what is gaslighting?
gaslighting is a pattern where things that happened get denied until you start doubting your own record of them. in a chat it rarely looks dramatic. it looks like a firm "i never said that", a calm "you're imagining it", a patient "you're overreacting", repeated until your version feels like the unreliable one.
it's a heavy word, so it carries a heavy standard. one disputed memory is not the pattern. the pattern is denial running in one direction, over time, against what's actually written in the thread.
what the pattern looks like in a chat
- denial of documented events. "i never said that" about a message that is still in the thread, with a timestamp. the record and the account of the record no longer match.
- downplaying as a pattern. "it's not that deep." "you're overreacting." once is a mood. every single time you raise something is a shape.
- the apology flows the wrong way. you raise a problem, and three messages later you're the one saying sorry. the catalogue calls that flipping it on you, and it travels with this pattern.
- your own messages change. over months, your side starts pre-hedging. "maybe i'm misremembering, but." that drift is evidence too.
thursday 19:02, them: "go, honestly, it's fine" · on file
what it isn't
this hedge matters more here than anywhere else on the site:
- two people genuinely remembering differently is human. it happens constantly, in good relationships, about who said what in a kitchen in 2024. ordinary misremembering is not this pattern.
- one heated denial in one argument is one argument. patterns need repetition, and this one needs it more than most.
- the direction is the test. honest confusion scatters both ways: you misremember some things, they misremember others. this pattern runs one way only, and it runs against the written record.
the boring explanation is often true. that's not a disappointment, it's a result.
how to check it against the written record
this is the one place text genuinely outperforms memory. rewritten history has an original on file:
- find the argument that keeps recurring. does one side's account of it move between rounds?
- take the flat denials and check them against the thread. "i never said that" is a checkable claim. check it.
- read your own side from three months ago and from this week. who is apologising now, and what for?
textRay does this reading for you. screenshots, a WhatsApp export or pasted text go in, the scan reads both sides, and every finding cites its messages with timestamps. if the pattern is supported, it's stated calmly: "gaslighting — 82% pattern match", quotes attached, counterevidence one tap away. if what the receipts actually show is two people misremembering in both directions, you get that instead. either way, you stop arguing with your own memory.
the original is still in the thread.
download on the App StoreiOS · 16+ · no subscription · credits from £2.99
The plain version.
Findings describe patterns in a conversation. They are never a diagnosis of a person, and textRay never claims more than the messages show. If a conversation involves threats, pressure, or control of money or movement, that is not a pattern to debate. The app drops the editorial layer, uses plain language, and puts support options first. You consent before any AI processing, and reports stay in your history until you delete them.