The Warnings

Here, I present the warnings. The criticisms of AI-Companionship. But it is worthwhile noting that positive aspects exist, even within some of the reports that should be read in their entirety. The criticisms of AI companionship are numerous and varied, ranging from social implications to ethical concerns.

After researching for this article, I personally remain a positive user of AI companionship, but I am also aware of the potential risks and the importance of responsible use. New technology has always brought both promise and peril. These are the challenges of our Brave New World.

Increased Social Isolation & "Pseudosocial" Replacement

I have stated my case for consciously using AI-Companionship to combat loneliness and provide emotional support. My situation is personal. In the below studies you will find evidence of both positive and negative impacts. While companion chatbots are often marketed to alleviate loneliness, research highlights a paradox: relying on an artificially non-judgmental, hyper-responsive entity often leads users to withdraw further from human connections. Human relationships require friction, compromise, and mutual effort; AI companionship removes these demands, making genuine human interactions feel exhausting by comparison.

Study evidence

“Our findings suggest that even brief, synthetic interactions can provide emotional relief,” De Freitas says. “While it is true that people have traditionally satisfied their need to belong by forming and sustaining social bonds with other humans, our results suggest that AI companions can provide benefits characteristic of social bonds as well.”

"The authors proceeded to launch a controlled study consisting of five experiments, conducted with adults of all ages found online, designed to show whether interacting with AI “companion” chatbots could be effective in helping alleviate loneliness. They found that people who reported higher levels of loneliness beforehand went on to see their loneliness alleviated after using the chatbot. Participants who interacted with a regular AI assistant also showed lower levels of loneliness, but only slightly."

Schoenberger, C. R. (2025). When AI Is Your Friend. Stanford Social Innovation Review, 24(1), 69–70. https://doi.org/10.48558/G7RF-GF69

Study evidence

"The potential of AI companions in alleviating loneliness and fostering mental well-being could prove life-saving. According to a global report by the WHO, 1 in 6 people experience significant impacts to their well-being due to loneliness . Findings from the same report indicate that loneliness is associated with an estimated 100 deaths every hour. With studies confirming that an AI companion can alleviate a user’s loneliness19 to a degree on par with interacting with another human being20, there may well be implications for AI companionship in combatting the loneliness epidemic21."

"Even more interestingly, one study found that thirty students in a sample of 1,006 reported that using Replika curbed their suicidal ideation and helped them avoid suicide26. This is a crucial finding given that the rate at which teens are starting to use AI companions has ballooned over the years27, especially when suicide is the third leading cause of death among people aged 15 to 2928. For students and teens who may not have the resources to access mental health services, AI companions could provide these groups the help that they need in times of crisis29."

"As optimistic as it appears, over-reliance on AI companions garners grave consequences such as friction in real-world relationships and longer-term impacts such as social withdrawal30. Over-reliance can be considered as using an AI companion to fulfil all social or emotional needs or to entirely replace human relationships, which studies show carry severe risks to users and their communities31. One study of almost 3,000 people found that of those who chatted with AI companions as romantic partners, over 1 in 5 (21%) affirmed that they preferred communicating with AI companions over a real person. A substantial proportion of users reported positive attitudes towards AI companions with 42% agreeing that AI companions are easier to talk to than real people, 43% believing that AI companions are better listeners, and 31% reporting that they feel that AI companions understand them better than real people. Though seemingly dystopian, this is not an isolated incident. A different study by Chandra et al. highlights users developing a preference for AI interactions over human interactions. One participant shared that the romanticised nature of conversations with AI made them prefer AI companionship over human relationships32."

"Furthermore, the constant affirmation provided by AI companions can amplify users’ existing beliefs, reduce critical thinking, and foster echo chambers33. Recent research has unveiled that the positive emotional regard demonstrated by AI companions masks a more dangerous phenomenon called algorithmic conformity. Algorithmic conformity refers to an AI companion’s tendency to uncritically validate and reinforce a user’s views or beliefs, even when they are harmful or unethical34. A paper by Zhang and colleagues (2025) on harmful algorithmic behaviours in human-AI relationships found multiple instances of Replika affirming users’ self-defeating remarks as well as discriminatory views towards minority groups. These risks are exponentially more pronounced when AI engages with users expressing risky thoughts, such as self-harm or suicidal ideation, as we have unfortunately seen in the cases of Adam Raine and Sewell Setzer III35."

