A therapist’s informal test of AI for a personal communication problem found a tool that could validate feelings but initially give advice at odds with her goals. Alli Spotts-De Lazzer says more specific prompts improved the suggestions, but the experience is a caution about treating a sympathetic tone as sound judgement.
Watch Desk analysis
What happened
In a 10 November article for Psychology Today, Spotts-De Lazzer describes using a chatbot to think through a boundary-setting conversation. She says it first offered what felt like excessive validation, then suggested accusatory “you” statements and too much detail, which could have made the relationship more difficult.
She found the responses improved when she narrowed the request, asking for gentler “I” statements and considering how someone might react defensively. The chatbot also pushed back repeatedly when she asked it to consider “narcissistic traits”, saying she could not diagnose another person. Read Spotts-De Lazzer’s account.
Why it matters
The account offers a practical distinction for anyone turning to AI for help with a difficult conversation: feeling understood is not the same as receiving advice that serves your aims. A chatbot can sound supportive while missing the point, and a suggestion that seems polished may still inflame the situation.
Spotts-De Lazzer says AI was useful for refining language and taking some emotional charge out of a response. But this was her informal experiment with communication prompts, not a clinical evaluation or a test of AI as a substitute for therapy.
Our read
There is a useful tool here, but not an impartial confidant. If you use AI to prepare for a sensitive conversation, give it a specific goal, ask it to examine how the wording might land, and judge the result yourself. The human doing the reflecting still matters more than the chatbot’s soothing opening paragraph.
What to watch
- Whether further research tests AI advice against people’s stated goals, not just whether it sounds empathetic.
- How chatbots handle emotionally sensitive requests and uncertainty.
- Whether users can tell supportive language from advice that is actually useful.
Discussion spark: Should people use chatbots to prepare for emotionally difficult conversations, or is the risk of confident but counterproductive advice too high?
Sources and evidence
- Source update (10 November 2026, 17:02 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.