You can tell an AI chatbot is helping the person you support when they seem more able to name what they feel, follow through on small daily routines, or handle a tough moment without shutting down. It is not helping if they come away more isolated, more self critical, or using the chat to avoid real life contact and professional care.
What positive changes should I look for?
Help rarely looks like a dramatic transformation. It usually shows up in small, ordinary shifts. You might notice the person starts putting feelings into words more easily, saying things like "I was overwhelmed at work and used that grounding thing the chat suggested" instead of just "my day was bad." They may mention a coping skill they tried, such as pausing to breathe, writing down three worries before bed, or stepping outside for five minutes.
Other signs include:
- They seem less alone at night or during gaps between appointments, because they have somewhere to talk things through.
- They bring up topics with you that they previously avoided, even if only briefly.
- They follow through on small commitments, like eating at regular times, going for a walk, or replying to a message.
- They distinguish between what the chatbot offers and what a person offers, and they still seek human contact.
- They describe the chatbot as one tool among several, not the only thing keeping them afloat.
What this can feel like from the outside is a slow loosening. The person may be a little less reactive, a little more curious about their own patterns, or slightly more willing to accept help. Progress is uneven, and a good week does not mean everything is fixed.
What are the red flags that it is not helping?
Some warning signs are easy to miss because the person may sound engaged. They might chat for hours and report feeling better in the moment, while their real world functioning stays the same or worsens. Red flags include:
- They use the chatbot instead of sleeping, eating, or leaving the house.
- They hide the extent of their use, or become defensive when you ask about it.
- They repeat the same loops without any change in behavior, venting nightly but never trying a new step.
- They treat the chatbot as a best friend, partner, or sole confidant, and pull away from people.
- They quote the chatbot as an authority on their mental health, such as saying it told them they have a condition.
- They seem more hopeless, more self critical, or more convinced that no human could understand them.
- They use the chat to rehearse self harm, suicidal thoughts, or other serious distress without telling anyone who can help.
If the person mentions thoughts of suicide or self harm, take it seriously. In the United States, you can call or text 988 to reach the 988 Suicide and Crisis Lifeline, or contact the NAMI HelpLine for guidance. Do not rely on a chatbot to manage a crisis.
How do I ask about their experience without pushing them away?
The goal is curiosity, not interrogation. People often shut down when they sense you are building a case against something they rely on. Start with an open question that assumes the chatbot might be useful.
Try something like:
- "What do you usually talk about with it?"
- "What does it do for you that talking to people doesn't?"
- "Have you noticed anything it says that actually sticks with you?"
- "Is there anything about it that feels off or unhelpful?"
Listen before you evaluate. If they say the chatbot helps them fall asleep, that is real information. If they say it never judges them, that tells you what they need from support. You can then ask a gentle follow up such as "Do you feel like it pushes you to try things, or mostly just listens?" That question gets at whether the tool is helping them move forward or simply absorbing hours of distress.
Keep your tone neutral. Saying "I'm glad you have something that helps at 3 a.m." can make it safer for them to also admit when it is not helping. If you only ask skeptical questions, they may defend the chatbot and hide their doubts.
When should I encourage traditional therapy?
AI chat can offer a place to talk things through, practise a skill, or feel less alone in a hard moment. It is not therapy, and it does not replace a licensed clinician. You might gently raise the idea of seeing a therapist when:
- The person's distress is intense, frequent, or interfering with work, sleep, relationships, or basic care.
- They are using the chatbot as their main or only form of support.
- They mention thoughts of harming themselves or someone else.
- They have a history of trauma, panic, deep depression, or other serious concerns that need a trained professional.
- They are not improving, or they are getting worse, despite regular chatbot use.
- They want to understand a diagnosis or explore medication, which requires a doctor or psychiatrist.
You can say something like "It sounds like the chat helps you get through the night, and I wonder if a therapist could help with the bigger picture." That keeps the door open instead of framing the chatbot as a mistake. If they are open to it, this guide on How to Choose an Individual Therapist: What to Look For walks through practical steps. The National Institute of Mental Health also explains what psychotherapy involves and how to find it.
For people who are curious whether AI chat has any evidence behind it, our page on Does AI Therapy Actually Work? separates what research suggests from what it does not.
How can I support without overstepping?
You cannot control how someone uses a chatbot, and trying to usually backfires. What you can do is stay present, stay honest, and keep your own boundaries clear.
- Ask what they find helpful, and affirm the parts that genuinely seem useful.
- Share your concerns once, calmly, without repeating them every day.
- Offer other forms of connection, like a walk, a meal, or a phone call, so the chatbot is not their only outlet.
- Know the limits of your role. You are a friend or family member, not their therapist or their monitor.
- If you are supporting someone with depression, this guide on What to Say to a Friend With Depression: Support That Helps offers concrete phrases that tend to land better than generic reassurance.
Notice your own reactions too. It can be painful to watch someone turn to a screen instead of you. But if you make your support conditional on them quitting the chatbot, they may hide both their chatbot use and their struggles. Staying in their corner, even while you disagree with one of their tools, keeps the door open for the moments when they really need a person.
Trust builds slowly. The fact that you are asking how to tell if an AI chatbot is helping means you are paying attention, and that attention itself is a form of care. Use it to stay curious, to ask better questions, and to guide them toward human support when the chatbot is no longer enough.