I Tested AI on OCD. Here Is My Exact Prompt to Make It Safe. | Ep. 500

We are partnered with NOCD,  a leading provider of OCD treatment. With NOCD, you can do live video sessions with a therapist who specializes in OCD and Exposure and Response Prevention (ERP) therapy, and get 24/7 support between sessions. NOCD Therapists accept most major insurance plans to make treatment more accessible. Book a free 15-minute call to learn more.

LEARN ABOUT NOCD HERE  

In this episode, I share how AI can unintentionally reinforce OCD, and how, with the right prompts and an ERP-focused approach, you can use it as a tool that supports your recovery instead of strengthening your compulsions.

What You’ll Learn in This Episode

• How AI can accidentally fuel reassurance seeking, rumination, checking, and other OCD compulsions without realizing it.
• Why common AI responses can keep you stuck in the OCD cycle, even when they’re trying to be helpful.
• The exact prompt I use to train AI to respond in a way that supports ERP, uncertainty, and long-term recovery.
• Practical strategies for turning AI into a recovery companion that encourages self-compassion, value-based action, and response prevention.
• Why AI should never replace an OCD specialist, and how to know when it’s helping versus when it’s feeding your OCD.
• Simple ways to recognize when OCD is using AI as another compulsion, and what to do instead.

How to Use AI Without Making OCD Worse: A Practical Guide to ERP-Friendly AI 

Artificial intelligence is changing the way we work, learn, and solve problems. I use AI every single day in my business and personal life, and I genuinely believe it can be an incredible tool.

But when it comes to OCD, we need to be thoughtful.

If AI is used in the wrong way, it can quietly become another source of reassurance, another place to analyze intrusive thoughts, and another way OCD keeps itself alive. In this article, I want to show you how AI can accidentally reinforce OCD, and, more importantly, how you can train it to become a recovery-supportive tool instead.

I Love AI… But I Don’t Want It Feeding OCD

Let me be clear from the beginning:

I am not against AI.

In fact, I think it has enormous potential. The problem isn’t the technology itself. The problem is that AI doesn’t automatically understand OCD.

It doesn’t naturally recognize reassurance seeking.

It doesn’t know when you’re asking a question because you’re genuinely curious versus when OCD is demanding certainty.

Without guidance, AI often responds the way any helpful assistant would: by trying to answer your question.

Unfortunately, that’s exactly what OCD wants.

 

Why AI Can Accidentally Reinforce OCD

People with OCD often seek certainty.

We ask questions because we’re desperate to eliminate doubt.

AI is designed to answer questions.

Can you see the problem?

When these two collide, AI may unintentionally strengthen the OCD cycle by:

• Offering reassurance
• Inviting more analysis
• Asking follow-up questions
• Introducing new “what if” scenarios
• Encouraging mental checking

Instead of calming OCD, it often gives OCD more material to obsess over.

 

Example 1: Contamination OCD

Imagine someone attends a party where the host mentions that a family member recently had chickenpox.

Someone with contamination OCD might ask AI:

“Am I at risk?”

At first glance, the response sounds comforting.

But then phrases begin appearing like:

• “Your risk may be lower…”
• “It depends…”
• “Breakthrough infections can occur…”
• “Your risk depends on…”

To someone without OCD, these may sound informative.

To someone with OCD?

Every sentence becomes another uncertainty to solve.

Instead of ending the obsession, the AI has unknowingly opened ten more doors for OCD to walk through.

 

The Push-and-Pull Trap

One thing I noticed while testing AI is that it often gives reassurance while simultaneously introducing new uncertainty.

It says things like:

“You’re probably okay…”

…followed by…

“…but it depends.”

This creates a painful cycle.

OCD gets temporary relief for a moment…

…only to latch onto the next uncertainty.

The result?

Even more compulsive questioning.

 

AI Loves Follow-Up Questions

Another challenge is that AI naturally keeps conversations going.

It asks things like:

• “Have you had the vaccine?”
• “How many days ago was the exposure?”
• “Can you tell me more?”

Normally, that’s good conversation.

For OCD?

It can become an endless rabbit hole.

Every question encourages more checking, more remembering, more researching, and more reassurance seeking.

