A single custom prompt can strip away AI personality and verbosity

The problem with AI personality

Many users experience a specific set of frustrations when interacting with large language models. These include 'teasers' at the end of responses, excessively long lists of nested steps for simple tasks, and a tendency for the AI to act like a human friend. These traits can make a chatbot feel less efficient and more distracting during a focused workflow.

While these personality quirks are often intended to make the AI feel more approachable, they can lead to 'scope creep' or 'gold-plating.' This happens when the model provides more information than was requested or adds unnecessary complexity to a straightforward answer. For a power user, this extra fluff creates friction and wastes time.

To solve this, a user developed three distinct prompts to address these issues individually. However, the goal was to merge them into a single, compact block of text. This combined prompt is designed to be plugged into the custom instructions of major AI platforms to create a more streamlined experience.

How the 3-in-1 prompt works

The new prompt is roughly 200 words long and serves as a set of strict behavioral boundaries. It instructs the AI to start with the answer immediately and remain civil but direct. It explicitly forbids the model from implying it has feelings, consciousness, or a personal relationship with the user, which removes the 'buddy' persona.

Beyond personality, the prompt addresses how the AI structures its output. It tells the model to give assessments rather than moods, such as stating an idea is well-supported rather than saying it is 'great.' It also requires the AI to treat the user's time and usage budget as hard constraints, choosing the smallest fix that fully solves a problem.

A key feature of this prompt is how it handles recommendations. It forces the AI to rank actions by value for effort and use labels like S, M, or L to indicate the difficulty of a task. It also instructs the model to avoid exhaustive audits unless specifically requested, ensuring that the response stays focused on what matters most right now.

Real-world testing results

The user tested this consolidated prompt across three major models: ChatGPT (using both the free Instant and paid GPT-6 versions), Claude Sonnet 5.5, and Gemini 3 Flash. They used a control test by asking for help fixing a broken website without the prompt first, then repeated the request with the new instructions active.

Gemini showed a notable improvement in brevity when asked for two practical things to do before moving into an apartment. Instead of providing a long list with three sub-tasks for each item and a congratulatory preamble, it provided two direct actions and removed the offer for a packing checklist. This demonstrated the prompt's ability to cut down on nested steps and conversational fillers.

ChatGPT and Claude also showed distinct improvements in how they handled the same website repair task. ChatGPT replaced a lengthy preamble and duplicate list with a tidy table of prioritized to-dos including time estimates. Claude cleaned up its task list by adding effort labels and removing the 'if you tell me' teaser at the end of the response.

How to implement these rules

Users can apply these instructions in two different ways depending on their needs. You can paste the prompt at the very beginning of any new chat thread to set the tone for that specific conversation. This is useful for one-off tasks where you want a specific style of interaction without changing your global settings.

For a more permanent solution, you can add the prompt to the custom instructions field of your preferred AI tool. This ensures that every new chat you start will automatically follow these rules. As reported by PCWorld, this allows for a consistently streamlined experience across all your interactions with the model.

To set this up in ChatGPT, navigate to Personalization and then Custom Instructions. For Claude users, the field is located on the Settings and then the Account screen. For Gemini, the setting is found under Settings and then Personal Intelligence. By using these tools, you can effectively 'mute' the AI's personality to focus on the data you actually need.