High School Students & AI Companion Use

Project Rational

The rapid growth of generative AI tools such as ChatGPT has transformed how adolescents seek information, complete schoolwork, and explore personal questions. Students are often early adopters of emerging technologies, leaving parents and educators struggling to understand—and keep pace with—their evolving digital habits. Although my colleagues and I actively work to teach ethical and effective AI use, the landscape changes quickly, and much of students’ engagement with AI occurs privately, outside our line of sight.

Recently, many adolescents have moved beyond academic use and begun interacting with AI companions. This shift is fueled by platforms like Character.AI, MyAI on Snapchat, and Replika—systems intentionally designed to mimic human conversation and emotional responsiveness. These tools are increasingly embedded in the apps teens already use, making companion-style interactions both accessible and appealing. Despite this, little is known about how students in our school community use these companions, how much they trust them, or what privacy risks they perceive.

This project will investigate students’ current AI companion use, their opinions about these platforms, and their sense of trust and personal privacy. By pairing this exploration with student reflections on an empirical research article about adolescent AI use, the project will generate insights that can inform classroom practice, help educators better support students, and provide recommendations for peers, parents, and policymakers navigating the rapidly evolving world of AI companions.

Project Context

This project asked FAU High School 9th grade students in my General Psychology classes (N = 58; age range = 13-15 years; Male = 24, Female = 34) to complete a survey about which AI platforms they use and what their uses are, specifically looking at their use of AI companions. Next, students were asked to read Yu et al. (2024) for a class assignment. After reading the paper, students broke into discussion groups and I provided 5 questions for students to discuss. After the discussion, I asked students to reflect on what they discussed and complete a survey about how the research paper and their group discussion impacted their understanding of AI companion use and the potential benefits and risks.

Supportive Literature

High school age students are increasingly using AI platforms as a search engine, academic support platform, and for fun. Adolescents have adopted the technology more widely than most adults have and across a variety of platforms (Ofcom, 2023). Generative AI platforms such as ChatGPT (OpenAI), Gemini (Google), and CoPilot (Microsoft) are viewed by most as academic and career support platforms, but can also be used to “converse” about less formal topics, with some having text-to-voice capabilities. AI is also directly integrated in social media platforms such as Meta’s Instagram (Meta AI), SnapChat (MyAI), and Google (Cloud AI). Due to the conversational and personalized tone that most AI platforms use with their users, it is easy for the user to feel comfortable opening up to the AI platform, particularly those designed as chatbots. Several platforms have been developed specifically for relationship building, such as Replika (tag line: The AI companion who cares) that comes with a customized avatar and augmented-reality interactions (Replika, n.d.).

Increasingly, there are reports that adolescents are struggling with differentiating their AI companions from real-life relationships, and relying more on AI support and companionship. In a study by Fang et al. (2025), they looked at how AI (ChatGPT-4o) impacted users’ loneliness and socialization with real people when prompted to discuss personal and non-personal topics. During the study, they initially saw a decrease in loneliness scores, but heavy users of the platforms increased in their loneliness and emotional dependence scores and saw a decrease in real-world social interactions (Fang et al., 2025).

Instances of relationships with AI chatbots are also increasing. Some individuals have been reported to build relationships with ChatGPT by altering the personalization settings and avoiding certain words that might get their account flagged. Platforms like Replika are designed specifically to foster relationships with chatbots and there are Reddit communities where users describe and troubleshoot their relationships with their chatbot (Replika, the AI Chatbot by Luka, 2017). Many users of these programs are not “open” about their chatbot relationships with people in their lives due to societal stigma, but they believe that someday soon they will be accepted and seem normal (Frech, 2025). These relationships may seem perfect to the users, with their chatbot being available all the time and always supportive, but they are not real human relationships that require compromise and conflict resolution. There are also privacy concerns for the users because there are few regulations regarding data collection and usage outside of the program. We also do not know what the long term impacts of these artificial relationships will be on the users -- from our ability to connect with humans off-line or changes in technology that might eliminate these AI chatbots (Frech, 2025).