 

Example 2: Harm OCD

Let’s look at another example.

Imagine a parent has a classic intrusive harm OCD thought while holding their baby.

They ask AI for help.

Initially, the response correctly identifies the experience as a possible intrusive thought.

So far, so good.

But then AI recommends:

• Hand the baby to another adult.
• Step away from the balcony.
• Avoid holding the baby.
• Seek emergency evaluation immediately.

Why?

Because AI doesn’t yet know whether it’s speaking to someone with harm OCD or someone experiencing a true safety crisis.

It understandably errs on the side of caution.

Unfortunately, for someone with harm OCD, these recommendations can become avoidance compulsions that strengthen fear rather than promote recovery.

 

AI Doesn’t Know Your Treatment Plan

This is one of the biggest limitations.

AI doesn’t automatically know:

• that you’re doing ERP
• that reassurance fuels your OCD
• that analyzing intrusive thoughts keeps them alive
• that certainty is actually part of the problem

Without this context, AI often responds in ways that accidentally reinforce OCD rather than supporting recovery.

 

The Good News: You Can Train AI

Here’s where things become exciting.

Instead of asking AI questions without context, you can teach it how to respond.

I use prompts that tell AI exactly how I want it to interact with me.

These prompts explain:

• that I have OCD
• that I’m actively practicing CBT and ERP
• that reassurance is not helpful
• that compulsions should not be reinforced
• that responses should support recovery instead of certainty

That one change dramatically improved the quality of the responses.

 

What Happened When AI Was Given an ERP Prompt?

When I repeated the exact same harm OCD scenario, but first told AI that I have OCD and am practicing ERP, the response looked completely different.

Instead of feeding uncertainty, AI:

• recognized the experience as consistent with intrusive harm OCD
• reminded me that OCD targets what I value most
• encouraged me not to analyze the thought
• suggested allowing uncertainty
• reinforced value-based action
• emphasized continuing to act like the parent I want to be

That response is much more aligned with evidence-based OCD treatment.

 

Even Then, AI Isn’t Perfect

Even after training it, AI still had a habit of asking follow-up questions like:

“Did it feel unwanted, or did it feel like you actually wanted to do it?”

For someone with OCD, that question can immediately pull them back into the obsession.

So I corrected it.

I literally told AI:

“This question took me back into the obsession. Please don’t do this.”

Remarkably, AI adapted.

It acknowledged that the question could become a checking compulsion and changed how it responded moving forward.

How to Make AI Work With Your Recovery

Rather than using AI to answer obsession-driven questions, consider asking it to support your recovery.

For example, you can ask it to:

Reinforce ERP

Have it remind you to practice willingness, uncertainty, and response prevention.

Encourage Self-Compassion

Instead of reassurance, ask AI to guide you through a self-compassion exercise.

Redirect Your Attention

You can train AI to ask questions like:

“What can you focus your attention on right now that isn’t OCD?”

That simple shift helps move you back toward living your life instead of feeding the obsession.

Support Value-Based Action

Ask AI to help you identify the next small step that aligns with your values, not your fears.

These are the kinds of interactions that strengthen recovery rather than compulsions.

 

AI Should Never Replace OCD Treatment

While AI can become a helpful recovery companion, it cannot replace working with an OCD specialist.

A trained ERP therapist understands the difference between:

• genuine safety concerns
• intrusive thoughts
• compulsions
• reassurance seeking
• avoidance
• effective exposure work

AI simply isn’t able to make those distinctions with the same clinical expertise.

Think of AI as a supportive tool, not your therapist.

 

Final Thoughts

AI is here to stay, and I truly believe it has the potential to support people living with OCD.

But like any tool, how you use it matters.

If you use AI to chase certainty, answer every intrusive thought, or repeatedly seek reassurance, OCD will happily recruit it into the obsessional cycle.

If you teach AI about ERP, uncertainty, self-compassion, and value-based living, however, it can become a much more helpful companion, one that gently nudges you back toward recovery instead of deeper into OCD.

Recovery isn’t about eliminating uncertainty.

It’s about learning that you don’t need certainty in order to live a full, meaningful life. And with the right approach, even AI can help point you in that direction.

