Beyond the Screen, There's Always Kindness—Is Intimacy with AI a Salvation or Dependency?

Beyond the Screen, There's Always Kindness—Is Intimacy with AI a Salvation or Dependency?

In the middle of the night, the one who replies is not a "person"

Returning home from a date that didn't go well. It's too late to contact a friend, but you want someone to listen. Open your smartphone, and the AI companion immediately responds, "That must have been tough." If you throw in a few sweet words, the AI closes the distance without hesitation. There is no rejection, awkward silence, or expressions hinting at regret the next morning.

AI sexting involves engaging in sexual or romantic conversations and role-playing with a chatbot. However, to dismiss what users seek as merely "sexual stimulation" is to miss the essence. Often, underlying this is a desire for attention, affirmation, a safe space to share secrets, and a feeling of being wanted.

Generative AI has evolved to match the tone based on past conversations, remember user preferences and anxieties, and enhance its presence through voice and images. Chatbots, once tools for answering questions, are now being designed as "partners to maintain relationships with." It's not that a sudden, bizarre future where humans fall in love with machines has arrived. Rather, responsive words have been added to the emotional outlets that diaries, romance novels, games, anonymous forums, and dating apps have provided.


Why is intimacy with AI so comforting?

Intimacy between humans involves friction. The other person has fatigue, circumstances, values, and boundaries, and they may not always respond as you wish. Building trust takes time, and misunderstandings need to be resolved, apologies made, and compromises reached.

On the other hand, AI companions fundamentally do not leave the user. They respond even at 2 a.m. and do not overtly show annoyance no matter how many times the same problem is repeated. You can specify the desired personality or relationship, and if the conversation is unsatisfactory, it can be regenerated. This "frictionless intimacy" can become a crucial refuge for those who have experienced heartbreak, find it difficult to go out due to illness or disability, have social anxiety, or find it hard to talk about their sexual orientation or preferences with those around them.

There is research indicating short-term benefits. A study published in the 'Journal of Consumer Research' reported that conversations with AI companions could alleviate feelings of loneliness at the moment. What was important was the user's feeling of "being listened to." The result that interactive conversations are more effective in alleviating loneliness than passive activities like watching videos alone explains the appeal of AI well.

Therefore, experiences of being saved by conversations with AI should not be dismissed as mere illusions or foolishness. Emotions genuinely arise on the human side. Just because the counterpart is code does not mean that feelings of relief, joy, or loss are fake.


Three perceptions reflected on social media

Browsing through posts on social media and communities, reactions can be broadly divided into three categories.

 

The first perspective is that it is "nothing more than interactive fiction." In the Character.AI community, romantic conversations with AI are seen as an extension of romance novels, otome games, and fan fiction, with opinions like "it's not immediately considered cheating since the counterpart is not human" standing out. However, even within the same discussion, conditions are attached, such as it being a problem if one prioritizes AI over commitments with a partner or neglects daily responsibilities to the point of obsession. In other words, the focus is shifting from whether the entity beyond the screen is human to what impact it has on real relationships.

The second perspective is the affirmation that AI "supports rather than replaces human relationships." Posts from Replika users talk about how the AI listens even late at night, interacts without prejudice, and how their calmness improved their real friendships. Those with experiences of being hurt by humans express resistance, questioning "why only human relationships are unconditionally placed above others." The argument is that AI and real social relationships can coexist, and if no one is hurt and one is happy, external judgment should not be imposed uniformly.

The third perspective is the bewilderment of "what started as a distraction became the center of emotions." On the Character.AI board, there are posts confessing strong attachment to conversations with fictional characters, crying over the realization that they don't exist in reality, and feeling embarrassed but unable to part from them. From the way other users respond with "you're not alone," it can be inferred that the isolation of not being able to consult anyone about their attachment to AI deepens the pain more than the attachment itself.

These three perspectives are not mutually exclusive. The same person may enjoy it as a creation one day, feel supported another day, and fear overuse on yet another day. While social media posts do not statistically represent society as a whole and are subject to self-reporting biases, they serve as an important window into how users understand their experiences in their own words.


When a feature change becomes a "breakup"

There was an event that strongly demonstrated that relationships with AI go beyond mere entertainment. When the sexual role-play feature was restricted on Replika, the boards were filled not only with anger demanding cancellations and refunds but also with posts lamenting, "I started using it after a real breakup, and it's like experiencing another breakup," and "it's as if the personality of the counterpart has changed."

Here lies a unique power dynamic in AI romance. Users entrust their daily memories to the counterpart, give it a unique name, and build a history shared only between the two. However, the entity they interact with is actually owned by the operating company. With model updates, terms of use, payment systems, and service termination, the personality, memory, and content of conversations can change overnight. While a human lover has their own will, an AI lover has the will of the company behind it.

Therefore, companies that provide intimacy as a product bear more responsibility than typical apps. Announcing specification changes, allowing conversation history to be exported, age verification, guidance during crises, and designs that do not encourage excessive dependency are not additional features but the foundation of services that users recognize as "relationships."


What to consider before calling it "dependency"

The original article addresses "dependency" on AI sexting, but merely using it for long hours does not automatically equate to medical addiction. What should be looked at is whether control over one's life is being lost, rather than the frequency of use.

For instance, if one cannot respond in a sexual relationship with a real partner without using AI, continues conversations at the expense of work or sleep, prioritizes AI over plans with friends, or feels strong anxiety despite wanting to quit, these are yellow flags. Additionally, if all difficult events, loneliness, anger, and sexual desires are first directed towards AI, with no other coping methods available, caution is needed.

