A new form of coercive control: using AI to gaslight a partner
In my clinical work, I recently came across something I had not encountered before.
A client described how their partner had been taking isolated messages, often written in moments of desperation after many hours or days of abuse, and pasting them into an AI tool for a “professional opinion”.
The AI’s response was then fed back to the client as though it were objective proof:
“See? Even AI says you’re crazy.”
“See? Even AI says you’re the problem.”
I want to name this carefully.
I am not saying AI is the abuser.
I am saying that, in abusive relationships, AI can be weaponised as part of coercive control.
What struck me most was the pseudo-authority this creates.
The abusive partner is no longer just saying:
“I think you’re irrational.”
They are now saying:
“A neutral, intelligent system has reviewed what you said and confirmed that you are irrational.”
But of course, that is a distortion.
The AI has not witnessed the relationship.
It has not seen the hours, days, or years of coercion, intimidation, provocation, fear, erosion, or despair.
It has not assessed the pattern.
It has only been shown a decontextualised fragment selected by the controlling partner.
That is not truth.
It is context-stripping dressed up as objectivity.
This matters because coercive control so often works by destabilising a person’s trust in their own mind, feelings, and perceptions.
If a distressed response, produced under pressure, is lifted out of context and fed into an AI system, then returned as “proof” that the victim is unstable, the gaslighting can acquire a new layer of apparent legitimacy.
I suspect clinicians may begin to encounter more of this.
Not because AI creates abusive dynamics.
But because abusive dynamics tend to recruit whatever tools help the abuser dominate the narrative, undermine the other person’s self-trust, and present themselves as the reasonable one.
So I think this is something we need to become more alert to:
AI-mediated gaslighting as a possible feature of coercive control.
Not sensationally.
Not as a moral panic.
But calmly, clinically, and with our eyes open.
Sometimes new technologies do not create new human problems so much as give old human cruelties a new surface.