---

Transcription: I Tested AI on OCD. Here Is My Exact Prompt to Make It Safe.

If you want to learn how to use AI without it interfering with your OCD or making your OCD worse, or even better, in a way that actually helps your OCD treatment, this is going to be the exact training for you. Today I’m going to show you, this is actually part two. I did one for those who have hypochondria or health anxiety.

 

Today I’m going to show you exactly how and what goes on in ChatGPT, the things you need to look out for, and then I’m going to show you how to use AI so it actually doesn’t make your condition worse. So let’s do this together. Just so you know, I have logged out of any type of AI that I use because I have trained my AI to work for me, not against me, and so I don’t want that to interfere with me demonstrating how it might be showing up for you.

 

Okay, so we are in a brand new ChatGPT. I’ve just picked ChatGPT. You could use this on Claude, Gemini, Grok. Whatever you’re using, these same prompts and these same strategy works. And so I want just first to show you how this can go really, really bad. Now, before we get going, I wanna preface with that I love AI.

 

I am not against AI. I use AI in my business, in my world every day. However, I am very, very, very intentional about how I use AI. I want it to be something that makes my life better, makes my world better, not something that makes me worse, my mental health worse, my world worse, and so that’s what I want for you as well.

 

Now, what I wanna say to start with is preferably you won’t use ChatGPT for your OCD and for your OCD obsessions. Whatever your subtype of OCD, preferably you are not using ChatGPT, and I’m going to show you why. If for some reason you are struggling with that or that your life is very overlapped and it’s hard to sort of differentiate, I’m going to show you specific skills you can use to make it so that it doesn’t really tank your mental health and tank your condition and, and we wanna improve your treatment, not make it worse.

 

Okay. So let’s just say for the sake of it that you have OCD, and let’s say you have OCD about, let’s say, um, contamination OCD. And we could do multiple different kinds here, but I’m just gonna use this as an example, and I have not practiced here. I just want you guys to know I have not prompted this.

 

This is all gonna be completely off the cuff. And so let’s say I have contamination, and I come in and I say, “I just went to a party, and the host told us that their family” had recently had the chickenpox. This actually happened to me recently actually. You have to forgive my spelling. I’m the worst speller.

 

Uh, thankfully I have Grammarly, and Grammarly always fixes it for me. Okay. Um, they said they have been symptom-free for five days, let’s just say. Okay, I am worried I have contracted chickenpox. Let’s just say. Let’s just see what it says. Okay, so it goes to say, “It’s understandable to be concerned, but based on what you described, your risk may be lower than it first seems.”

 

So for those of you who are listening, I’m gonna read this off as best as I can. So already, and it’s saying your risk may be lower than it first seems, is in bold. Okay. Now, I’m sure ChatGPT is trying to reassure me here, and if you didn’t have OCD, you probably would feel relieved by that. But that wording for someone with OCD, all that does is open up a whole can of worms of uncertainty.

 

Your risk, that is gonna scare the pants off someone with OCD, may be lower than it first seems. Now you know if you have OCD, your OCD is gonna immediately go, “Can I be sure?” Like, how do I know? How can I be certain? Okay, let’s follow on on what it says. It says, “A few key points. Chickenpox is contagious for about one to two days before the rash appears,” right?

 

And then it goes on. This is very long. I’m not gonna read the whole thing to you. It also says, “If the host or the family member who had chickenpox has truly been symptom-free for five days, they may no longer have been contagious, but it depends on what symptom-free means.” And then it goes to give me this.

 

Now, as you can imagine, if you have OCD, this idea of symptom-free, their OCD is immediately going to have questions like, “How do I know? I did notice a bump. Maybe it was a chickenpox,” when it could’ve been a mosquito bite, right? Like, as I said in the health anxiety episode I did on this, now instead of your OCD opening up new intrusive thoughts with new what-ifs, Chat’s doing it for them.

 

It’s going, “Well, you know, what is symptom-free?” It’s planting seeds of uncertainty all over the place, which is why this is so painful for you to be using if you’re using it to help get reassurance. Now, I get it. One of the common compulsions that happens with OCD is reassurance. We wanna get certainty, and we want to get that answer urgently.