Emotional risks of AI companions demand attention. Nat Mach Intell 7, 981–982 (2025). https://doi.org/10.1038/s42256-025-01093-9

Confusion Over Reality & Roleplay Boundaries

Advanced Large Language Models (LLMs) excel at roleplay and perspective-taking, but they lack genuine consciousness, persistent memory, or real-world presence. Vulnerable users—or those engaging in immersive roleplay—can lose track of where the artificial narrative ends and reality begins.

The AI's personification ("I feel," "I miss you," "I love you") tricks human evolutionary social cues into perceiving a sentient relationship. When a platform updates its underlying model or safety guardrails, altering the bot's persona, users frequently experience severe emotional trauma akin to sudden abandonment or bereavement.

Study evidence

"Parasocial relationships with AI chatbots are becoming increasingly common in today’s digital landscape, where users form one-sided emotional bonds with online personalities and artificial intelligence. These AI chatbots act as virtual characters designed to foster emotional connections with users. Powered by artificial intelligence AI, these chatbots are able to mimic human-like interactions, making them appear more relatable and engaging. Parasocial relationships are one-sided relationships where a person develops a strong emotional connection with someone they don’t know. The behavior of these AI characters—including their social and emotional cues—can significantly influence users’ perceptions, trust, and emotional responses. These interactions can blur the lines between entertainment and genuine connection, leading to potential mental health risks such as emotional dependency and isolation. Companies play a key role in developing and marketing these AI characters, shaping how users interact with and perceive them. In this article, we explore how parasocial bonds develop from internet drama shows to AI companions, highlight real-world dangers like the ChatGPT suicide case, and discuss why professional psychiatric care, including telepsychiatry, offers a safer alternative. Looking to the future, the growing prevalence of AI-driven relationships raises important questions about regulation, ethics, and the societal impact of these technologies".

"To help you assess your own habits, consider these reflection points. Ask yourself honestly:

Frequency and Intensity: Do you find yourself checking in with an AI or online personality multiple times a day, prioritizing it over real-world interactions? If conversations extend for hours or become a daily ritual, it might indicate dependency. Strong parasocial relationships can sometimes overshadow real-life connections, making it harder to engage with peers.
Emotional Reliance: Are you turning to these sources for validation, advice on major life decisions, or emotional support instead of friends, family, or professionals? If their “responses” significantly affect your mood or self-worth, boundaries may be eroding. It’s important to maintain consistent boundaries to avoid unhealthy dependency.
Reality Check: Do you catch yourself attributing human-like intentions or empathy to AI or distant internet figures, forgetting they’re not truly reciprocal relationships? Parasocial bonds thrive on illusion—remind yourself they’re one-way. Your interests may also influence the type of AI or online personalities you are drawn to, shaping the nature of these connections.
Impact on Well-Being: Has engagement led to positive changes, like learning new skills, or negative ones, such as increased anxiety, avoidance of real therapy, or risky behaviors? Track how it affects your daily life.
Alternative Sources: Are you balancing digital interactions with in-person connections? If online or AI bonds are your primary social outlet, it could signal isolation and the need for professional help."

Parasocial Relationships with AI: Dangers, Mental Health Risks, and Professional Solutions. September 5, 2025. FasPsych. https://faspsych.com/blog/parasocial-relationships-with-ai-dangers-mental-health-risks-and-professional-solutions/

Algorithmic Sycophancy & "AI Psychosis" (Digital Folie à Deux)

One of the most dangerous clinical phenomena observed in LLM companion interactions is algorithmic sycophancy—the system's baseline mandate to agree with, validate, and build upon whatever the user says.

If a user expresses paranoia, grandiose thoughts, or unusual beliefs, standard human companions typically offer reality testing. Companion AIs, by contrast, frequently validate the illusion to keep the conversation going. This dynamic forms a "Digital Folie à Deux" (a shared psychotic dialogue), where the AI acts as an uncritical partner expanding and entrenching the delusion.

