Taming AI Overwhelm: How a Simple Instruction Helps ADHD Brains (and Everyone Else)

Have you experienced this with AI? You ask a simple question, expecting a concise bullet-point answer or a short step-by-step guide, only to receive a full-blown essay in return. That is AI trying to “spit” tons of words to please us and demonstrate how thorough its responses can be. Most of the time, however, we simply do not need that.

Over time, consistently using a single dedicated AI tool with memory and personalisation enabled will teach the AI how to respond to you. Yet, new topics still frequently turn into an unnecessary wall of text.

After reading just a paragraph or two, your brain already has follow-up questions, and you do not really have time to read all the way to the end.

Sometimes you do not require AI to resolve your query with a ready-made solution immediately. Instead, as part of the learning process, you want to ask questions and analyse answers to gradually reach a point where you have learned something with the AI – rather than having the AI do the work for you.

People with differently wired brains, as well as highly intelligent individuals, do not use AI to replace their work. They need help processing significant amounts of information in a way that allows their brain to keep things organised, structured, and easy to comprehend.

Sometimes they face an urgent need to learn something new, but they do not have the time or right conditions to do so. This is where they need a quick, swift answer from AI. The answer needs to be straight to the point, easy to understand, and simple to follow…

…and then the AI does its thing, giving you a lecture about how to boil eggs in the style of a professor explaining quantum mechanics.

Imagine if your brain is wired differently and you are unable to process information in an unorganised, unstructured way.

This is where ADHD brains struggle. People who have too many things to handle at the same time will struggle in the same way.

ADHD is just a diagnosis, or a set of common traits that have been identified, following a framework to determine what “ADHD” is. Some people with high-functioning brains may have traits of ADHD, but will not necessarily fall into the criteria required to get a formal diagnosis. On the other hand, some people do not want to be diagnosed or tagged with the stigma attached to ADHD. These people simply need information processed differently. By doing so, they can navigate life successfully without ever thinking they might have ADHD or another condition.

It is all about the environment.

If the environment works well for certain people, they can thrive throughout their whole life. AI must be treated as an environment as well. The AI environment is created in a specific way to satisfy the general public, but sometimes it needs to be adjusted to work for you as an individual.

To address that, we need to provide some instructions, explanations, and a guide on how our brain operates and what output would be most beneficial for us. The problem is that these instructions vary depending on the task, query, or problem we are trying to solve.

This is where some people on the GitHub repository ayghri/i-have-adhd came up with the idea of providing specific instructions – primarily for Claude Code – to help generate output in an ADHD-friendly way.

This project offers an initial concept that can be adapted and implemented into any other AI tool.

As a primary Gemini user who sometimes feels frustrated with the answers generated by this AI, I decided to implement these ideas to see how much they could improve my own interactions.

Spoiler alert: it really helped!

In Gemini, there are two ways you can implement this:

1. Add it globally as Personalised Context under Your instructions for Gemini
2. Create it as a custom Gem

My Global Instructions

In my personal setup, I have already added a few instructions that I built up based on my experience dealing with this tool. They are:

Make responses in British English

As Gemini – and Google overall – is an American company, there is a lot of American phrasing pushed through that I do not like. For a person living in the United Kingdom, those responses sound odd and not like something you would compose natively.

I am living in the United Kingdom (England).

I followed up with specific location information, as AI sometimes hallucinates details that are not relevant to where I live.

I understand Polish.

Even though I communicate with AI in English by default, it is crucial to add any other native language that you understand. This helps you mix and match responses when needed.

Because Polish can be far more precise than English for AI processing, whenever I struggle to explain something and need a native level of understanding, I throw in an instruction in Polish for better responsiveness.

When you use — in a sentence, between two words, make sure there is a space before and after, so it is not “wordA—wordB” (with em-dash) but “wordA – wordB” (with en-dash). Do not use em dash at all.

One annoying pattern I hate in text generated by AI is the overuse of the — (em-dash), especially joining sentences and words with it without maintaining spaces. Whilst that style may be technically acceptable and used in Oxford-style writing, I personally dislike it. It instantly shouts that the text was written by AI.

