TL;DR
Get gifts for the two of you delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A Gottman Institute report describes clients using AI to interpret partners’ behavior, draft communication scripts and analyze text messages. The author argues that AI may make users feel more certain without establishing whether their interpretations are accurate or helping couples connect.
A therapist writing for The Gottman Institute says clients are bringing AI-generated explanations of their partners’ behavior, communication scripts and interpretations of text messages into therapy. The report warns that these tools can leave someone more certain without being more accurate, because an AI system may hear only one person’s account and cannot independently establish a partner’s intent.
The author says AI can be useful for organizing thoughts, considering different perspectives and questioning assumptions. The concern is not that people use tools such as ChatGPT, but that a fluent answer may be treated as an authority on a relationship. The report offers an example of a client who used ChatGPT to interpret a partner’s behavior as “trickle truthing.” The author says the concern about dishonesty could be legitimate, but the system had received only the client’s description, not the partner’s perspective or the couple’s interaction.
The report also describes a couple caught in a familiar conflict over appreciation. One partner wanted to feel valued; the other prepared appreciative statements and read them aloud. The effort did not create the feeling of being seen, and the first partner withdrew while the second became frustrated that her effort was not recognized. In the author’s account, having the right words is not the same as building connection: the couple needed to understand how each was experiencing the exchange, rather than treat appreciation as a completed task.
Another pattern involves uploading text exchanges and asking AI what a partner “really meant,” or who was wrong. The author notes that a brief reply such as “Fine” can carry different meanings depending on the speaker and circumstances. AI may offer possible readings, but the report says it cannot independently determine the sender’s intention. The supplied report ends mid-sentence while describing a therapist’s alternative to continued message analysis, so the specific suggestion that followed is not available.
Confidence Is Not Mutual Understanding
The report highlights a practical risk for people using AI during relationship conflict: an answer that sounds confident may reinforce one person’s interpretation before the other person has been heard. That can matter when someone is already worried about dishonesty, dismissal or manipulation. Their concern may be valid, but an AI response based on one account cannot, by itself, confirm what happened or why.
Scripted phrases can also fall short if they are delivered without attention to how they land. As the therapist describes it, a list of appreciative statements may meet the request literally while leaving the recipient feeling unseen. The larger issue is the difference between communication advice and the live exchange between two people: tone, silence, body language and each partner’s understanding of the moment.
This is a therapist’s clinical observation, not a study establishing how often AI use changes couples’ relationships or whether it improves or worsens outcomes. Still, the examples give readers a reason to treat AI-generated interpretations as possibilities to examine, rather than proof about a partner. The report also argues that clinicians who respond with judgment may discourage clients from discussing how they use AI at all.
Popular right now
Bose QuietComfort Headphones (2nd Gen), Wireless Over-Ear Noise-Cancelling Headphones, Black
Ninja CM401 Specialty Coffee Maker with 10-Cup Carafe and Fold-Away Frother
Sony WH-1000XM4C (M4 2nd Gen) Wireless Noise-Cancelling Headphones, Black
As an affiliate, we earn on qualifying purchases.
How the Therapist Sees AI Use
The report is written by a therapist affiliated with The Gottman Institute, which offers relationship education and describes Gottman Method interventions for couples. The author says they use ChatGPT personally to organize thoughts and consider other perspectives, while also stressing that therapists need to stay curious about the resources clients use outside sessions.
In the author’s account, a therapist can observe a couple interacting in real time: changes in tone, moments of withdrawal, missed attempts to connect and chances to repair a conversation. Those signals do not automatically reveal what a person is thinking; the therapist says they prompt questions. The distinction is between asking about an observed reaction and treating an outside tool’s account of a relationship as a complete explanation.
The examples in the report are presented as the author’s experiences with clients, not as findings from a quantified survey or controlled research study. The source material provides no sample size, dates for the client examples, or comparison with couples who do not use AI.
“They may leave an AI conversation feeling more certain without necessarily becoming more accurate.”
— The report’s author, a therapist writing for The Gottman Institute
What the Report Cannot Establish
The source presents a therapist’s observations and examples, not research measuring how common these patterns are or whether AI advice changes relationship outcomes. It does not identify the AI systems or prompts involved, nor provide independent accounts from the partners described. The author’s examples therefore cannot establish that AI caused the couples’ conflicts or that a tool’s interpretation was necessarily wrong.
In any specific dispute, the report does not establish what a partner intended, whether dishonesty occurred, or whether a message was manipulative. Those questions depend on details and perspectives the AI conversation may not include. The source also cuts off during its final example, leaving the author’s proposed alternative to text-message analysis incomplete.
Bring AI Interpretations Into Conversation
The report does not announce a new study, clinical rule or scheduled follow-up. Its practical direction is to keep AI use open for discussion in therapy, while examining what information shaped an answer and whether the conclusion has been mistaken for proof. A therapist may ask what was happening for each partner during an exchange and what each understood the other’s behavior to mean.
For readers, the distinction offered by the report is straightforward: AI can help generate questions or possible wording, but it cannot hear both sides unless they are provided, observe the couple together or independently verify intent. Whether a particular script or interpretation helps will depend on the couple and the situation. The report leaves broader questions about effectiveness and risks for research beyond the examples it describes.
Key Questions
What concern does the Gottman Institute report raise about AI and couples?
The author says AI can make a person feel more certain about a partner’s behavior without confirming that the interpretation is accurate. A system responding to one person’s account may lack the partner’s perspective and the context of their interaction.
Does the report say AI should not be used for relationship questions?
No. The author says they use ChatGPT and sees value in tools that help organize thoughts and consider perspectives. The concern is treating an AI-generated conclusion as proof or assuming a suggested script will create connection on its own.
Can AI determine what a partner meant in a text message?
The author says AI can suggest possible interpretations but cannot independently establish the sender’s intent. Short messages can mean different things depending on the person and circumstances.
Is this report based on a study of couples?
The supplied report describes a therapist’s observations and client examples. It gives no sample size or measured outcomes, so it does not establish how widespread the patterns are or whether AI use caused the relationship difficulties described.
Source: gottman.com
