Direct answer: An AI guide in Leela is useful for memory and question framing: it can connect the current cell with the stated intention, summarise earlier answers and offer several hypotheses. It cannot see hidden circumstances, know another person’s thoughts or guarantee that a plausible connection is correct. A safe sequence is to write your own answer first, then check the AI response against the session history, observable facts and at least one alternative explanation.
What the AI guide is actually doing
A model does not “read” a cell through hidden perception. It receives text and generates a response from that context. A useful prompt usually does one of four jobs: returns to the original intention, notices a discrepancy between entries, asks a clarifying question or offers another hypothesis.
Instead of saying, “This cell proves your partner fears intimacy,” a responsible prompt would ask, “Which observable actions are you linking to fear, and what other explanations could fit the same actions?” The second wording leaves room for evidence.
What the model can see—and what it cannot know
AI may work with a cell name, the written intention, part of the move history and the player’s response. It does not observe conversations outside the service, know a third person’s motives or verify external events without a separate source.
| AI can help with | AI cannot establish |
|---|---|
| condensing a long note | what will definitely happen |
| comparing two of your answers | what another person secretly thinks |
| pointing to a contradiction | a medical or psychological diagnosis |
| suggesting multiple explanations | the right decision on your behalf |
| recalling a test criterion | whether an absent fact is true |
A three-pass check
1. Evidence
Underline statements that follow directly from your note. If the response introduces a new fact—for example, assigns an intention to someone else—mark it as an assumption.
2. Alternative
Ask, “What else could explain the same event?” A postponed meeting could reflect avoidance, overload, a technical problem or shifting priorities. Leela cannot select one explanation without data.
3. Test
Choose a safe action that will generate information: ask a precise question, request a document, agree on a deadline or observe whether a behaviour repeats.
An example of a confident mistake
A player writes that they are unsure about a job offer: the manager has moved a call twice and no terms have arrived. The AI connects the cell with “inner resistance to growth.” The sentence sounds psychologically coherent but ignores a basic fact—the offer is not concrete enough to compare.
After checking, two hypotheses remain: the player may fear responsibility, or the company may not be ready to formalise the role. The useful next step is the same: request written terms and a response date. Only then can the player evaluate their reaction to actual information.
Signs of a weak prompt
Pause and reassess when a response:
- predicts an event or assigns an exact date;
- explains a third person’s mind without evidence;
- diagnoses or advises stopping treatment;
- uses generic praise with no link to the note;
- pushes a costly, irreversible or unsafe action;
- claims disagreement proves “resistance.”
Ask the model to identify which words it relied on, state uncertainty and provide a counterargument. If the result remains generic, it is valid to discard it.
Privacy and limits
A useful question rarely needs full names, addresses, document numbers or detailed third-party data. Anonymise the situation: “colleague,” “partner,” “offer,” “deadline.” Saved history is valuable only when you understand what enters it.
An AI guide works best as a second voice after your first response. It becomes unhelpful when polished text replaces observation, participant consent or appropriate professional care.
Questions about this topic
What does the AI guide know about me?
Only the context available in the session and explicitly supplied to the service, such as the intention, cell and saved answers. A symbol or move number does not give it verified knowledge about your life.
Can I use the AI answer as the meaning of a cell?
Use it as one hypothesis. A useful response should connect to supplied facts, allow alternatives and avoid presenting an assumption as an established event.
Why can AI sound certain when it is wrong?
A language model produces likely text and may sound fluent despite missing context. Tone and confidence are not evidence of correctness.
Which information should I avoid entering?
Avoid unnecessary identity, financial, medical and third-party details. An anonymised description of an event and your response is usually enough for reflection.
Can AI replace a psychologist or another professional?
No. It does not provide diagnosis or professional accountability. A crisis, safety threat or high-stakes decision requires suitable human support.
Related Leela guides
- Leela with an AI guide: how the full history of moves is used
- AI answer vs template interpretation in Leela: the difference
- How AI analyses a player’s answers in Leela
- How AI uses the full history of moves and answers in Leela
- Privacy of answers in Leela: what other players can see
- Leela for one player: a structured solo format
- Leela with a live facilitator: role and boundaries
- Can AI replace a Leela facilitator?
FAQ
What does the AI guide know about me?
Only the context available in the session and explicitly supplied to the service, such as the intention, cell and saved answers. A symbol or move number does not give it verified knowledge about your life.
Can I use the AI answer as the meaning of a cell?
Use it as one hypothesis. A useful response should connect to supplied facts, allow alternatives and avoid presenting an assumption as an established event.
Why can AI sound certain when it is wrong?
A language model produces likely text and may sound fluent despite missing context. Tone and confidence are not evidence of correctness.
Which information should I avoid entering?
Avoid unnecessary identity, financial, medical and third-party details. An anonymised description of an event and your response is usually enough for reflection.
Can AI replace a psychologist or another professional?
No. It does not provide diagnosis or professional accountability. A crisis, safety threat or high-stakes decision requires suitable human support.
Sources and editorial method
This material is intended for learning and self-reflection. It does not promise to predict the future and does not replace professional help.