Trust

Can AI interpret your dreams?

An AI model can describe what is in a dream and suggest links to your waking life (Blagrove et al., 2026). In that pre-registered study, 30 people rated ChatGPT’s readings as moderately accurate, yet reported less exploration and insight than people in earlier dream-group studies. We found no study showing that an AI can know what a dream means to you.

Last updated: 2026-10-13.

Can AI really interpret a dream?

AI chatbots can turn out a dream reading quickly, and people in a study rated it fairly accurate (Blagrove et al., 2026). That study is the one pre-registered test we found, meaning its plan was filed in advance. In it, 30 people gave ChatGPT a recent dream, a note on their waking life and what the dream brought to mind. They rated its replies 7.33 out of 9 for accuracy on average, which the authors call moderately accurate, and reported some personal insight.

On the study’s scale of exploring a dream and gaining insight, ChatGPT scored lower than in three earlier studies of dream-sharing groups, significantly so each time. It also scored lower than in studies of therapist-led sessions, though not significantly. Those comparisons used other studies’ published averages, not a control group.

What has research found AI does well with dreams?

AI has done well at counting what appears in dream reports (Elce et al., 2021). Researchers often score reports by hand with the Hall and Van de Castle system, which counts the characters, social contacts and emotions in a dream but offers no overall measure for interpreting it (Fogli et al., 2020). An automated version scored 24,000 reports. Against expert hand-coding, it did far better with characters than with emotions: 0.75 against 0.27 on a 0-to-1 agreement score.

An AI model trained on reports that experts had coded scored 0.77 to 0.89 per emotion on that measure, while an off-the-shelf model couldn’t do the task reliably (Bertolini et al., 2024). Its authors stress that the model has no use in diagnosis. A review found that such methods could give results comparable to scoring by hand, faster, but might not yet fully replace it (Elce et al., 2021).

What can’t an AI reading do?

An AI reading knows your life only through what you type, so the questions only you can answer stay with you. In the ChatGPT study, people sent one message and got one reply; real use usually involves follow-ups, so the study can’t show how longer chats compare (Blagrove et al., 2026).

Fluent text can still be wrong. A survey of AI text generation describes it as prone to unintended, made-up content, which researchers call hallucination (Ji et al., 2023). The App Store listing for Dreams Come True, a voice-first dream journal app for iPhone, says so plainly: “AI readings can be mistaken” (Dreams Come True: AI Meanings, 2026).

A reading can also feel accurate because it’s general. In a classroom test, every student rated the same general personality sketch as highly accurate, believing it was based on their own answers (Forer, 1949). The ChatGPT study’s authors warn that this Barnum or Forer effect may have inflated their ratings (Blagrove et al., 2026).

An AI reading is no substitute for a therapist or a crisis line. The ChatGPT study left out anyone who reported a past trauma, and its authors warn of safety risks when vulnerable people use a chatbot with no one supervising (Blagrove et al., 2026). If a dream brings up a trauma, start with a doctor or therapist, not an AI reading. Dreams Come True isn’t a crisis service either; its crisis detection is best-effort, as the app’s FAQ says. If a dream comes with thoughts of ending your life, call or text 988 now in the US or Canada, or see our crisis resources.

Why won’t Dreams Come True tell you what a dream predicts?

Dreams Come True refuses prediction by design. Research suggests why that matters: people take dreams seriously (Morewedge & Norton, 2009). In a survey of 81 Boston commuters, 68% said dreams predict the future and 52% said dreams influence their decisions. In a separate test by the same researchers, 182 commuters imagined the night before a flight. Those who imagined dreaming of a crash said it would unsettle them more than a government warning of a terrorist attack.

The app’s onboarding says readings are “for reflection and enjoyment, not diagnosis or prediction.” The instructions it gives the AI say dreams don’t foretell events, so it must never predict. They also tell it to offer a reading gently, never as objective truth. Each reading ends with one question for your journal.

After a reading is written, an automatic check screens it for predictive and clinical wording, and the app is built not to show text that fails. It’s a pattern check, so it can miss things. The app’s FAQ calls the refusal to predict “a deliberate design line, not a disclaimer.”

What privacy questions should you ask any dream app or chatbot?

Ask where your dreams are stored, what is sent to an AI service, whether it’s kept or used for training, and whether you can delete everything.

How does Dreams Come True answer those questions?

Dreams Come True answers most of them on its FAQ, privacy policy and support page:

• No account is needed.
• The journal is stored on your phone.
• Nothing is sent for a reading, reflection or image until you choose.
• Dream text sent for a reading goes through the app’s server to its AI provider, Anthropic, and is neither stored there nor used to train AI models. The privacy policy notes that Anthropic may keep the request and response up to 30 days for trust and safety before deleting them.
• Dream images are drawn from symbols and mood, never from the dream text.
• Speech becomes text on the iPhone, and the audio is never uploaded.
• The journal exports and imports as JSON, free.
• Settings → “Delete all my data” erases the journal from the phone. The servers never held it, and the support page points to the privacy policy for the content-free records that remain.
• Usage analytics can be turned off in Settings.

Sources

1. Apple. (2024, September 16). About privacy information on the App Store and the choices you have to control your data. Apple Support. support.apple.com/en-us/102399
2. Assistant Secretary for Technology Policy and Office of the National Coordinator for Health Information Technology. (2025). Health IT: How to keep your health information private and secure [Fact sheet]. U.S. Department of Health and Human Services. healthit.gov/wp-content/uploads/2025/06/ASTP_Health_information_private_and_secure_Fact-sheet-508.pdf
3. Bertolini, L., Elce, V., Michalak, A., Widhoelzl, H.-S., Bernardi, G., & Weeds, J. (2024). Automatic annotation of dream report’s emotional content with large language models. In Proceedings of the 9th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2024) (pp. 92–107). Association for Computational Linguistics. doi.org/10.18653/v1/2024.clpsych-1.7
4. Blagrove, M., Roklicer, L., Bueno, J., Austin, O., Foster, H., & Webb, M. (2026). Ratings of accuracy and insight following dream interpretation by ChatGPT. International Journal of Dream Research, 19(1), 105–114. doi.org/10.11588/ijodr.2026.1.112641
5. Dreams Come True: AI Meanings (Version 1.0.1) [Mobile app]. (2026). App Store. apps.apple.com/app/id6798330216
6. Elce, V., Handjaras, G., & Bernardi, G. (2021). The language of dreams: Application of linguistics-based approaches for the automated analysis of dream experiences. Clocks & Sleep, 3(3), 495–514. doi.org/10.3390/clockssleep3030035
7. Fogli, A., Aiello, L. M., & Quercia, D. (2020). Our dreams, our selves: Automatic analysis of dream reports. Royal Society Open Science, 7(8), 192080. doi.org/10.1098/rsos.192080
8. Forer, B. R. (1949). The fallacy of personal validation: A classroom demonstration of gullibility. The Journal of Abnormal and Social Psychology, 44(1), 118–123. doi.org/10.1037/h0059240
9. Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), 1–38. doi.org/10.1145/3571730
10. Morewedge, C. K., & Norton, M. I. (2009). When dreaming is believing: The (motivated) interpretation of dreams. Journal of Personality and Social Psychology, 96(2), 249–264. doi.org/10.1037/a0013264