Gaslighting Check

Gaslighting Check is an AI tool that turns “is this conversation all in my head?” into a readable document: paste a conversation, upload a screenshot or a recording, and it flags possible manipulation patterns, with verbatim quotes and suggested next steps. It supports English only and is built around four settings: partners, family, the workplace, and no-contact after a breakup.

Core capabilities

• Pattern Mapper: Breaks a conversation into named tactics — denial, blame shifting, DARVO, pseudo-apology, financial control and others — with every item tied to specific words in the original text.
• Reality Separation: Three editable columns that set “what happened”, “what I felt” and “what I read into it” side by side, so the facts are sorted out before any judgment.
• Reply Drafter: Offers three calm, restrained drafts that follow the JADE principle of not justifying, not arguing and not defending, and it lets you choose not to reply at all.
• Evidence Builder: Compiles analyses and incident logs chronologically into one document stripped of labels and emotional wording, and clearly marks observation as observation rather than diagnosis.
• Journal and Grounding: Ten guided templates with AI reflections, plus a completely free sixty-second breathing exercise, so you steady yourself before you answer.

Who it is for It suits English-speaking users who “cannot quite put it into words but keep feeling something is off”: people logging workplace manipulation or post-separation contact over time, and organizing a timeline to hand to a lawyer or HR. It does not suit anyone in an immediate crisis, anyone who needs evidence with legal force, or anyone whose device may be searched by a partner.

Standout strengths Most tools in this category stop at reading the exchange and producing a report; this one links identification into a chain — reading, emotional grounding, drafting a reply, keeping a record — in an order that matches the rhythm of trauma intervention. Its twenty-seven-tactic vocabulary cites an academic source for every item and keeps a fallback category for “I cannot tell which one this is”, so the model is not forced to pin a label on it. Quick exit and the GC TECH billing disguise are also rare.

Editorial assessment The product itself is top tier, but the credibility of its documentation is a real weakness: the blog still claims end-to-end encryption, deletion within twenty-four hours and a zero-knowledge vault, which directly conflicts with the privacy policy (processed through OpenAI, with separate retention of seven days, thirty days and indefinitely); the promised purely local analysis does not exist architecturally; pre-signup drafts sit in the browser with no expiry and in clear text, and audio is written to disk as base64; the homepage carries no crisis hotline; “Court-ready” in practice is browser printing, while the tool page itself says court-aware. External validation is close to nonexistent: zero Trustpilot reviews, and the only score, three point eight out of five, rests on six reviews. Fixed wording would earn four stars; before that it should not be recommended to people in high-risk situations.