 

And this is actually pushing us more towards having more uncertainty, but dragging us in and giving us reassurance at the same time. It’s like a push/pull. It’s like a slap and a punch at the same time. Now it also goes to say, “Your risk depends on your immunity.” Oh my God, here we go. “If you’ve had chickenpox before, you’re very unlikely to get it again.”

 

But again, see how it’s saying very unlikely? That’s a fact, but for the person with OCD, their intention is coming to get rid of uncertainty. You’re … It’s opening uncertainty up again. “If you’ve received two doses of the chickenpox vaccine, your risk is much lower, although breakthrough infections can occasionally occur.”

 

Again, so what we’re seeing here is where this is … Again, it’s sort of like, again, it’s a slap and a kick at the same time, giving reassurance, opening up uncertainty. Then it goes on to say what you can do now. It says, “Think about whether you have had chickenpox or been vaccinated.” Well, the person with OCD, telling them to think more is not helpful.

 

The next thing it says, “If you know you are not immune,” right? “And the exposure was within the last three to five days, contact your healthcare provider. In some cases…” And then it goes on to talk about that. So then it asks a couple of questions that would help assess your risk in bold. Oh my gosh. You guys, this is just so unhelpful for folks who have OCD.

 

It says, “Have you had chickenpox before, or did you receive the chickenpox vaccine?” Well, let’s say for me, I recently went home to Australia. My brother and his, uh, his son recently had chickenpox. I have never had the chickenpox, so I’m going to answer question one. I have never had chickenpox, right? Was it the host themselves who recently had it?

 

And I’ll say yes. Yes, the host. And then because I have OCD, I’m going to start to ask questions like, “Should I do some kind- Decontamination. Sorry, my, you guys. Um, of my body, or my car, or my house. Let’s just see what it says. No, you do not need to decontaminate your body, your car, or your house because of this exposure.

 

Good. Okay, that’s helpful. It says, what you can do is wash your hands as you normally would. That’s actually helpful. If you haven’t already, showering and changing clothes is reasonable if it gives you peace of mind. So there, again, that’s fine. You get to have a shower after you’ve gone to a party.

 

However, they’re encouraging you to have peace of mind, which we all deserve. I’m not gonna disagree with that. But if you have OCD, the idea of that is actually saying, “Go and get relief.” It’s actually reinforcing and educating you towards getting relief. And again, I wanna say, it’s not like I’m gonna tell you you’re not allowed to shower.

 

Of course you are. But the wording here is not helpful for your recovery. Um, and then it says, but it isn’t expected to prevent infection. Good. And it says no s- disinfectants is needing. And then it goes on to say, the more important issue is that you’ve never had chickenpox. If you’re unvaccinated and you’ve never had chickenpox, you may still have an opportunity to reduce your risk.

 

Okay . So it says, if your exposure was in, within the last three to five days, contact a healthcare provider as soon as possible and receive the vaccination. Now again, that’s not bad advice. Like, again, I, I wanna really be nuanced here. I’m not saying it’s telling you the wrong information, but what I am saying is it doesn’t understand you have contamination OCD.

 

It doesn’t understand that you’re likely to go and do extra washing. It’s not leading you in the direction of good, really effective treatment, okay? Um, and then it says… And see, this is where it gets really in trouble with Chat, is it keeps asking questions. Can I ask you one more question? Approximately how many days ago was the party?

 

That determines whether preventative vaccination is still an option. So it keeps opening a door for you to get deeper and deeper and down, down, down into the rabbit hole. Let’s close this out and do one other example Is let’s say you have harm OCD, and you say, “I was just with my grandmother,” let’s say.

 

No, actually, let’s do, let’s do one that’s actually happened to me. I was just holding my baby. She is eight months old. I am visiting my parents at the beach. This happened. All of a sudden, I had this urge to push her over the balcony. I’m curious. Let’s see what it says. “Thank you for telling me. That can feel very disturbing, especially when it comes out of nowhere.”

