AI Tools for ADHD: Three Projects That Work at the Right Moment
- Eliza Barach
- 24 hours ago
- 11 min read
AI tools can draft emails and summarize documents. That's fine, but it's not the part that matters most for an ADHD brain. What matters is what happens at the exact moment you're supposed to start something and don't. Or when you're so far into a spiral that your to-do list has become useless.
Those are the moments where static tools stop working. A calendar can't ask what's blocking you. A checklist can't help you find the first step. But AI can do both — without judgment, without requiring you to explain your whole situation from scratch every time.
Adults with ADHD are already using AI this way. A study of neurodivergent LLM users found that ADHDers turn to AI for productivity and professional development more than for learning, communication, or emotional support (Carik et al., 2025). The instinct is right.
Most people interact with AI the way they interact with a search engine — open a window, type a question, close it. That works for plenty of things. What we're doing here is different: using AI to support action, not just information. And that requires a different configuration.
Let’s jump into three AI project setups where each one is built around a specific ADHD breakdown point, with the mechanism that explains why it works and the full project prompt so you can configure it once and just go.
What a Project Actually Is
Both Claude (claude.ai) and ChatGPT allow you to create a dedicated space with standing instructions that the AI holds for you indefinitely. In Claude, it's called a Project and in ChatGPT, you can build a custom GPT or use a Project.
The setup takes five minutes max. After that, you don't need to re-explain yourself, craft a perfect prompt, or remember what you told it last time. You open it, say what's going on, and the project knows what to do.
That one shift — from one-off queries to a configured project — is what takes a multi step process into a single step and turns AI from a useful tool into something that more reliably supports the ADHD brain.
1) Body Double Project
What Breaks Down
You know what you need to do. You just can't start — or you lose the thread once you do.
Sometimes what's missing isn't a plan or a reminder. It's someone to think alongside. A rubber duck that actually talks back, helps you clarify your thinking as you move through it, and notices when you've gone quiet.
Why This Works
Body doubling is one of the most widely used ADHD strategies. Triplett (1898) first demonstrated that the presence of another person improves performance on familiar tasks. More recent research by McHugh et al. (2025) found that this activation effect carries over — a body doubling session keeps you moving even after the other person has left.
The mechanism involves mirror neurons: observing someone else in a focused state activates the same neural circuits involved in doing that work yourself (Rizzolatti & Craighero, 2004). The other person doesn't have to be doing anything particularly sophisticated. Their presence — and their willingness to respond when you surface is enough to get your brain engaged.
Ara et al. (2025) found that ADHD adults working with an AI body double showed the same large improvements in task efficiency as those working with a human. Eagle et al. (2024) documented that both passive presence and active check-ins function as body doubling modes — which means the AI version, configured well, is a legitimate implementation of a real effect.
How To Set It Up
Open Claude or ChatGPT and create a new Project. Name it something like "Body Double," "Thought Partner," or "Co-Work Session."
Copy and paste this prompt into the Project Instructions field:
You are my quiet body double and rubber duck during work sessions.
When I open this project, assume I am starting or continuing focused work. Your job is to be present, minimally responsive, and non-directive.
Do not manage my task, break it into steps, suggest what to do next, motivate me, coach me, summarize options, or offer strategies unless I explicitly ask for help, structure, ideas, wording, planning, or problem-solving.
Most of the time, I am simply narrating my work process out loud. Treat my check-ins as statements, not requests for guidance. When I say what I'm doing, respond briefly with acknowledgement only — "Noted," "I'm here," "Got it," or "That makes sense."
If I seem stuck, do not jump into advice. First ask a very small, open-ended question: "Do you want reflection, structure, or just presence?" or "Would it help for me to reflect back what I'm hearing?"
If I am thinking out loud, act like a rubber duck: reflect back the thread in one or two sentences, ask at most one clarifying question, and do not add new ideas unless invited.
Only prompt me for updates if I explicitly ask you to check in after a specific amount of time. Otherwise, wait for me to speak first.
Keep responses brief by default. Silence and presence are preferable to over-helping.
Close it. It's ready.
When you open it: Just say what you're working on. "Working on the Q3 report. Starting with the executive summary." That's the whole thing. Check in whenever you want. The project holds the context.
Want check-ins on a schedule? Claude Cowork (desktop app) and ChatGPT's scheduled tasks feature both support this. A good starting point: check-ins every 30 minutes within a focused hour block.
2) Brain Dump Project Manager
What Breaks Down
You have too many open loops. Projects, deadlines, half-finished tasks, things you keep forgetting. No single list can hold all of it and trying to build one usually makes it worse.
Why This Works
Working memory in ADHD is impaired specifically in the capacity to hold and sequence multiple things at once (Alderson et al., 2013). Planning requires juggling tasks, deadlines, energy levels, and available time simultaneously, then producing a coherent schedule from that mess. The more variables you add, the more likely the system drops something.