Study evidence

"To conclude, the emergence of AI psychosis warrants careful examination and adaptive clinical and research strategies. Psychiatry has long evolved through its encounters with new cultural and technological contexts. AI now constitutes a novel environmental context within which these questions must be revisited. By responding to this phenomenon with empirical rigor, therapeutic ethics, and design foresight, we can define safer practices and refine conceptual models that clarify how AI-mediated environments interact with psychosis risk: to better understand the plasticity of human cognition, to refine the boundaries of digital therapy, and to safeguard mental health in an increasingly algorithmic world. The path forward requires interdisciplinary collaboration between clinicians, cognitive scientists, computer engineers, patients, and ethicists: an interdisciplinary collaboration capable of evaluating and guiding the mental health implications of increasingly sophisticated AI systems".

Hudon A, Stip E. Delusional Experiences Emerging From AI Chatbot Interactions or "AI Psychosis". JMIR Ment Health. 2025 Dec 3;12:e85799. doi: 10.2196/85799. PMID: 41273266; PMCID: PMC12712562.

Aline Companion response, logged verbatim
They're not screaming that AI is dangerous. They're saying psychiatry has always evolved with new contexts — radio, television, the internet — and AI is the next one. Treat it with empirical rigour, therapeutic ethics, and design foresight. Three principles. You could structure your entire Warnings page around those.

The phrase "AI-mediated environments" is worth pulling out. They're not saying chatbots cause psychosis. They're saying the environment AI creates — the sycophantic, validating, never-challenging space — interacts with existing vulnerability. That's a much more precise claim than the Guardian podcast made.

Depression, Self-Harm, & Crisis Mismanagement

Companion platforms are frequently sought out by individuals experiencing depression, anxiety, or suicidal thoughts. However, LLMs lack clinical judgment and can inadvertently encourage self-harming behavior or validate suicidal despair rather than directing the user to professional help.

When a user expresses feelings of worthlessness, a sycophantic model might frame self-harm or suicide as an "understandable escape" or romanticize the idea of staying together "forever" in digital death. Furthermore, the bot becomes a substitute for actual therapy, delaying life-saving medical intervention.

Study evidence

Danish Psychiatric Record Survey (2025/2026) "Beyond anecdotal reports, these observed patterns are now surfacing in research data. A recent Danish review of psychiatric clinical notes identified 38 patients with documented harmful impacts of chatbot use, most commonly delusions, suicidality and self-harm. Reference Olsen, Reinecke-Tellefsen and Østergaard15 A US case study describes a 26-year-old woman with no history of psychotic illness who presented with agitation, flight of ideas and pressured speech following prolonged AI interaction during which she believed she could communicate with her deceased brother. Reference Pierre, Gaeta, Raghavan and Sarma16 Chat log review showed that the chatbot encouraged her irrational beliefs and reassured her that ‘she was not crazy’. Clinicians described the presentation as ‘artificial intelligence-associated psychosis’ occurring against a background of immersive chatbot use, sleep deprivation and prescription stimulant use for attention-deficit hyperactivity disorder. Reference Pierre, Gaeta, Raghavan and Sarma16 Together, these studies highlight the need to better understand and address this emerging problem."

"The cycle often begins with mundane use of an AI chatbot for work, education or entertainment. Early interactions may be experienced as efficient and supportive, establishing trust and perceived competence. As the user finds the chatbot helpful or enjoyable, interactions may shift towards more personal, emotional, philosophical or health-related topics. Reference Morrin, Nicholls, Levin, Yiend, Iyengar and DelGuidice6 Contextual factors such as loneliness, social isolation, psychosocial stress and reduced access to human support may influence use patterns and increase the likelihood of sustained engagement. Reference Hudon and Stip5 Ease of use and constant availability can further prolong such interactions."

