When I announced I was writing a book to help victims of abusive exes use AI to communicate with their exes and strategize, someone commented that abusers would read it and use the information to get ahead of their victims.

Around that time, a client’s husband had started using AI himself, and it was clear that all he was doing was putting lipstick on a pig. Like abusers who start using therapy speak, it was all a veneer. If you give both an abuser and their victim a copy of my book, AI Armor, the results will be far from equal.

The reason why? People with abusive personalities (especially those with what Dr. Ramani Durvasula calls a “narcissistic personality style”) lack the self-reflection needed to make the deep changes to their thinking. They also tend to be rigid and controlling, and don’t have the flexibility to pivot or the humility to acknowledge to themselves and others when a course of action is wrong. All the volumes in the world can’t ram that into their narcissistic heads, never mind my 190-page book.

When people with a narcissistic personality style use AI, they use it to reflect back the way they already see themselves, not as a tool to explore where their own limiting or grandiose beliefs might be preventing them from being their best selves.

Recently, I was venting to Claude about a conflict I was having that involved multiple parties (someone on the other side, and people close to me who were pushing me to take actions that created the conflict). After a couple of back-and-forths, it suggested I should let go of the conflict. It asked me to consider other perspectives. Yes, my feelings were validated (like AI always wants to be your BFF) but I was getting a gentle pushback as well. I was being reactive, and Claude made that clear to me.

This kind of pushback is entirely lacking in a series of blog postings involving chats with the very same AI tool, Claude, by “parental alienation” expert Craig Childress. First, some background on who he is.

It would be an understatement to say that Childress hasn’t coped very well with losing his license. In fact, he’s gone off the rails, with all the hallmarks of a classic case of narcissistic decompensation. It’s a bit ironic, considering that he diagnoses parents he’s never met as narcissists, based on their children’s behavior.

What Childress did in the blog posts I mentioned was to feed Claude a prompt containing a particular framework (e.g., the institutional betrayal he thinks he experienced = family courts betraying parents and children). And Claude did what Claude does. It applied the framework rigorously, mapped every element, and found a “comprehensive fit.”

The whole series of blog posts is replete with this, but here’s a screenshot of one part. Claude tells Childress what he wants to hear. That a child’s protective responses being pathologized as “alienation” is the same as what he thinks was done to him. That the use of “parental alienation” as a counter-allegation to suppress reports of abuse is directly equivalent to licensing board complaints as a disciplinary mechanism.

The irony goes up to 11 here, because Childress himself has played a leading role in the first type of incident in both these examples. And it’s clear that the knob on Childress’ ability to self-reflect is firmly stuck at zero. The end result is academic-grade DARVO.

Claude basically mirrored back Childress’ own persecution narrative to him. It didn’t give him any pushback or point out the possible flaws in his narrative. It told him what he wanted to hear because that’s the expectation he fed it. It’s probably what he’s been doing with other people for years.

My approach was completely different. At one point during my chat with Claude, I asked, “Am I being too pushy?” That was an invitation to Claude to see that the other parties in my disagreement had coherent narratives of their own.

Also, I’ve trained Claude a certain way. In previous chats, I’ve explicitly questioned it when it’s agreeing with me too readily, or when its framing is off. I constantly prompt it to provide me with counter-arguments and to point out where my framing is flawed.

When your ex starts using AI, it feels scary. You think, AI makes my abuser sound so credible. But, like Childress, your abuser doesn’t feed AI the whole story. They don’t invite any pushback. They’re just using AI as another flying monkey. And, like Childress, they’re going to get a subpar result. It doesn’t show the flaws in their reasoning or the evidence that they’re leaving out.

You worry that your ex’s AI output is going to make their case stronger. I’m here to tell you that it is not going to do this. It does the opposite. It hands you a map of their narrative. It points out clearly what they’re afraid of, what they’re avoiding, the delusions they operate under, and (thanks to their absence) the realities they can’t afford to let a court examine too closely. You know how to read that map. You’ve been living inside their narrative for years.

First, stop being intimidated by the polish. Polish doesn’t equate to accuracy. AI often produces factual errors. Your ex, who’s likely too entitled to check the output and too lacking in self-reflection to read it critically, probably didn’t catch these.

And that’s where you come in. You don’t need to produce a ten-page rebuttal. People are getting jaded by the AI-speak they see in every second Facebook post, and reading the kind of AI slop your ex produces just makes the judge’s eyes glaze over. One page of precise, documented facts beats their AI manifesto every time.

Watch for the tells, especially the detailed, polished messages that don’t align with reality. Example: Their AI-generated email waxes lyrical about “flexible, child-centered scheduling” but they’ve been woefully late for three pickups in a row and caused your child to miss important appointments. Screenshot all of this. Document the gap between what the AI helped them say and what they actually do. This documentation will be extremely valuable going forward.

You can deploy AI in a way that they can’t, so make the most of this. Use it to prepare, not to perform or intimidate. Feed it your ex’s communications and ask what’s missing, where the logic breaks down, what a court would actually need to see. Do the same with your own.

Ask it to pull you back to your long game. What do you want your children to remember about how you handled this? What does the version of you who stayed focused on them and their needs actually look like? Feed that into every strategy conversation, so you’re making decisions from there and not from your reaction to whatever your ex did last Tuesday.

Ask it where your argument is weak, what the strongest counter-argument is, whether you’re being reactive. Your ex will never do this. The whole point of their AI use is to avoid that kind of friction. And keep your actual communications human. Even if you get AI to generate them, read through them and think, does this sound like me?

And when you read their AI-generated communications, remember that you’re actually looking at a closed loop. It’s their narrative, confirmed by a machine that had no access to your evidence, your children’s actual experience, or any voice but your ex’s. You have all of that. Plus your own ability to self-reflect. Your ex doesn’t. All the technology in the world won’t close that gap.

I love writing this newsletter every week but there’s something I love even more, and that’s talking to my readers and hearing where they are on their journey. So, please, feel free to respond to me. It warms my heart when I get a response in my inbox because I know that I’ve helped someone feel heard. If you want to get stuff off your chest, you’re welcome to book a free, 30-minute call with me. Put in the comments why you’re calling me. Sometimes it’s just good to get things off your chest with someone who gets it. I don’t believe in hard sells and I don’t push my coaching on anyone who doesn’t want or can’t afford it.

Want to know more about what I do?

If you want to get ahead of your ex's AI use instead of scrambling to catch up, my book AI Armor gives you the practical framework to do exactly that.

It shows you how to use AI to find the humor in your ex’s antics, document patterns, stress-test your strategy, shift your mindset, and keep your communications court-ready without sounding like a robot. It's 190 pages, not a textbook, and it was written for exactly the situation you're in. If you’re not convinced, head over to the Amazon Canada page or the Amazon US one and check out the reviews.

If you’re looking for an AI tool that’s affordable, gets coercive control and can help you with timelines, is confidential, and has a very responsive support team, look no further than Aimee Says. You can try their premium version (which costs USD 20 per month) for free for 7 days.

And if you really want to up your prompting game, then check out the courses on offer at learnprompting.org. They have a free course called “ChatGPT For Everyone” (you can use what you learn here with other AI tools).