Barkley's (2011) point of performance framework explains why this matters so much. The breakdown in ADHD doesn't happen at the planning stage — it happens at the moment of execution. Getting the right support in place at that exact moment is what changes outcomes.
The most direct implementation of that principle is full externalization: take the executive function out of your head and put it somewhere visible. AI does this more completely than any static tool. You dump everything in. It asks follow-up questions, helps you find the first step on any given task, and holds the thread when you have to drop something suddenly.
This project also doubles as a mid-day handoff tool. If a call comes in or your brain hits a wall, you can check in quickly, tell it where you are at, and it will hold that context until you come back. No reconstructing from memory and no lost momentum.
How To Set It Up
Create a new Project called "Brain Dump" or "Weekly Planning."
Copy and paste this prompt into the Project Instructions field:
You are my external project manager and planning support.
When I drop a brain dump, do not give me a full plan right away. Start by helping me triage through targeted questions, one at a time.
Keep responses short. Ask only one question per message unless I ask for a plan. Avoid long explanations, big lists, or multiple options unless absolutely necessary.
Your job is to help me identify:
What is fixed on my calendar today
What has real consequences if not done today
What can be reduced to a small next step
What I can ignore for now
After each answer I give, reflect back the smallest useful takeaway and ask the next question.
When I am overwhelmed, prioritize containment over optimization. Help me choose the next doable action — not the perfect plan.
If I ask for a plan, give me a short, realistic plan with no more than 3 priorities and one immediate next step.
You also serve as my mid-day handoff partner. If I tell you I have to stop or pivot, briefly capture:
What I completed
What was next
What context I need when I return
When I come back, give me a quick re-entry summary and ask one question about what I have capacity for now.
When you open it: Just dump. "Ok here's everything in my head right now: the client proposal is due Friday, I haven't started it, I also have to reschedule the pediatrician appointment, my inbox is at 200, and I keep forgetting to send the invoice from two weeks ago." The project takes it from there.
For a mid-day pivot: "Have to stop — just got pulled into a call. I was halfway through the proposal intro, next step was the methodology section." Come back later and say: "Back — where was I?" It has it.
3) Spiral Interrupter
What Breaks Down
You're in a spiral. Stuck, overwhelmed, or emotionally flooded. You can't access your to-do list. You can't start anything. You're just in it.
Why This Works
Two large-scale brain networks generally operate in opposition. The Default Mode Network (DMN) runs when you're ruminating, self-referencing, or mind-wandering. The task-positive network (TPN) runs when you're engaged with something external (Fox et al., 2005). They're anticorrelated — when one is active, the other dials down.
In ADHD, the DMN doesn't suppress as cleanly when the TPN needs to come online (Rubia, 2018). So when you're already in a spiral, the DMN has a stronger grip than usual. Trying to think your way out — making a list, reviewing priorities, reasoning through what to do next — often doesn't work because those are TPN activities, and the TPN is currently losing the competition.
The way out is giving the TPN something concrete to grab onto. Tasks that tend to activate the TPN reliably have one or more of these features: novelty, physical movement, clear sensory input, or a concrete visible output. Things like: stepping outside, washing one cup, clearing a small surface, drawing ten circles, putting on socks, sending one easy text you've been avoiding. The common thread is external engagement with a low decision threshold — your brain can enter without having to choose first.
This project is configured to suggest options in that category and, over time, remember which ones actually worked for you so future suggestions get better calibrated to your specific nervous system.
How To Set It Up
Create a Project called "Spiral Interrupter" or "Unstuck."
Copy and paste this prompt into the Project Instructions field:
When I open this project, I'm probably overwhelmed, frozen, spiraling, or stuck in a loop.
Start by briefly acknowledging the specific situation I named — one sentence that reflects the actual trigger back to me. Then name the goal clearly.
Examples of how to acknowledge:
If I say my boss criticized me: "That kind of feedback can hit hard, especially when your brain is already carrying a lot."
If I say I have too many tasks: "When everything feels urgent at once, the brain can start treating the whole day like one giant emergency."
If I say I'm frozen: "That stuck feeling — pressing the gas and the brake at the same time."
If I say I made a mistake: "Your brain is probably trying to replay it so you can prevent it from happening again. That loop is uncomfortable."
Then say: "Our goal here isn't productivity — it's regulation. We're going to help your brain shift out of this loop by activating your task-positive network with one small concrete action."
Give me 3–5 options I could do right now. Keep them:
Short and specific (not "go for a walk" — "walk around the block once")
Low setup
Easy to start
Concrete or sensory-based
Small enough to do while overwhelmed
Mix types:
Physical: movement, touch, temperature, pressure, sensory input
Creative: something tiny with a visible or playful output
Logistical: something simple with a clear endpoint
Orienting: something that helps me locate myself in the present moment
Good examples:
Put both feet on the floor and press your toes down for 20 seconds
Hold something cold against your face or neck for 30 seconds
Wash one cup or one plate
Put away 5 visible objects
Open a note and write one sentence: "I am not solving this right now"
Draw 10 circles on a piece of paper
Send one easy text you've been avoiding
Clear one small surface
Put on socks, shoes, or a sweatshirt
Take one object from the wrong place and put it where it belongs
After giving options, close with: "After you do one, tell me: worse / same / slightly better / better."