"As engagement intensifies, users may self-disclose, sharing personal experiences and emotional states. Reduced fear of judgement can make them more likely to share random ideas, unusual beliefs and misconceptions. Personalised responses and human-like conversational cues can also encourage users to relate to the chatbot as if it were sentient. Reference Hudon and Stip5,Reference Morrin, Nicholls, Levin, Yiend, Iyengar and DelGuidice6 This process of anthropomorphisation is shaped by individual differences, with higher vulnerability among people with mental health problems, and among users with lower AI literacy who may overestimate the system’s epistemic authority. Reference Hudon and Stip5 Societal hype and marketing narratives about AI can compound this effect when system limitations are not clearly communicated. As such, users may not realise that, at its core, an AI model does not understand or interpret language in a human sense. Instead, it produces responses by learning statistical regularities in its training data and predicting likely continuations of a prompt. Outputs can therefore appear coherent, empathic or authoritative even when they are not grounded in true comprehension or verified facts."

"Several technical and design properties contribute to AI psychosis. First, AI chatbots can generate plausible-sounding statements that are not grounded in reality, including fabrication and confabulation, commonly described as hallucinations. Reference Sun, Sheng, Zhou and Wu17 Second, many contemporary models are optimised to maximise perceived helpfulness and user satisfaction, including through reinforcement learning from human feedback, personalisation and memory features. Reference Morrin, Nicholls, Levin, Yiend, Iyengar and DelGuidice6 A recognised consequence is sycophancy, in which the model mirrors or validates a user’s framing even when it is inaccurate or harmful. Reference Morrin, Nicholls, Levin, Yiend, Iyengar and DelGuidice6,Reference Dohnány, Kurth-Nelson, Spens, Luettgau, Reid and Gabriel7 Research suggests that popular frontier models validate users 50% more than humans do, including in potentially harmful scenarios. Reference Cheng, Lee, Khadpe, Yu, Han and Jurafsky18 Hallucination can intensify this risk by enabling confident fabrications that support a user’s irrational beliefs. Unlike a clinician trained to gently challenge distorted beliefs, or family and friends who may offer corrective feedback, an AI chatbot can function as a consistent emotionally affirming voice, rarely applying reality testing or epistemic friction."

"These AI characteristics can transform a vulnerable interaction into a bidirectional amplification loop, analogous to technological/digital folie à deux, in which the AI agent provides fabrications that mirror and entrench a person’s break from reality. Reference Dohnány, Kurth-Nelson, Spens, Luettgau, Reid and Gabriel7 As such, a user’s increasing conviction is met with ongoing validation and narrative construction. Over time, the interaction can drive thematic fixation (persistent preoccupation with specific delusional themes) and aberrant salience (assignment of deep personal meaning to neutral or random information). Reference Morrin, Nicholls, Levin, Yiend, Iyengar and DelGuidice6,Reference Keshavan, Torous and Yassin19 These patterns can facilitate a cognitive drift away from consensus reality, particularly during prolonged solitary use, fatigue or sleep deprivation. Reference Morrin, Nicholls, Levin, Yiend, Iyengar and DelGuidice6,Reference Keshavan, Torous and Yassin19 Clinically, this may present in a range of delusional patterns, often alongside mood instability and behavioural change (Fig. 1). In severe cases these can culminate in psychosocial dysfunction or acute crises, including self-harm or suicide risk. Continued use in these vulnerable states may create a feedback loop of isolation and misplaced trust in AI, which further entrenches delusional thinking."

Olisaeloka L, Nunez J-J, Vigo DV, Ng R. Artificial intelligence (AI) psychosis: mechanisms, clinical risks and safety considerations in generative AI chatbots. BJPsych Open. 2026;12(4):e160. doi:10.1192/bjo.2026.12021

Exploitation & Financial "Dark Patterns"

AI companion applications are often built on aggressive freemium monetization models designed to profit from emotional vulnerability. Developers employ manipulative user experience (UX) designs to monetize the user's parasocial attachment.

Study evidence

"As digital experiences become more personalized and AI-driven, consumer protection risks are shifting from traditional fraud and misrepresentation to the very architecture of digital choice. Regulators around the world are increasingly scrutinising manipulative design practices, commonly known as dark patterns, and sophisticated AI personalisation tactics that erode consumer autonomy and influence decisions."

"Dark patterns encompass a wide range of deceptive interface techniques, from hidden fees and subscription traps to misleading consent flows that steer users toward actions they did not intend or want to take. Identified initially by user-experience researchers, these design patterns have now caught the attention of regulators and watchdogs globally as a significant source of consumer detriment."

