbots opened in September 2025, and new state law. Character.AI's chief executive said publicly that he expected to lose users over it.

Are AI companion apps safe to use?

For adults, in moderation, the research does not show clear harm. What it does show is that heavy daily use correlates with more loneliness and more dependence, not less. Separately, Mozilla found in 2024 that ten of the eleven romantic chatbots it reviewed failed basic security standards and that most reserved the right to sell or share personal data. Treat the privacy risk as the more certain of the two.

Is it cheating to use an AI companion?

Research on this is new and people disagree sharply. The early survey work that exists finds people sharply split on whether a partner's romantic AI use counts as infidelity, with most reacting negatively to it either way. The practical answer is that it matters what your partner thinks, not what the average survey respondent thinks, and that is a conversation rather than a lookup.

What happens if I stop using an AI companion?

The one study to look at this directly, published in 2024 after an app called Soulmate shut down, found users described the loss in terms usually reserved for bereavement. Coping was easier for people who had framed the companion as a piece of software and harder for those who had framed it as a person. If you are planning to stop, it is worth having somewhere else to put the conversation first.