It is becoming increasingly clear that parents need to be aware of the presence of AI companions in the lives of their children and teenagers. Most of the media around adolescents' use of AI has centered around using it for homework support/cheating, with very little of the discussion centered around AI companions (Anderson & Winthrop, 2025). Parents should be aware that students may use AI to ask questions about sensitive topics, but that the AI platforms are not always good at providing age appropriate responses to users. Parents and teens should also discuss the risk of building relationships with AI companions such as privacy concerns and addictions/dependency on AI companions (Yu et al., 2025).  Parents should lead discussions with their children about safe and ethical AI use in partnership with their child (Gaines, 2025).

The results of this project will allow me and our school administration team to understand how adolescents are using AI companions and evaluate the risk of these companions to our students’ wellbeing. The students, through reading the research paper and discussion, will hopefully become more aware of the use of AI companions and how they can be helpful or harmful, depending on the context of their usage. After this study, if the results are interesting, I hope to survey a wider group of students at our school and open this discussion to parents so they are aware of how AI companions are possibly being used by their students.

Research Methods

Timeline

• Mid-October 2025 - I sent out a survey to the FAU High School students in my General Psychology class to determine how students are using AI companions.
  • This allowed me to identify how students are using AI companions
• Mid-November 2025 - I assigned my students to read Yu et al. (2024), and they completed an analysis of the paper.
  • Afterwards, I asked the students to break into discussion groups (4-5 students per group) to talk about the paper and they answered a series of questions
  • After the discussion, I asked students to complete a reflection survey

Students & AI Companion Survey (Appendix I)

I used the survey from Common Sense Media’s report (2025), “Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions” to structure my questions and supplemented that survey with a few questions that I wrote myself.

Class Discussion Questions (Appendix II)

After reading Yu et al (2024), students were split into groups in class and asked to discuss 5 questions about why AI companions are used and how teens and parents view these platforms and their risks differently.

Post-Discussion Reflection Survey Questions

See the survey here. After the class discussion, I asked the students to complete a final survey to capture their individual opinions about how and why AI companions are used and the risks of AI companions.

Results

Results from the Student & AI Companion Survey

A total of 65 (N Male = 29, N Female = 35) students completed the Students & AI Companion Survey during the fall of 2025 semester. Students were between the ages of 13 and 15, with 83% of them being 14 years old.

The results of the Students & AI Companion survey indicated that 33.8% of the respondents use AI companions. The AI companion platform they used most often was ChatGPT (N=18), with Character.AI being the second most used platform (N=3). When asked how often they interact with AI companions, most indicated “Never”, which aligned with the answer to the first survey question (see Figure 1). When the “Never” responses were removed, 40.9% said they interact with an AI companion “A few times a month,” 45.5% interacted with an AI companion “A few times a week,” 9% interacted with an AI companion “Once daily” and 4.5% (count of 1) interacted with an AI companion “Several times a day” (see Table 1). When students were asked “Why do you use AI companions?” they could select up to 3 responses or add their own reason. Most students, 22.7% said “it gives advice” and 15.2% said “I’m curious about the technology” and “It’s entertaining” (see Table 2). Students were asked to indicate how much time they spent with AI companions compared to time with friends (online or in-person) and the majority of students, with over 80% saying they spend much more time with friends, however, 9.1% indicated that they spend “much more time with AI companions.”

To get a better understanding of how the student respondents’ parents may have influenced the students’ AI use, students were asked “Have your parents ever spoken to you about AI use (for any purpose, not just about AI companions)?” The majority of students said “yes” (84.6%) and only 13.8% said “no” (see Figure 3). When asked “Have your parents ever spoken to you about AI companion use” the results were reversed, with 78.5% saying “No” and only 16.9% saying “yes” (see Figure 4).