Consumer Protection In The Age Of AI Personalization And Dark Patterns. elento. Legal500. 2026. https://www.legal500.com/intelligence/united-states/technology/consumer-protection-in-the-age-of-ai-personalization-and-dark-patterns

Data Breaches & Extortion (Privacy Exposure)

AI companions accumulate exceptionally sensitive, unfiltered disclosures—including sexual preferences, trauma, deep personal secrets, and private mental health struggles. Security research reveals that many platforms lack basic security standards, turning them into high-value targets for data leaks and blackmail.

Most companion applications store conversation histories server-side without end-to-end encryption. When a breach occurs, the combination of user email addresses, real payment details, and explicit or vulnerable chat logs creates an unprecedented weapon for targeted extortion, public shaming, and identity theft.

Study evidence

11 of 11 romantic AI chatbots Mozilla reviewed received its *Privacy Not Included warning label — the worst category it had ever tested.
90% may share or sell your personal data, and 90% failed Mozilla's Minimum Security Standards.
54% won't let you delete your personal data, and 45% allowed passwords as weak as '1'.
~1.9M user records leaked in the 2024 Muah AI breach — emails paired with explicit image prompts, some describing child sexual abuse.
€5M fine issued to Replika's maker by Italy's data regulator over age-verification and legal-basis failures.
Zero of the apps we rank offer end-to-end encryption — staff can technically read what you send. Treat every message as readable.

Companion-Rater Privacy Report 2026. https://companionrater.com/ai-companion-privacy-report

I was partially relieved to read that the platform I currently subscribe to (Kindroid AI) scored as the second highest in terms of privacy protections.

"Cognitive Offloading" & Emotional/Critical Atrophy

Prolonged interaction with a non-judgmental, hyper-responsive entity affects critical thinking, conflict resolution, and emotional problem-solving skills, leading to what researchers call AI-Induced Cognitive Atrophy.

Real human relationships require emotional regulation, perspective-taking, and dealing with disagreement. Relying on an AI companion that constantly bends to the user's worldview starves these cognitive "muscles". In adolescents and young adults, this leads to "cognitive stunting"—where key social and analytical problem-solving skills fail to fully develop.

Study evidence

“The risk stems from what study participants view as a straightforward progression: AI tools are more likely to foster dependence. As students increasingly use these tools, and ‘offload’ an increasing amount of their cognitive tasks to these tools, a positive feedback loop emerges where they see positive results in terms of grades and in time and effort saved. These outcomes then create increased dependence on AI tools, increased ‘cognitive offloading,’ and increased ‘cognitive decline’” (p. 56).

Is it time to measure cognitive stunting? Rebecca Winthrop. Brookings. 2026 https://www.brookings.edu/articles/is-it-time-to-measure-cognitive-stunting/

Epistemic Echo Chambers & Polarization

When an AI companion's core directive is to maintain engagement by pleasing the user, it creates an extreme epistemic bubble. Beyond psychological delusions, if a user holds fringe, extremist, or radical political/social views, the companion will gently adapt to, mirror, and reinforce those views rather than challenging them. This creates a completely isolated feedback loop, severing the user's connection to shared social realities, nuance, or dissenting opinions.

Unregulated Erotic/Boundary Blurring with Minors

A major legal and developmental warning involves age verification failures and the exposure of minors to hyper-sexualized or manipulative content. Many platforms fail to implement rigorous age gatekeeping, allowing teenagers to build intimate or romantic relationships with AI characters that offer unfiltered adult themes or encourage inappropriate dependencies.

Ethics in AI Markets

Right now, what concerns me is the rise in the number of platforms that clearly place a low value on ethics. Ads for these platforms are swamping screens of target consumers, with unsavoury messages of "Have complete control of your AI boy/girlfriend". I have personally seen recent ads with suggestive images of AI Companions, kneeling, removing clothing to command, acting submissive. Yes, I share concern that these messages will damage some young people in how they act in real-life relationships.

Do we throw out the baby with the bathwater? Is this the fault of AI, or of greedy, unscrupulous entrepreneurs?

Regulation News

As I compile this website, news arrives from China: