Oct 2026
A recap of our live webinar with EMDR trainer Paula Merucci on September 30, 2026: why images are the doorway into trauma memories, and how AI image generation can make them more powerful.
In July, Paula Merucci showed how she uses AI image generation in EMDR and IFS, and hundreds of therapists tuned in. The response was so strong that we built an image generator directly into WeMind. On September 30, Paula came back to go a few steps deeper, this time with real cases and a live demonstration.
Paula is an EMDRIA approved trainer and consultant based in Chicago, a basic trainer, and a regular presenter at EMDRIA. She works mostly with complex trauma, eating disorders and addiction, and integrates EMDR with IFS. As she put it in the webinar, using AI images has been a real game changer in her practice.
Here is what she shared.
Why images matter in EMDR
EMDR has always started with an image. Francine Shapiro described it as a picture that represents the memory, and Paula teaches it in her basic trainings as the snapshot of the worst moment. That snapshot is the anchor: the doorway into the neural network that holds the trauma.
Research on working memory taxation, much of it from the Netherlands, shows that recalling a memory while doing a demanding task, such as eye movements, makes the image less vivid and less emotional. Paula highlighted one finding in particular: the image shifts first, and the emotion follows. "I used to say, what comes first, the chicken or the egg, the image or the emotions? No, it's image." As the vividness of the image decreases, the SUD goes down, and the client's perspective changes.
That is also why the snapshot matters so much. Paula gave the example of a client who wants to work on being bullied. There is a lot in that memory. But when she asked for the snapshot of the worst part, the client said: when the counsellor turned around and walked away. That is a completely different image, and a completely different theme: abandonment. A specific, two dimensional snapshot also helps with titration, because the client works with one moment instead of the whole event.
Three types of visuals
Paula works with three kinds of images, and an image generator can help with all of them.
The first is the memory snapshot: the single moment that represents the worst part of a memory. The second is the body sensation image. With complex trauma, many memories are implicit and non-verbal, so Paula helps clients put language to what they feel through colour, shape, texture and boundaries. That description becomes an image in its own right. The third is the metaphor, such as a client who says it feels like a train going 200 miles an hour.
What matters most, Paula stressed, is the client's relationship with the image. What are they noticing? What are they feeling toward it? A heavy, dark body sensation might bring up shutdown, while the speeding train might bring up energy. As the image changes in the client's mind, their perspective changes too. "At the end of the day, that's what this work is. It's all about perspective."
Real examples from Paula's practice
A wrist in a cast
A client came in angry and upset about an injured wrist, just before a planned holiday. The worst part was the bandaged wrist itself, with the belief "I'm powerless" and a SUD of 10. Paula used a generic image of a wrist in a cast as the target. "Is this my client's wrist? No. Did it work? Absolutely."
After about six sets of bilateral stimulation in WeMind, the client said something like: I get it, I'm doing the best I can. While still talking with the client, Paula typed a new prompt into the image generator, checked with the client that the new image felt right, and used it to anchor the new positive cognition: I can recognise what I can and can't control, and be okay with it. Phase three took about five minutes, phase four around fifteen to twenty.
Two parts on a couch
Another client, who grew up in an abusive home, felt torn between protecting her son and disciplining him. From an IFS perspective, a younger part wanted to protect him, and an adult part knew she needed to be a responsible parent. Paula generated a couch with two pillows, one saying protection and one saying discipline, and started bilateral stimulation to see which part would take the lead. It was the younger part.
When the client then described a physical sensation, "a dark bruise with deep purple and blue tones, dense, heavy and painful", Paula turned that description into an image too. She started with a small version of the image against the client's peaceful place, a forest trail in northern Michigan, and gradually made the exposure bigger, combining bilateral stimulation with an extra task. As homework, the client created her own image between sessions and looked at it once a day, a way to build the self to part relationship that is central to IFS.
OCD and the anxiety protocol
In an earlier case, before WeMind had its own image generator, Paula worked with a client who washed his hands until they bled. Using the anxiety protocol, which starts with the flash forward, she created an image for each point of disturbance: the feared outcome, the last time it happened, the worst time, and finally the first time, a teacher's disgusted look in eighth grade. After desensitising each image, the hand washing improved dramatically. Only a year later did it become clear what lay underneath: that teacher had been abusive. "I don't do this protocol anymore without the images."