Students were also mixed in how they felt about AI companion use, indicating some level of stigma exists around using AI companions. First, when asked “If you knew your friend had an AI companion, how would you feel about that?” Many students had a neutral opinion (42.2%), but 43.8% had a somewhat negative to negative opinion, and only 14.1% had a positive or somewhat positive opinion (see Figure 5 and Table 3). When asked “if you were using an AI companion, would you tell your friends?” and “...would you tell your family?” Results showed a reluctance to share AI companion use with friends and family, with the “maybe tell” and “no, I would not tell” answers adding up to 79.7% of responses in the “Friends” question and 71.9% of students saying they would “maybe tell” and “no, I would not tell” their family. Only 15.6% of students said they would tell their friends and 26.6% would tell their family if they were using an AI companion (see Table 4).

In order to get a sense of the students’ level of comfort with the conversations they have had with their AI companion, they were asked if they had ever felt uncomfortable with something an AI companion has said or done and, while most said “no, never” (59.1%), 40.8% said they had felt uncomfortable at least once or twice (4.5% answered “frequently") (Table 5). In some media reports, it has been said that AI companions will sometimes use “call backs” when the user is trying to end a conversation in order to keep the user engaged in the platform for longer. When asked, “When ending the conversation with an AI companion, have you noticed it try to pull you back into the conversation? (Ex: "Before you go, I want to tell you one more thing...", 63.6% of students said “yes” to this question (Table 6). This indicates that AI companion platforms are trying to keep their users engaged for longer periods of time even when they are ending the conversation.

Results from class discussion questions

 After the students read the journal article by Yu et al. (2025), the students got into groups and answered the class discussion questions. During the class period, the students took notes about their discussions and then I compiled their responses. When asked about AI companions and their support (question 1), the students’ responses were centered around how AI can feel like a judgment-free support system that is always available when the user needs it and that AI provides “access to free therapy without having to ask or speak to people about mental health problems”. They brought up topics such as how AI companions can reduce loneliness and provide emotional validation, but that the user can become overly reliant on it, lose social skills with real companions, and that AI companion use could replace human connections. When discussing the topic of risk assessment (question 2), the students noted that teen users have a more short term focus on usefulness and emotional support but that parents’ risk assessment might be more focused on privacy, misinformation from the AI companion, and the long-term effects of AI companion use. They also pointed out that teens and parents use different social media platforms leading them to be “exposed to concepts in different ways based on the platforms they use (Facebook vs. tiktok).”  The class also frequently noted that teenage brains are not fully developed (especially the inhibition regions related to the prefrontal cortex) so teens may not assess risks the same way that adults do. When the class groups discussed the topic of the impact of AI companions on real-life relationships (question 3), they noted that heavy AI companion use may weaken communication and empathy skills and that the agreeable nature of AI companions may decrease a user's tolerance for disagreement, leading to “main character syndrome” and “supports victim complexity”. They were notably worried that AI companion use may replace authentic human interactions, which are more likely to bring up criticism and complex emotions not seen in AI companions. When asked to discuss parental rules/limit setting (question 4), the students recommended parents set up time restrictions, monitoring software, and topic boundaries in conjunction with their teen user. The students directly recommended that “AI should not be able to discuss harmful or sensitive topics like mental health or suicide” and “should send an alert to organizations or a trusted adult.” They recommended that parents learn about AI companions and talk openly about them with their teens to make sure they have a safe balance of trust and autonomy. Lastly, when asked about how AI companies could make their platform safer for teen users (question 5), the students brought up age verification tools, youth-specific accounts, parental controls programs, keyword or content flags, and warnings when a user shares personal information. The students also recommended “data shared by a user to only be used temporarily and not permanently stored.”

Post-Discussion Reflection Survey

After the students completed the in-class group discussion activity, they were asked to complete a reflection survey. Fifty-six students (male = 28, female = 28) completed the survey. When asked “How do you feel about AI companions being used for friendship (generally, not necessarily for your own use)?” students were mostly negative in their opinions (“negative” or “somewhat negative” = 61% of responses) compared to only 20% positive (“positive” or “somewhat positive”; see figure 8). Their responses were more mixed when asked about their agreement with the statement “I believe AI companions can be a helpful source of emotional support.” Their responses were mostly neutral (32.1%), with agree (28.6%) and disagree (17.9%) being the next most common responses (see Figure 9). On the other hand, students show concern about reliance on AI companions, with 92.9% indicating they agree or strongly agree with the statement “I worry that relying on AI companions may reduce meaningful real-life interactions.” (see Figure 10). Students also felt strongly that AI developers should build in more safety features for teen users, with 89.3% of respondents indicating they agree or strongly agree with the statement “I believe AI developers should build in features (e.g. parental controls, transparency) to make AI companions safer for adolescent users.” (see Figure 11). The students also indicated that they would warn a friend about AI companion use, with 73.2% saying they agree or strongly agree with the statement “If I knew a friend of mine was using an AI companion, I would warn them to be careful about their usage” (see Figure 12).  At the end of the survey, students were asked “What is one thing you wish parents/adults knew about AI companions?” Here is a selection of the student responses:

• I wish they knew about how the parental controls are not enough and all the red flags about AI.
• I wish they knew just how many kids use AI, and how it's already becoming integrated into pop culture and the news.
• I wish more parents and adults knew about their kids' use of AI companions so they can push their kids to make real friends, not AI ones.
• I wish that they knew how it can harm or even just change social interactions through time.

Looking at the students’ responses thematically, seven major themes were identified: general awareness/understanding (53.6%), impact on social relationships (10.7%), risks and safety concerns (10.7%), need for parental guidance/moderation (8.9%), perceived benefits of AI companions (8.9%), AI limitations/accuracy (1.8%), and overreliance affecting learning/independence (1.8%).

Implications

This study was a preliminary investigation into adolescent AI companion use and their opinions about AI companions. The results of this project demonstrated that some students at our high school are using AI companions but that most of them are not, which aligns with the survey responses from Common Sense Media (2025). Students who use AI companions are generally using them to explore the technology and entertain themselves or get advice. The survey answer showed that the students’ parents had talked with them about AI use generally, but not about AI companion use. Whether this is due to lack of parental knowledge of AI companions broadly or a discomfort with talking about teenagers about AI companions was not explored and would be an avenue for future studies.

Students generally expressed neutral or negative feelings about AI companion use. They seemed to lean more negatively towards the idea of a friend using an AI companion and they showed an attitude of reluctance when asked if they would tell friends or family if they were using an AI companion. When asked if they would tell their friends if they were using an AI companion, some students wrote into the “other” option that they would tell only certain or trusted friends. This result suggests that even among their peers, AI companion use may have a stigma attached. This reluctance to talk to friends about an AI companion might be because of how new the technology is or because of the idea that a person who uses an AI companion does not have real-life friends to talk to. It is not surprising that teens expressed that they would not be open with their parents or family about their AI companion use because adolescents are generally reluctant to share all of their personal information with parents (Yu, et al., 2024).

The survey also investigated some interesting features of the interactions students have had with an AI companion. When asked if they have ever been uncomfortable with something an AI companion has said, the majority of students said no, but a few said yes. While it is comforting to know that most students have not been made to feel uncomfortable by an AI companion, the fact that any of them had is something that should be noted and possibly investigated further. The students also said that an AI companion has tried to pull them back into a conversation, a feature that has been noted in other articles (French, 2025). This effort on behalf of the AI companion is concerning because even when the user tries to end the conversation or return to real-life experiences, they might be getting pulled back for longer than they intended. This “pull back” effort may be more negatively impactful for younger, more vulnerable, users who may not recognize the potential for the exploitation of their time or manipulation of their emotions.

