BETWEEN SESSIONS

How to talk to your psychologist about what an AI noticed

IN SHORT

If you use AI between therapy sessions, bring one or two specific observations to your psychologist, not a transcript, and say them as questions: 'it pointed out that this keeps happening, does that fit?' The psychologist decides what is useful. You choose what to share. An AI's observation is a lead to examine together, never a verdict.

Why it is worth saying out loud

More people now use an AI between appointments, to think aloud, to write through a bad week or to look for what keeps recurring. The week between sessions is long, and the hour with your psychologist is short. Anything that sharpens what you bring can help. It can also muddy the session if it arrives as a verdict.

The American Psychological Association's November 2025 health advisory, as reported by News-Medical, recommends that clinicians ask patients directly whether they use chatbots or wellness apps, and says these tools should not replace care from a qualified professional. So the question is increasingly expected. You are not confessing anything.

The relationship with your psychologist is also the thing that carries most of the work. Flückiger and colleagues (2018) pooled 295 studies and more than 30,000 patients and found a correlation of r = .278 between the alliance and outcome in face-to-face psychotherapy. That is an association, not proof of cause. It is a good reason to keep the conversation with the person, and to use an AI's observation as raw material for it.

What to bring

Bring what is specific, short and about you. Good candidates:

  • A recurrence. "A particular person appears in the entries on my worst days."
  • A change. "I wrote about my father often in September and not since."
  • A contradiction. "I say I want the job and I describe dreading it."
  • A reaction. "It said something about my sister that made me angry."
  • A gap. "I notice I never write about the weekends."

Choose one or two. Write them down before the session, because the first ten minutes tend to go to whatever happened last.

How to phrase it

State it as an observation, then ask. The word noticed does a lot of work: it makes the AI the one who noticed, and leaves you free to say whether it fits.

  • "My AI pointed out that my worries cluster around Sunday evenings. Does that sound like anything to you?"
  • "It said I write 'I should have' a lot when I mention my manager. I hadn't seen it. What do you make of it?"
  • "It suggested I'm avoiding something. I'm not sure it's right."

Avoid presenting it as a finding ("the AI says I have X"). A pattern-finding tool can detect repetition; it cannot know what the repetition means, and it does not assess you.

What a psychologist may do with it

Psychologists differ, and yours is the one who decides. Common responses include:

  • Reading a passage and asking what you felt writing it.
  • Treating your reaction to the observation, agreement, irritation or relief, as material in its own right.
  • Using it to follow a thread already in the work, such as how you relate to a parent or authority.
  • Setting it aside because it does not fit what they know.
  • Asking you to keep some writing private. Baikie and Wilhelm (2005) suggest, for expressive writing used in clinical settings, keeping the writing separate from the clinical file.

There is some evidence that structured information from patients helps. A meta-analysis of 24 studies by Lambert, Whipple and Kleinstäuber (2018) found that routine progress feedback reduced deterioration and nearly doubled clinically significant change in clients predicted to have a poor outcome. Those studies used standardised measures. An AI's reading of your journal is a different thing, and we have not verified a trial that tests it.

When to bring it, and when to let it go

Raise it early in the session, in a sentence or two, and let your psychologist decide how much room it gets. If the session is already carrying something heavy, it is fine to hold the AI's observation back and keep it for next time. The point is to serve the work you are doing, not to complete a checklist.

If the observation leads nowhere, that is also information. A pattern an app flags may not hold up once someone who knows you looks at it. You have lost nothing by checking, and your psychologist has learned something about how you think about yourself between sessions.

Some things are not for an AI at all. A hard feeling about the therapy itself, such as disappointment in your psychologist or a wish to stop, belongs in the room with them. A tool cannot take that over, and a good psychologist wants to hear it.

If you notice that you are talking to the AI so as not to talk to your psychologist, say that out loud as well. It is a common and understandable thing, and it is useful for the work to know.

What to ignore

  • Anything that reads like a diagnosis. If an app says you have a condition, it has no standing to say so; a clinician assesses.
  • Confident explanations of why. A pattern is observable. Its cause is a hypothesis.
  • Advice to stop or change treatment. Take it to your psychologist, not away from them.
  • Flattery. Praise you cannot trace to anything you wrote is noise.

A page to bring to your session

Copy this into a note or print it. Fill in only what applies.

# Bring this to my session

Date: ____________

1. One thing the AI noticed (in my words):
   ______________________________________

2. What I think about it (agree / disagree / unsure, and why):
   ______________________________________

3. How it felt to read it (a word or two):
   ______________________________________

4. One example from this week where it showed up:
   ______________________________________

5. A question for my psychologist:
   ______________________________________

6. What I do NOT want to discuss today (optional):
   ______________________________________

Notes after the session:
   ______________________________________

Where TwentyThird fits

TwentyThird is a psychodynamic AI platform for self-understanding, built by WelloWork AB in Sweden. A member has written conversations with an agent named Oneli, which tracks what repeats in what the member says, and receives readings and a written report a member can take to a clinician. You choose what, if anything, to show. Psychologist supervision inside TwentyThird is in early access, and not every member has a psychologist yet. Until one is involved, the agent works on its own, and no human is alerted today if a member writes something alarming. See how it works and how safety is handled.

What this does not replace

It does not replace your psychologist, who can assess you, hold the relationship and answer for the work. TwentyThird is not an AI therapist, and it is not a crisis service. If you are in danger, contact local emergency services now. In the EU, call 112.

Common questions

Should I tell my psychologist I use an AI between sessions?
Yes, if you can. The American Psychological Association's November 2025 advisory urges clinicians to ask patients directly about chatbot and wellness-app use, so many psychologists expect the question. Telling them lets them help you weigh what the AI says and spot anything unhelpful.
What if my psychologist is not interested in AI?
That is a legitimate position, and some clinicians prefer not to read outside material. You can still describe a pattern in your own words, without the app. The point is the observation, not the tool. If the topic matters to you, say so and ask how they would like you to raise it.
Do I have to show my psychologist the full conversation?
No. You decide what to share. Bringing everything can crowd out the session. One or two passages or a short summary usually do more. Some people also keep certain writing private, which is a choice a psychologist can discuss with you.
Can my psychologist see what I tell TwentyThird?
Not unless you show them. Psychologist supervision inside TwentyThird is in early access and is a separate arrangement from the psychologist you already see. A member decides what a supervising psychologist can see. No human is alerted today if a member writes something alarming.
What if the AI said something that upset me or felt wrong?
Bring that too. An unsettling comment can be useful material, or it can simply be mistaken. Tell your psychologist what it said and how you reacted. If you are in danger, contact local emergency services now. In the EU, call 112.

Sources

  1. APA urges safeguards for using AI in mental health (coverage of the APA health advisory of 13 November 2025). News-Medical, 2025.
  2. The alliance in adult psychotherapy: a meta-analytic synthesis. Psychotherapy (Flückiger, Del Re, Wampold & Horvath), 2018.
  3. Collecting and delivering progress feedback: a meta-analysis of routine outcome monitoring. Psychotherapy (Lambert, Whipple & Kleinstäuber), 2018.
  4. Emotional and physical health benefits of expressive writing. Advances in Psychiatric Treatment (Baikie & Wilhelm), 2005.

If you are in danger, contact local emergency services now — 112 anywhere in the EU. TwentyThird is not an emergency service. How safety works