How to Describe a Dream to AI: A Prompt Template That Works

Why the description decides the answer

Most disappointing readings fail before the model speaks: “I dreamed about water, what does it mean” has lost the place, the mood and the ending, so the answer can only be a dictionary entry.

The description preserves the dream and tells the AI which details matter. Do both and the reading becomes something you can argue with.

The eight fields of the template

You do not need a long prompt. You need eight facts, in the order a stranger would need them.

• Scene. Where you were, whether the place was familiar, the time of day.
• Sequence. What happened first, next and last, in three to six beats.
• People. Who appeared and how you knew them; a role if the name is gone.
• Emotion. What you felt inside the dream, not what you should have felt.
• Last image. The final frame before waking. Endings carry unusual weight.
• Waking feeling. The mood of the first minute awake.
• Real-life context. Two or three sentences about this week: work, a decision, a loss.
• Your question. One specific question, such as why the flooded kitchen keeps appearing.

Facts first, meaning last. Put your interpretation at the top and the model will agree with it and stop looking.

A filled example

Finished, it looks like this.

• Scene: my grandmother's kitchen, flooded to ankle height. Lights on, stove cold.
• Sequence: I walked in, saw the water, looked for the tap, then noticed the table set for six.
• People: my grandmother, dead four years, at the sink with her back to me; a cousin I argued with last year.
• Emotion: calm at first, then a quiet shame, as if I had arrived late.
• Last image: water rising over the table edge while nobody moved.
• Waking feeling: heavy and slightly guilty; I checked my phone immediately.
• Context: I am selling her flat this month and avoiding that cousin's calls.
• Question: why does the dream put shame in a kitchen where I always felt welcome?

No symbol dictionary, no “this probably means”, no request for a prediction.

The same night, described badly

“I dreamed about my grandmother and water and a big table, what does it mean?” Three details survive and twenty disappear. The model cannot know whether the water was warm, whether you were afraid, or why the dream arrived this week. It will produce a warm paragraph about family: the safest guess available.

Six mistakes that flatten a dream

• Describing the interpretation, not the dream. “I am anxious about money” is a conclusion; a till that will not close is evidence.
• Mixing several nights. Two dreams in one prompt become one blended story that never happened.
• Being too vague. “A dark place” gives it nothing to hold; one odd detail beats three adjectives.
• Writing a novel. A thousand-word prompt invites a summary, not a reading.
• Hiding the emotion. A report with no feeling reads like a police statement and gets treated like one.
• Asking for a verdict. “Should I quit my job” turns an interpreter into an oracle. Ask what the dream emphasises, then decide yourself.

Privacy, language, and what to leave out

Remove names, addresses and anything identifying a third person who did not consent. Roles work better: “my manager” says more than a name. Prefer a tool that encrypts your text and lets you delete it, as described in how dreamOut works.

Write in the language you dream in, or the closest one, then ask for the answer in that language. Symbols do not translate cleanly: a word meaning both “debt” and “fault” loses its double edge in English. For a religious or cultural reading, name the tradition and ask the model to separate source, inference and guess.

The three-pass loop

Frequently Asked Questions