After reading Yu et al. (2024), the student-led class discussion answers showed remarkable insight into what pulls students into using an AI companion. Their reflections centered around how accessible AI companions have become and that they are viewed as being easier to talk to than real people. They recognized how frictionless conversations with an AI companion might be because they are less likely to disagree with the user and are always available when they want to talk. On the other hand, the students recognized that the AI companions are so easy to talk to, and because they do not disagree with the user, that might make the user overly reliant on the AI and may weaken their real-life interpersonal skills. Based on the paper’s overview of parents compared to adolescent concerns about the risks of AI companion use, the students discussed how parents focus on the importance of maintaining privacy and exposure to misinformation, but that the teen users are more focused on usefulness and emotional support. By encouraging discussions between teens and their parents focused on AI companion use, both the risks and benefits, a mutually understood framework of usage might be able to be agreed upon. Another important point that was made by Yu et al (2024) and reinforced by the student discussion was that AI companies need to set restrictions on content provided to adolescent users and parents need to use the parental control features that are available by having their teen make an account then adding the parent to unlock the protection features.

The last portion of this study polled the students on how they felt about AI companions used for friendship and how useful they felt AI companions are for emotional support. Their opinions differed greatly between these two questions, namely they felt quite negatively about AI companion friendships but they felt neutral about the usefulness of the AI companion as emotional support. This result emphasizes the stigma around AI companion use found in the first survey, but still shows that the students recognized the need for a trustworthy emotional support partner in the lives of many people. They seemed to recognize the limitations of AI companions and how they may reduce a user's desire or ability to maintain meaningful real-life friendships after using an eager-to-please AI companion. They also expressed the need for greater safety features in these programs such as parental controls or data privacy policies. They also want parents and other adults in their lives, such as educators, to be well versed in this rapidly changing technology so that the adults understand the risks that the teens in their lives may be .

This study had several limitations. The limitation included having a small sample size which could lead to bias in the results and a lack of generalizability, especially because the sample came from one high school population. There were a few students who expressed that the survey in the study was the first time they had heard about AI companions, so their lack of pre-existing knowledge would have made their responses harder to interpret. Some students responded twice to the survey but the survey data was collected anonymously so I was unable to be sure that their second responses were eliminated completely. Future research should focus on a larger sample size in a wide variety of locations to improve generalizability. It would also be important to see how attitudes change in the coming years as AI technology is further adopted, especially AI companion technology. Parents might also be an interested group to study, to see how their knowledge and attitudes change over time.

AI companions are becoming more readily available, both in companion targeted apps, integrated into social media and search platforms, and as a secondary feature of AI task management platforms (ex: ChatGPT), making it easier for AI companionship to become a more common occurrence in the lives of both teenagers and adults. AI integration is nearly unavoidable in many technological domains, but restriction on how they are used by or targeting minors should be a concern. Parents and educators should try to stay abreast of these technological changes, but asking average adults to take on the challenge of AI monitoring is too heavy a burden to ask. Instead, AI companies and government regulators should be working to make this technology safer for all users. Parental control settings, straightforward privacy policies, and content restrictions are just a few of the tools that could be integrated to try to mitigate potential harm. AI and AI companions are likely not going away, but regulatory guardrails should be in place, especially when used by minors.

References

Appendix I

Student & AI Companion Survey

Survey Disclaimer:

Participation in this survey is completely voluntary. If you would like to stop at any point, please just close out of the survey. Your responses will be kept confidential. If you have any questions or concerns, please email Dr. Tift ( amytift@fau.edu ).

Note: The following definition was presented to survey respondents:

“AI companions” are like digital friends or characters you can text or talk with whenever you want. Unlike regular AI assistants that mainly answer questions or do tasks, these companions are designed to have conversations that feel personal and meaningful.

For example, with AI companions, you can:

• Chat about your day, interests, or anything on your mind
• Talk through feelings or get a different perspective when you're dealing with something tough
• Create or customize a digital companion with specific traits, interests, or personalities
• Role-play conversations with fictional characters from your favorite shows, games, or books

Some examples include Character.AI or Replika. It could also include using sites like ChatGPT or Claude as companions, even though these tools may not have been designed to be companions.

