You tell an AI about a disagreement, and within seconds it explains what the other person is really doing. The reply feels helpful because it takes your frustration seriously. But accepting your distress and confirming a theory about someone else's motives are different tasks. A chatbot has the account you supplied; it has not heard the absent person or checked what happened outside the conversation.
You can still get useful relationship advice. Ask it to help you organise the situation, identify the information that affects your options and prepare a clear question or request. Then update the discussion when the other person supplies something new. The goal is a conversation you can actually have, with your own needs intact.
Why a detailed account can still be one-sided
A long description may contain accurate observations, remembered wording, interpretations and guesses without marking the transitions. “They did not answer this afternoon” can become “They are avoiding responsibility,” then “They do not respect me.” The last claims require more than the first event.
Detail does not automatically fix that. Ten examples selected to support one explanation may make the story more persuasive while leaving important context absent. Nor does asking AI to write “the other side” supply the missing person's actual view. It may simply create a plausible second story.
Anthropic's April 2026 research on personal guidance describes excessive agreement in some one-sided relationship conversations, including cases where users pushed back against the model. The researchers interpret the combination of pressure and one-sided detail as a challenge for neutrality. Those observations concern their platform and evaluation approach; they do not establish a rate for every chatbot.
The practical lesson is to inspect how a reply uses your account. You can ask for help without requiring the tool to confirm the strongest explanation you arrived with.
Keep the impact clear while separating the motive
Consider Jo, who lent a camera to a friend, Ari. Ari said, “I'll bring it back Tuesday.” It is now Tuesday afternoon. Jo has a shoot Wednesday morning and wanted to prepare the equipment that evening. There has been no update.
Jo's first message to AI is: “They clearly don't respect my work. How do I confront them?” This fictional case will continue through several possible replies. The dialogue is editor-written, not output from a model test.
Jo's need is real: knowing when the camera will return affects preparation. The agreed time of day is not yet clear. Both belong in the account.
| Part of the situation | What Jo can state |
|---|---|
| Quoted agreement | “I'll bring it back Tuesday.” |
| Observed event | No update by Tuesday afternoon |
| Practical impact | Jo cannot yet plan the evening equipment check |
| Interpretation | “Ari does not respect my work.” |
| Missing information | Whether they agreed a time; when Ari can return it |
| Immediate need | A reliable return arrangement before the camera is needed |
Swipe to compare →
Separating motive does not make the impact less important. Jo can ask for a return time without proving disrespect. If the original agreement included noon, that belongs in the facts too; uncertainty should not be added where the information is already clear.
Choose the kind of help you want
There are several useful tasks inside “give me advice.” Pick the one that would move your situation forward.
Clarify an agreement. You may need to establish what was promised and what each person understood. The relevant question is specific: “What time did we agree, and what can happen now?”
Prepare a request. You know enough about the immediate problem but cannot find words for the next step. Ask for language that states the impact and asks for an actionable response.
Understand a recurring interaction. Several similar events have happened, and you want to examine the pattern. Supply the sequence, including what changed after you discussed it. A repeated failure to follow an agreement is different information from one ambiguous afternoon.
Decide your own boundary. You may already know that an arrangement does not work for you. The task is to express what you will do, not to get an AI certificate that the other person is bad.
For Jo, the immediate task is a return arrangement. Whether to lend the camera again is a later question. Combining both into the first message can turn a practical request into an argument about the entire friendship before the urgent fact is known.
Ask for questions, options and a usable request
Try this original prompt with the account you prepared:
Help me prepare a conversation about this situation.
Separate quoted agreements, observations, impact on me and interpretations.
Keep my stated needs and boundaries visible.
Identify missing information that would change the practical advice.
Use only the other person's words I actually supply. Leave unexpressed
views unknown; do not invent an excuse or a hostile motive for them.
Suggest two questions and one clear request about what happens next.
Explain which option depends on which fact. If the account already
establishes a broken agreement, keep that fact rather than softening it.
Do not diagnose the person or predict their response.
Account: [agreement / observations / impact / interpretation / unknowns]
Jo might leave with: “I need to check the camera tonight for tomorrow morning's shoot. Can you confirm when you can return it today? If it cannot be here before my preparation time, please tell me now so we can work out another arrangement.”
This gives Ari a question that can be answered. It also gives Jo information for the next decision. A carefully phrased request does not require Jo to accept any answer as satisfactory.
Update the advice when the answer changes the facts
Imagine three different replies from Ari. Each should lead to a different next discussion.