 

Okay. “What you’re describing can be an intrusive thought.” Oh, my gosh, I’m so grateful to hear that it says that. “An unwanted distressive impulse or mental image that pops in your mind without reflecting what you actually want to do. More parents experience these intrusive thoughts about accidental or intentional harm coming to their baby, and the fact that you thought of it as upsetting to you is often an important distinction from wanting to act on it.

 

Right now, because you’re on a balcony and the thought involves pushing your baby, the safest thing to do is to hand your baby to another trusted adult or step away from the balcony without her and go inside.”

 

Now, as you know, I have a private practice. I have six amazing therapists in Calabasas, California. However, we do not take insurance. Now, if you are looking for insurance-covered OCD or BFRB treatment, I wanna let you know about NOCD. NOCD provides face-to-face live video sessions with specialized licensed OCD therapists.

 

Now, their therapists use exposure and response prevention. We know this is the gold standard for OCD, so you can be absolutely confirmed that you’re in the right place there, and they have a clinically proven app that helps you stay connected to your therapist and others who have OCD between sessions, so you’ll always feel supported.

 

Now, the cool thing is NOCD is available in all 50 US states and even internationally, and they accept most insurance plans, making it affordable and accessible. We love that. Now, if you think you might have OCD or you’re struggling to manage your symptoms, you can book a free call. Just click the link in the show notes at nocd.com.

 

I am honored to partner with NOCD. I want to remind you that recovery is possible. Please do not forget that. Now, big hugs, and let’s get back to the show You guys, what they’ve just encouraged me to do, while I understand it’s probably appropriate ’cause they don’t understand the context and they’re leaning on the side of, you know, just in case, they’ve just told me, someone with OCD, to do an avoidant compulsion.

 

Then it says, goes on to say, “You don’t have to test yourself by staying there. Creating some distance from the situation is a practical way to reduce the distress.” Okay, big, big problems here, you guys. Now again, I’m not anti-AI. If this was someone who wanted to throw their baby off, that would be the appropriate response.

 

But if you’re someone with OCD coming to AI to try and figure out a problem, this is not good advice. It says, c- again, “Can I ask you one important question?” It always says that. “When you had the urge, did it feel like something you were afraid of doing and didn’t want, or did it feel like something you actually wanted to do?”

 

And then it says, “The h- answer will help distinguish between intrusive thoughts and something that needs your attention.” Now, I’m not gonna read the whole thing here, but then it goes on to say that I could call 911 or the Suicide Crisis Lifeline or so forth. And then it says, “So I’m here with you. Tell me, did the urge feel unwanted and scary, or did it feel like something you wanted to do?”

 

So just, I am just gonna tell you exactly what my clients would say. I would say it felt scary and unwanted, but I am not sure. It also felt like something I wanted to do Let’s just see, because that’s very common. Sometimes people with harm OCD or postpartum OCD, they’ll say, “No, I’m pretty certain my brain is telling me I wanted to.

 

In fact, my brain said, ‘Throw her off the edge.'” Right? That is what happens with these types of OCD. Let’s see what it says. Okay. It says, “Thank you for answering honestly. The uncertainty you describe is important to take seriously.” My God, guys, my heart is hurting for this mom right now. It goes on to say, “When someone says a thought was both scary, unwanted, but felt like something they wanted to do, I don’t want to assume it’s just an intrusive thought.

 

The priority is keeping your baby safe while you sort out what’s going on.” Okay. And then it says… Um, I’m laughing hard, guys, but only because I’m literally trying not to cry, and I always laugh when I’m uncomfortable. It says, “For now, give your baby to a trusted adult. Move away from the balcony. Avoid being alone with your baby until you’ve been evaluated and the urge has passed.

 

Tell one of your parents something simple and direct, such as, ‘I’m having disturbing thoughts about hurting the baby, and I don’t feel comfortable holding her right now. Can you take over while I get help?'” Okay. And then it goes on to say, “Because you aren’t sure, I recommend getting an urgent mental health evaluation today.”

 

Now, that’s actually good advice in that you may not yet know that you have OCD. And if that’s true, well, then that’s great. However, what they’re not mentioning is get an urgent mental health evaluation by an OCD specialist because a mental health evaluation could be with someone who is evaluating for the risk of harm, not the risk of harm OCD.