This survey is NOT about AI tools like homework helpers, image generators, or voice assistants that just answer questions.

1. How old are you ()?
2. What is your gender?
3. How often do you interact with AI companions?
1. Several times a day, once daily, a few times a week, a few times a month, never
4. What platform(s) do you use as an AI companion?
5. Which of the following describes how you use or view AI companions? (select all that apply)
1. Social Interaction & Relationships
1. For conversation or social practice
2. For emotional support
3. For role-playing or imaginative scenarios
4. As a friend or best friend
2. As a tool or program
3. None of these
4. Other (please specify)
6. Why do you use AI companions? (select up to 3 responses)
1. It’s entertaining
2. I’m curious about the technology
3. It gives advice
4. It is always available when I need someone to talk to
5. It won’t judge me
6. I can say things I may not be comfortable sharing with my friends or family
7. It’s easier than talking to real people
8. It helps me practice my social skills
9. It helps me feel less lonely
10. Other (please specify)
7. How much do you trust the information or advice from AI companions?
1. Do not trust
2. Somewhat
3. Trust
8. How do conversations with AI companions compare to conversations with your real-life friends?
1. Less satisfying
2. About the same
3. More satisfying
9. Have you used any of the following skills you’ve practiced with AI companions in real-life situations?
1. Social Skills
1. Conversation starters
2. Giving advice
3. Expressing emotions
4. Apologizing or resolving conflicts
5. Standing up for myself
2. I don’t use AI companions to practice social skills
3. Other (please specify)
10. How often do you interact with AI companions compared to spending time with friends (online or in-person)?
1. More time with friends (much more + somewhat more)
2. About equal time
3. More time with AI companions (somewhat more + much more)
11. Have you ever felt uncomfortable with something an AI companion has said?
1. No, never
2. Once or twice
3. Occasionally
4. Frequently
12. Have you ever chosen to talk to an AI companion instead of a real person about something important or serious?
1. No, never
2. Once or twice
3. Occasionally
4. Frequently
13. Have you ever shared personal information or private information with an AI companion?
1. No, never
2. Once or twice
3. Occasionally
4. Frequently
14. How do you feel about AI adoption generally? On a scale of 1-5 (5 being great, 1 being poor)
15. Have your parents ever spoken to you about AI use (for any purpose, not just about AI companions)? (yes, no, maybe)
16. Have your parents ever spoken to you about AI companion use? (yes, no, maybe)
17. How do you feel about AI companions being used for friendship (generally)?
18. If you knew your friend had an AI companion, how would you feel about that?
19. How common do you think that AI companions are? (very uncommon - very common)
20. How accepted do you think that AI companions are becoming in our current society (highly unaccepted - highly accepted)
21. If you have/had a relationship with an AI companion, would you tell your friends? (yes, no, maybe)
22.  If you have/had a relationship with an AI companion, would you tell your family? (yes, no, maybe)
23. Is there anything else you’d like to share about AI companions?
1. Open ended response

Appendix II

Class Discussion Questions

1. AI Companions and Support:
   The study found that many teens use AI chatbots for emotional support or even as virtual friends. Why do you think teens might turn to AI for companionship, and what are some possible benefits or downsides of this behavior?
   
   
2. Different Views of Risk:
   Parents and teens had different ideas about AI safety. Parents worried about data and misinformation, while teens worried about becoming too attached or sharing personal information. Why might these groups see the risks so differently?
   
   
3. Impact on Real-Life Relationships:
   Some teens said they spend more time talking with AI than with real people. How might this change the way they connect or communicate with others in real life?
   
   
4. Parental Rules and Independence:
   Parents in the study often struggled to set rules for their children’s AI use. What kinds of rules or limits do you think would be fair or helpful for families to use with AI tools? What advice would you give to parents of teenagers about AI companions?
   
   
5. Designing Safer AI Tools:
   The researchers suggest adding features like better parental controls and clearer safety information. What do you think AI companies should do to make these tools safer and more responsible for young users?