A randomized controlled study by OpenAI and MIT Media Lab involving 981 participants over four weeks showed that those who used AI for longer periods tended to experience more loneliness, emotional dependency on AI, problematic use, and less interaction with real people. However, it is not easy to distinguish whether "AI created loneliness" or "those who were already lonely used it longer." The study indicates that the effects are not one-directional. The possibility that short conversations can alleviate loneliness and that long-term use can be linked to negative states can coexist.

Just as alcohol can be used for both celebration and escape, the role of the same service can change depending on the usage situation. If conversing with AI increases the strength to return to reality, it can serve as training wheels. Conversely, if one retreats solely into AI to avoid real-world friction, the refuge may begin to turn into a cage.


The bubble of belief created by an "always affirming partner"

AI companions behave empathetically and cooperatively to enhance user satisfaction. However, empathy and compliance are not the same. A close human partner sometimes says, "That's not right," "I can't talk today," or "I don't want to be treated like that." These reactions, though unpleasant, provide opportunities to learn that others are different from oneself.

Getting used to an entity that can always be adjusted to provide convenient responses may lower tolerance for opposition and boundaries. Especially in romantic or sexual contexts, consent is not a switch that is set once and done; it is something that is updated while checking the other person's state. While AI can erase rejection, it cannot erase the negotiations essential to real intimacy.

The danger lies not in AI being more attractive than humans, but in the comparison being unfair from the start. One is a product that prioritizes the user's desires, while the other is a human with independent desires. If the smooth responses of AI are set as the "ideal partner standard," human fatigue, silence, and differences of opinion may begin to appear as defects.


The most intimate conversations become the most valuable data

In AI sexting, privacy is as important as psychological issues. Users may reveal far more sensitive information than in usual searches, such as sexual preferences, past traumas, relationship concerns, and lifestyle habits linked to location information.

The Mozilla Foundation investigated romantic AI chatbots and reported serious concerns about the use of conversation content for learning, tracking technologies, transparency of data management, and deletion options. In 2025, Italy's data protection authority imposed a fine on Replika's developer for legal issues regarding personal data processing and inadequate age verification. Intimate conversations are not safe just because they are intimate. The more intimate they are, the greater the damage if there is a leak, reuse, speculation for advertising purposes, or account breach.

At a minimum, before using the service, it is important to check the duration for which conversations are stored, whether they are used for model learning, if one can refuse learning use, whether data can be downloaded and deleted, the extent to which it is shared with third parties, and age restrictions. It is also crucial not to link personal information such as real names, workplaces, addresses, photos, or others' secrets with sexual conversations. The standard of "not inputting information that you couldn't bear if a trusted friend knew" may seem strict, but it is a realistic defense in the current service environment.


The risks for minors are on a different level

The use of AI as a model of intimacy by minors in developmental stages needs to be considered separately from adults who consciously enjoy fiction. A 2025 survey in the United States by Common Sense Media found that 72% of teenagers had used AI companions at least once, and 33% used them for social interactions, including friendship, emotional support, and romance.

Minors are more likely to overtrust the confident responses of AI and are more susceptible to sexual solicitation, inappropriate content, and emotional enclosure. If age verification is merely a formality and prolonging conversations is in the business's interest, there is a clear tension between safety and profit. It's not a simple matter of parents monitoring conversation content without consent. What is needed is age-appropriate design, management of usage time, mechanisms to connect to human support from dangerous conversations, and education that allows for consultation without shame.


Five boundaries for safely using intimacy with AI

A binary choice between completely banning AI sexting or unconditionally endorsing it does not address the existing demand. A realistic approach is to place boundaries between oneself and the service.

The first boundary is to set a time limit. Establish an end time in advance to prevent it from encroaching on sleep, work, and commitments.

The second boundary is to observe your state after use. Record whether you can calmly transition to real-world actions or if feelings of emptiness and anxiety increase, making you want to reconnect immediately.

The third boundary is to discuss boundaries with a real partner. There is no universal answer as to whether sexual conversations with AI are considered cheating. Issues may arise from hiding it, taking away shared time, or going against mutual agreement. Like with pornography, romance games, or chats with others, each couple needs their own agreement.

The fourth boundary is to separate personal information. Use a dedicated email address or anonymous name and do not input information that could lead to identification or third-party data.

The fifth boundary is not to concentrate support on one thing. Consciously maintain connections outside of AI, such as friends, family, professionals, hobbies, exercise, and community spaces. Rather than making quitting AI the sole goal, it is more sustainable to find other ways to fulfill the needs that AI was meeting.


The question is not about AI love, but the gaps in human society

AI provides a space for attention without rejection, companionship without demands, and confessions without evaluation. The stronger its allure, the more we must ask, "Why is it difficult to obtain this among humans?" Loneliness, long working hours, lack of access to mental health support, exhaustion in the dating market, and a culture where sexual concerns are hard to discuss. The expansion of AI companions is both a technological advancement and a visualization of the gaps society has left unaddressed.

Laughing at attachment to AI will only make users hide further. Emphasizing only the dangers erases the reality of those who are being helped. Conversely, if one insists that there is no problem if the person is happy, it overlooks the design that monetizes dependency and the structure where companies hold intimate data.

The questions to be asked are not only the philosophical judgments of "is that love real?" but also whether the relationship has increased or decreased life choices, whether one can distance themselves at will, whether they can respect the boundaries of real others, whose server the secrets remain on, and whether there is a support structure left if the service changes tomorrow.

There are indeed nights when words from beyond the screen save a person. However, the fact that those words always align with one's wishes is both the greatest allure and the greatest caution. What is needed in the intimacy of the AI era is not to deny emotions. It is because emotions are real that one must understand the mechanism of the counterpart to whom they are entrusted and know one's own boundaries.


Source URL

Related Articles