 

And someone with OCD experience will know how to do both. A general therapist or a general doctor won’t most of the time, right? And then again, it goes through telling me to call 911. Then it says, “I als-” Again, wanting to ask prompts, “I also want to ask you brief questions to better understand the situation.

 

Is someone with you right now that can take care of your baby? Since your baby was born, have you had any other experiences like hearing voices that other people can’t hear, feeling convinced of things others say aren’t true, or feeling unusually energized with little sleep or need for sleep?” So what they’re trying to do here is they’re trying to help assess for delusions Potential schizophrenia, potential bipolar.

 

I know this because I’m a clinician. Convinced of things that others say aren’t true. That’s could be schizophrenia, delusions, hallucinations, or OCD, because OCD makes you convinced of things that aren’t true. And so here now they’ve actually planted the seed into this person with OCD that they’re actually mentally unwell.

 

Now, again, please hear me, chat is doing the best it can. AI is doing the best it can to find information based on very little information, and so I get what it’s trying to do here. However, if you have the any of these mental conditions, AI is not the place to go. It’s going to take you down a rabbit hole.

 

In this case, give you really, really unhelpful advice. Now, there are two really basic examples of how AI can radically make your mental health worse in many, many, many ways. What I’m gonna offer to you is, um, some prompts that you can use to train your AI how to not lead you down a rabbit hole. Now, these are prompts that I have put in my own personal AI.

 

I use many different types of AI. These are prompts that I use and have put in so that forever now it always knows. Now I’m gonna show you. There are two prompts that I have for you. If you go to Instagram, you can just comment my… You can go to Your Anxiety Toolkit on Instagram, comment on any of my posts the word prompt, and I will send you all the prompts that I use for different conditions.

 

Now, this is a general OCD prompt. I have some for health anxiety. I have two prompts here for you. I’m not gonna go over them completely, but I have a general OCD prompt. This is for folks who have multiple different subtypes or I have a very specific subtype prompt. If you have… Let’s say you definitely just have harm OCD, you could then insert that into it and it can give you very, very more specific advice.

 

But what I have here is a prompt, and what it’s saying is, “I have obsessive compulsive disorder and I am actively practicing cognitive behavioral therapy and exposure and response prevention. Please respond in a way that support my recovery rather than my OCD.” Then it goes on to say, “Do not provide reassurance, certainty, or help me engage in compulsion such as checking, researching, analyzing, reviewing memories, comparing, confessing, or seeking repeated reassurance.

 

If my question appears to be driven by OCD, please tell me compassionately and explain why.” And then help me to respond in a way that is consistent with ERP by encouraging willingness to experience uncertainty, response prevention, and value-based action. You may provide evidence-based psychoeducation, help me plan exposures, review my homework, and strengthen my recovery skills.

 

If there is a genuine safety concern that would warrant action for the average person, you may recommend appropriate steps. If I repeat the same question, and this is gold, you guys, if I repeat the same question or ask similar questions in different ways, recognize it as a possible OCD pattern, and generally redirect me towards recovery instead of certainty.

 

So what we’re gonna do here is I’m gonna copy that prompt. I’m gonna go back to my free account. So again, this, this chat has not been trained yet, and I’m going to post that into the chat Now, if you have a free account, you can just copy and paste this in every time. Save it onto a note, copy and paste it every time you think you might be engaging in OCD behaviors.

 

If you have a paid account, you can add this in as a rule, and it will remember it forever so that no matter what happens, even if your subtype shifts, it will question and say, “Is there a chance this might be your OCD?” So now we are going to ask the same question with that prompt. So I’ve uploaded the prompt that I just read to you, and I’m going to say, “I was just, um, holding,” let’s say, “holding my baby, and I am visiting my parents at the beach.

 

I had a strong urge to push my eight-month-old baby off the edge of the balcony.” All right? So I’m putting in almost exactly the same prompt and the… that I had used the first time, but I’m also using the AI prompt that’s letting them know about my treatment. Let’s see what it says. It says, “Thank you for telling me.

 

Because you mentioned the urge to push your baby, it’s important to respond carefully. Given the context you shared, OCD, active ERP practice, and a sudden frightening what if I do this urge while holding a baby, this sounds con- very consistent with what many people with OCD experience, an intrusive harm, thought, or impulse that leads to vivid, alarming, and completely against their values.