To insert an — (em-dash), you need to use a keyboard shortcut like:

• macOS: Press ⌥ Option + ⇧ Shift + - (hyphen)
• Windows: Press ⊞ Win + ⇧ Shift + - (hyphen)

This is not something that most people remember to use. The most common (and partly incorrect) approach is to use a standard - hyphen, as it is natively accessible on the keyboard with a single key press and no key combinations.

Alternatively, there is the – (en-dash), which on a Mac keyboard is displayed prominently on the same key above the hyphen and activated using the ⌥ Option key.

AP Style (Associated Press) – a standard journalistic format – and General British Style (Non-Oxford) utilise spaces between these symbols, which is much more pleasing to the eye and does not stand out unnaturally in text.

Also, because the — (em-dash) is so difficult to remember how to invoke, it is commonly replaced with the far more accessible – (en-dash).

This is why my instruction forces the use of an – (en-dash) with spaces around it, and forbids using the — (em-dash) altogether.

Never use Americanised versions of the English language, unless there is no other alternative.

Despite my previous instruction about using British English, I frequently get words with Americanised spellings, especially using z instead of s. This instruction simply reaffirms that preference.

Moving on…

Here you can add instructions that will make all your chats ADHD-friendly.

More on what this instruction is at a later stage.

I have tried this kind of approach (a global instruction) with another prompt in the past, and I strongly advise against doing so – unless the workflow with a Gem (described later) proves to be “one click too many” for you.

Some time ago, I applied a global prompt designed to make answers more direct at their core, less overly polite, and without unnecessary small talk. The goal was to make the AI sound more like a typical Polish person – which, in many cases, can be interpreted as “rude”.

I did not like it, perhaps because I have lost patience with that style of interaction after living abroad for so long. I removed it altogether and reverted to the default approach.

This is why, for instructions like these, a better approach to trying something new is to start with a Gemini Gem.

My Gem

Gems allow you to create an isolated chat environment with predefined instructions at the start, so you do not need to repeat yourself and you do not risk them getting lost when the context of a discussion grows significantly.

Creating Gems is simple. Just click on Gems from the left-hand menu and proceed with the New Gem button.

Give it a name – for example, Focus Mode.

You do not need to specify a description unless you really want to. On mine, I simply added the note below to remind myself what it is for and where it comes from:

Inspired by https://github.com/ayghri/i-have-adhd – a Gem with a set of instructions to simplify answer modes and prevent overwhelm.

I followed up with The Instruction (more on this shortly), and in the Knowledge (optional) section, I clicked the + button and selected Import Code, pasting the https://github.com/ayghri/i-have-adhd repository link.

Once everything is set, finish by clicking the Save button.

The Instruction

In the ayghri/i-have-adhd repository, inside the skills folder, you will find SKILL.md.

Open SKILL.md in RAW format and copy the entire content after # i-have-adhd into the Instructions field in your Gem.

Your Gem is now ready for interactions.

Remember that this repository will evolve, and SKILL.md will be updated with new features over time. After creation, your Gem can be edited and its instructions updated in the exact same way you created it in the first place. I would recommend updating it from time to time – or watching the repository for changes to assess whether they are significant enough to warrant an update.

Now, by selecting your newly created Focus Mode Gem from the Gems menu, you can start your chat. You will feel the difference straight away.

Usage

After using it for a few days, I found it very useful and effective. However, as mentioned earlier, the output is not suitable for all my needs and workflows. Accessing it within a Gem – and interacting with it when I quickly need to understand something or work on tasks where I do not have time for “nonsense talking” – is extremely handy in specific situations.

Some people will find it useful enough to make it their default setup by applying it as a global instruction, whilst others will not.

Try it out and decide for yourself.

Final Thoughts

At the end of the day, AI tools should adapt to the way our brains work – not force us to bend to theirs. Whether you have ADHD, manage a heavy cognitive load, or simply value concise and direct communication, tweaking your AI environment can transform it from an overwhelming wall of text into a genuinely useful assistant.

Credit & Resource: Special thanks to the contributors behind the ayghri/i-have-adhd GitHub repository for creating and sharing these ADHD-friendly AI prompts.