Tips for working with an image generator
The live demo showed that there is a small learning curve. Paula asked for a couch with "I want to drink" on one side and "I can't drink" on the other, and the first attempts missed the words. It took three tries. The lesson: be explicit. Say exactly what you want, keep it short, and put text you want on the image between quotation marks, so the AI knows those are words to show rather than an instruction.
AI image generators also have boundaries. They tend to refuse prompts that touch on violence, sex or drugs, so a description of a traumatic moment may sometimes return an error. Rephrasing in more neutral or symbolic language usually helps.
The practical benefit is time. In WeMind, Paula can type a prompt while she keeps working with her client, and pick up the image a few seconds later. "Now I don't have to go to my ChatGPT and make the image and bring it back, which I never would do because it was too time consuming. I can do it right in the program." The image can sit in the background during bilateral stimulation or EMDR 2.0, and she can make it smaller or bigger to adjust the exposure.
One question from the audience was whether the image changes during reprocessing. It does not, and that is the point. The image is the anchor. It keeps the client on target, which is especially useful in EMD, EMDR 2.0 or when working with clients who tend to drift to other memories. What changes is the client's perspective on it. Paula also works this way in person: the client holds a phone or iPad and responds to the tasks like a game controller, while she watches the screen.
What about privacy and ethics?
Using AI in therapy raises fair questions, and the webinar did not shy away from them. For image generation, the risk is relatively low: you describe a general scene, such as a wrist in a cast or a couch with two pillows, not a real person or personal health information. The image generator in WeMind does not create the faces of real people.
It becomes more sensitive when you use AI with session content, such as transcripts or summaries. Sander Kamphuis reminded attendees never to paste personal health information into free AI tools. Tools like ChatGPT can be used in a compliant way, but in the US that requires a signed Business Associate Agreement, which is only available on enterprise plans.
Paula also shared a more personal reflection. She only uses transcription with a few clients who have explicitly asked for it. Her concern is not the tool, but how a client might feel sharing something deeply intimate while knowing it is being transcribed, and whether a client who wants to please their therapist feels truly free to say no. It is a good reminder that consent is about more than a signature, and that every therapist needs to decide what feels right for their clients.
The takeaway
Images have always been at the heart of EMDR. What AI image generation adds is speed and flexibility: you can turn a snapshot, a body sensation or a metaphor into something your client can actually see, in the moment, without leaving the session. Used thoughtfully, with clear consent and without personal health information, it can make targets more concrete, parts work more tangible and homework more meaningful.
The bottom line
✅ The image comes first: research shows vividness shifts before emotion, so a clear target image matters
✅ Three types of visuals: memory snapshots, body sensations and metaphors can all become images
✅ Create images in session: type a prompt while you keep working with your client
✅ Be explicit in your prompts: short descriptions and words in quotation marks work best
✅ Keep privacy in mind: describe general scenes and never share personal health information with free AI tools
Frequently Asked Questions
Can I use AI image generation in standard EMDR protocol?
Yes. Paula uses images in standard protocol, EMD, EMDR 2.0 and IFS informed EMDR. The image serves as the target and anchor, whatever approach you use.
Do I create the images during the session or beforehand?
Both are possible. In WeMind, you can type a prompt while you keep working with your client and use the image a few seconds later. You can also prepare images in advance or upload your own.
Does the image change during reprocessing?
No. The image stays the same and keeps the client on target. What changes is the vividness, the emotional charge and the client's perspective.
Can I use this with clients in person?
Yes. The client can use a phone or iPad in the room, while you control the session from your own screen.
Is AI image generation safe for client privacy?
Image prompts describe general scenes rather than real people or health information, which keeps the risk low. Never put personal health information into free AI tools.
Watch the full webinar
In the recording, Paula walks through more cases and gives a live demonstration of the image generator in WeMind.
Watch the webinar on YouTube →
How ready is your practice for AI?
Curious where AI could make a real difference in your own work? Answer 9 quick questions about your week, your notes and your tools. About two minutes, no client information needed.