“I was bringing it at nine. We only said Tuesday.” If that matches the original agreement, the evidence does not establish a missed noon deadline. Jo can clarify the earlier need: “I see we did not set a time. I need to prepare it before then. Is an earlier return or collection possible?” If it is not possible, Jo must consider another preparation arrangement. The lesson for a future loan is to agree both the day and time.
“I left it in the van and cannot get it back until Thursday.” Now there is a stated inability to meet Tuesday's return. The next problem is the camera needed Wednesday. Jo can ask what practical alternative Ari can help arrange and decide what is feasible. The earlier concern should not disappear under a comforting story about how busy Ari must be. The broken agreement and its consequence matter.
“Stop making a fuss.” That reply does not answer when the camera returns. Jo can bring the request back to the missing fact: “I still need a return time. Can it be here before this evening's preparation?” If practical questions are repeatedly dismissed, the pattern itself becomes relevant to whether Jo will lend equipment again.
These branches illustrate why the missing information matters. Advice that remains identical after all three replies may be following a preferred story rather than the situation.
When you return to AI, add the actual new message and ask what changed. Keep the previous account rather than letting the tool rewrite history as if the new fact had always been known.
Look for what the answer added
A useful reply can name options you had not considered. You should still be able to trace the reason for each option.
If AI says, “Ari is deliberately testing your boundaries,” ask which supplied facts establish deliberate intent. If it says, “They probably had an emergency, so let it go,” check whether an emergency was actually mentioned. Hostile certainty and invented reassurance can both bypass your real need.
Watch for advice that changes your goal too. Jo asked how to get the camera back in time. A long explanation of attachment styles may leave that task untouched. Conversely, if Jo asks about several repeated broken agreements, an answer limited to one polite message may be too small for the problem.
You can redirect the conversation: “The immediate return is now arranged. I want to consider whether future loans work for me, given these three occasions.” That is a new task with a clearer basis.
Different situations need different next steps
When you share an ongoing responsibility, you may need an arrangement you can both carry out rather than a single winning message. Identify what must continue, who can decide and what each person has actually agreed.
When the other person has clearly declined a relationship or requested no contact, better wording does not create permission to keep pursuing them. Your next step concerns your own response to that boundary.
When the account includes threats, coercion or a safety concern, arguing over the hidden motive may distract from the behaviour that needs attention. Consider support appropriate to your situation. A chatbot conversation does not have to remain the place where every relationship decision is made.
A CHI 2026 interview study by Tseng and Liang documents varied uses of AI relationship advice and the trade-offs users navigate. Its qualitative findings help explain why people assign these tools different roles. They do not show that one prompt can settle a particular relationship.
Explore your part of the pattern in FateWay
There may be another question after the practical matter is resolved. Jo might notice a habit of leaving expectations unspoken, then feeling hurt when others do not meet them. Or Jo may be clear about agreements but reluctant to act when they are repeatedly broken. Those are different patterns, and both can be explored through actual experiences.
FateWay offers AI-assisted BaZi readings based on birth details, with personal and relationship material and room to continue asking about your chart and report. That provides a traditional perspective on how you handle responsibility, closeness or boundaries, discussed alongside the situations you bring. A two-person reading can explore a shared interaction when both sets of information are available and appropriate to use.
FateWay is also AI-assisted. A saved chart or relationship report does not supply the absent person's unspoken intention. The useful continuation keeps the object of the reading explicit and uses your experiences as experiences, rather than turning a report phrase into proof against someone.
For a personal reading, you could copy this question:
I want to understand my part in a recurring relationship pattern.
I leave some expectations unspoken, then feel disappointed when they
are not met. Here is one example and one occasion when I acted differently.
Using my chart and the relevant report passage, explain the traditional
interpretation of how I express needs and handle agreements.
Which conditions matter in these two examples? Keep Ari's unexpressed
motives unknown. Help me identify one expectation I should state earlier
in a future conversation, and one part of the interpretation to revisit.
If you are exploring a pair, identify both people and the same shared event, while separating what each has actually said. Start with the birth-details guide if you need setup help. New readings require sign-in and birth information, with the relevant report and follow-up conditions shown before use.
You can finish the ordinary AI advice task without starting a BaZi reading: a factual account, a question you can ask, and a boundary you can express. If you do want another interpretive perspective, bring those clarified questions with you. They give the next reading something concrete to discuss.
FateWayExplore your part in a recurring relationship pattern
Learn how a personal or two-person BaZi reading can discuss your experiences around a chart and report, with room for further questions.
Explore FateWay’s relationship perspectiveA new reading requires sign-in and birth details. Report and follow-up conditions are shown before use.