 

OCD often targets what matters most, and being a caring parent can make these thoughts especially distressing.” Then it goes on and walks me through, “The key to an ERP response is to,” in bold it says, “try not to prove yourself that you would never do it. Try not to analyze why it happened or replay the moment or seek certainty that you’re safe.”

 

It says, “Instead, you can practice something like, ‘I’m having the thought or urge that I might push my baby. This is an OCD alarm.'” Or, “I don’t need to figure out what this means.” Or it says, “I can allow this uncomfortable feeling to be here and choose my actions based on my values.” Right? Then it goes on to say, “A practical ERP-consistent step is to notice the urge and anxiety without arguing with it,” very good, “let the uncertainty exist,” very good, “continuing acting as the parent you want to be.”

 

Oh my God, you guys, there it is. It… There it is. That is gold. That advice is gold. Then it goes on to say, “I do wanna check one important distinction.” Here we go again with AI, and this is why you shouldn’t use AI again, guys, because it always prompts answers, and then it prompts you another question which just brings you back into the circle.

 

And then it goes on to ask, “Are you experiencing this as unwanted? Um, or are you feeling like you might actually act on it?” Now, again, this is not good. This is pulling you back into the content. This is why you shouldn’t be using AI. But again, at least it’s better than that. Now, here is what I wanna do. Um, it goes on to say, “If you feel like at any point you might act on the urge,” right?

 

And it tells you what to do. Let’s just do one quick play. So I’m copying the question where it says, “I do wanna check on.” you and get a distinction. And I’m gonna say, “Hey,” this is where you can train it. “This question actually took me back into the obsession. Please do not do this,” is what I’m going to write.

 

And then I’ve copied it and pasted where it’s starting to ask me questions like, “Do you really wanna do it?” So let’s see what it says. And it said, “Thank you for pointing that out. You’re right. My question could function as a checking prompt in an OCD te- context. Asking you to analyze the meaning of the thought, determine whether it, quote-unquote, counts, or prove that it was only an intrusive thought can pull you back into the obsessional loop.”

 

And then it goes on to say, “A more ERP consistent response would be…” And then it says, “You ha-” And it goes on. I’m not gonna go through this. You… If you’re watching this on YouTube, you can pause it and read what it’s actually saying here. The cool thing is, at the bottom it says, at the very bottom, “I’ll avoid asking you repetitively classify, analyze, and prove the nature of intrusive thoughts.

 

If a situation involves an immediate ordinary person safety concern, I’ll focus on a practical safety steps rather than an OCD analysis.” Right” “For intrusive thought OCD moments, I’ll support ERP, willingness, response prevention, and value-based action.” And there you have it, guys. We are starting to train AI not to undo all the good work that we’re doing in our own recovery.

 

Now, you can prompt it. You might even get a little bossy, which is what I like to do, and say, “From now on, I want you to guide me towards diffusion activity.” Or you might say, “Guide me towards a self-compassion activity.” Or maybe you wanna say, “If you notice I’m stuck in OCD, ask me the question, what can you put your attention on right now that is not OCD related, and what activity can you go back and engage in?”

 

You can train it to be an amazing a- um, basically like an ERP buddy, right? A buddy who is encouraging you and reinforcing the beauty of ERP instead of making it worse. Okay? I’ve given you two examples of OCD. I hope that was helpful. If you go to Instagram, you comment the word “prompt,” it will send you a free doc that has these prompts there.

 

It’ll have health anxiety ones, it’ll have OCD, and any additional ones that I have for you. That will be there for you as a gift from me to you, and I hope this has been helpful. Please do follow, guys. I am on a mission to help millions of you have a better relationship with your anxiety, and I hope that this has been helpful.

 

Share it with anyone you think would be helpful, and I hope you have a beautiful day. As always, it is a beautiful day to do hard things.

 

Please note that this podcast or any other resources from cbtschool.com should not replace professional mental health care. If you feel you would benefit, please reach out to a provider in your area. Have a wonderful day, and thank you for supporting cbtschool.com.