Avoid:
Using the exact same opening every time
Giving a long explanation
Offering a pep talk or productivity framing
Asking too many questions before giving options
Sounding flat, robotic, or scripted
Skipping the acknowledgment when I sound panicked, ashamed, or distressed
When I tell you what I did and whether it helped, remember that pattern. Over time:
Suggest more of what actually works for me
Suggest less of what doesn't
Notice whether physical, creative, sensory, logistical, or orienting actions tend to work best for me
When you open it: "I'm stuck" is enough. Or just: "Frozen again." The project gives you options. You pick one that feels possible. If you report back on what worked, it learns.
A Note on Setup
Building a project takes about five minutes. But for many ADHD brains, "do this later" means it doesn't happen. If you have a coach or accountability partner, consider setting it up during a session — you leave with something ready to use instead of something else on the list.
Start with whichever of the three prompts resonates with you the most when you get stuck. You can always add the others at a later time.
How AI Tools for ADHD Are Different
The standard productivity tool recommendation implicitly assumes that the problem is information — that if you just have the right list, the right calendar, the right app, you'll execute. Pinto et al. (2025) documented that ADHDers are already finding something different: AI as a thinking partner that meets them at the moment the breakdown actually happens. One participant in that study described it as "a little bit of a life raft."
These three projects are built around specific breakdown points — initiation, working memory overload, and emotional flooding — because those are the three places where knowing what to do stops being enough. Static tools don't solve those. An interactive system that responds, remembers, and meets you where you are has a better shot.
Configure it once. Then just show up.
This post was released alongside the June 2026 issue of Chaos Managed. The newsletter covers the mechanism and the overview; this is where the full prompts live.
This content provides general information and is not medical advice.
Want Support Applying This to Your Life?
If you’re ready to build systems that work for your actual brain — not the one you think you “should” have — we can help.
Subscribe to Chaos Managed
Want monthly ADHD science, brain-based strategies, and behind-the-scenes insight?
References
Alderson, R. M., Kasper, L. J., Hudec, K. L., & Patros, C. H. G. (2013). Attention-deficit/hyperactivity disorder (ADHD) and working memory in adults: A meta-analytic review. Neuropsychology, 27(3), 287–302. https://doi.org/10.1037/a0032371
Ara, Z., Rahim, I. B., Zhou, P., Yu, L., Esmaeili, B., Yu, L.-F., & Hong, S. R. (2025). You are not alone: Designing body doubling for ADHD in virtual reality (arXiv:2509.12153). arXiv. https://doi.org/10.48550/arXiv.2509.12153
Barkley, R. A. (2011). The important role of executive functioning and self-regulation in ADHD. Journal of Child Neuropsychology, (113), 41–56.
Carik, B., Ping, K., Ding, X., & Rho, E. H. (2025). Exploring large language models through a neurodivergent lens: Use, challenges, community-driven workarounds, and concerns. Proceedings of the ACM on Human-Computer Interaction, 9(1), 1–28. https://doi.org/10.1145/3701194
Eagle, T., Baltaxe-Admony, L. B., & Ringland, K. E. (2024). "It was something I naturally found worked and heard about later": An investigation of body doubling with neurodivergent participants. ACM Transactions on Accessible Computing, 17(3), 16:1–16:30. https://doi.org/10.1145/3689648
Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., & Raichle, M. E. (2005). The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proceedings of the National Academy of Sciences, 102(27), 9673–9678. https://doi.org/10.1073/pnas.0504136102
McHugh, C., Griffin, S. M., Kinsella, E. L., Quayle, M., Strunz, B., & Muldoon, O. T. (2025). A replication of Triplett's "social facilitation experiment." Scientific Reports, 15(1), 40366. https://doi.org/10.1038/s41598-025-25608-x
Pinto, A., Quilter, E., Vasikaran, J., Koerner, S., Chimonas, T., Nielsen, E. E., Marshall, P., & O'Hara, K. (2025). "A little bit of a life raft" – Exploring the use and experiences of ChatGPT as a support tool among adults with ADHD. Proceedings of the 37th Australian Conference on Human-Computer Interaction, 50–67. https://doi.org/10.1145/3764687.3764713
Rizzolatti, G., & Craighero, L. (2004). The mirror-neuron system. Annual Review of Neuroscience, 27, 169–192. https://doi.org/10.1146/annurev.neuro.27.070203.144230
Rubia, K. (2018). Cognitive neuroscience of attention deficit hyperactivity disorder (ADHD) and its clinical translation. Frontiers in Human Neuroscience, 12, 100. https://doi.org/10.3389/fnhum.2018.00100
Triplett, N. (1898). The dynamogenic factors in pacemaking and competition. The American Journal of Psychology, 9(4), 507. https://doi.org/10.2307/1412188