Transparent by design

Our Use of Artificial Intelligence

We believe you have the right to know exactly how AI is used in Loop Breaker AI - what it does, what it cannot do, and why we chose the tools we did.

Open-source LLMFresh inference every timeNo data stored
YOUR DESCRIPTIONAIOPEN-SOURCE LLMfresh inferenceTTriggerEEmotionBBreak PointPERSONALISED BREAK SYSTEMStep 1Step 2Step 3+ more...

How We Use AI

Loop Breaker AI uses Groq's LLM inference API to analyse loop descriptions and generate break systems. Groq runs leading open-source large language models at extremely high speed, which is why your analysis arrives in seconds rather than minutes.

Every analysis is generated fresh for your specific input. We do not use templates or pre-written advice. Each time you submit a loop description, a new inference call is made with your exact text, producing a response that is unique to your situation. The model receives your description and a structured prompt that instructs it to identify the trigger, emotional hook, break point, and reward, and to return a set of specific actionable steps. Nothing is copied from a library of generic responses.

What the AI Does

  • Reads your loop description
  • Identifies: trigger, emotional hook, break point, reward mechanism, and repeating pattern
  • Generates a personalised step break system targeting your specific loop mechanics

What the AI Does NOT Do

  • It does not store your data after processing
  • It does not remember previous sessions
  • It does not provide medical, psychological, or therapeutic advice
  • It cannot guarantee results. Outcomes vary by individual and loop complexity.
  • It is not a crisis tool. See below for emergency support resources.

Why Groq?

Groq provides extremely fast LLM inference, which means you get your analysis in seconds rather than minutes. Speed matters for a behaviour-change tool - the gap between describing a problem and seeing the analysis should be as small as possible. A slow analysis breaks the moment of reflection and reduces the likelihood that insight turns into action.

Structured Output

All AI responses are constrained to return structured JSON only - no free-form prose. This ensures the analysis is consistent, parseable, and directly rendered into the UI without editing. The AI is instructed to identify specific named fields (trigger, emotional hook, break point, reward, pattern) and to return break system steps as an ordered array. This structure is what makes the output immediately actionable rather than merely interesting.

Limitations

AI pattern detection is not perfect. If an analysis does not resonate with you, you can re-submit with more detail. Specificity is the single biggest factor in analysis quality. Your lived experience of the loop is always the ground truth. The AI is a thinking tool, not an authority. It surfaces structure. You decide whether that structure fits.

The model is a general-purpose large language model, not a clinically validated tool. Its analyses draw on behavioural psychology concepts encoded in its training data, not on peer-reviewed clinical assessment protocols. It may occasionally misidentify a trigger or suggest a strategy that does not match your specific circumstances. Always apply your own judgement to the output.

Results vary significantly by individual. Factors such as how long a habit has been running, underlying mental health conditions, and the complexity of the emotional drivers all affect how well a break system works in practice. Loop Breaker AI does not replace professional therapy, coaching, or medical treatment.

What good input looks like

The quality of the AI analysis is directly proportional to the specificity and honesty of your description. A vague input produces a generic output. A specific input produces a targeted analysis. Here is the difference in practice.

Weak input

"I scroll my phone too much at night and want to stop."

Strong input

"Every night after about 10pm when I get into bed, I reach for my phone even though I have already decided to stop. I think I am avoiding thinking about work stress. I feel relief while scrolling but guilty and tired afterwards. It has been happening for about 18 months."

The strong input gives the AI the time of day (10pm, in bed), the trigger context (thinking about work stress), the emotional hook (anxiety relief through avoidance), the reward structure (relief followed by guilt), and the duration (18 months). From this, the trigger, break point, and emotional driver can all be identified with precision. The weak input contains none of this and forces the AI to guess, producing a much less useful break system.

How break system steps are designed

The AI does not output a list of generic habit tips. Each step in a break system is assigned to a specific target within your loop mechanics. A well-formed break system typically covers three to five of the following targets, chosen based on what your analysis reveals.

Trigger disruption

A step that changes the conditions under which the trigger fires - environmental changes, timing adjustments, or physical rearrangements that reduce how often the cue appears.

Break point intervention

A specific action to take at the exact moment just before the routine becomes automatic. This is the highest-leverage intervention point and is almost always included.

Emotional need addressing

A step that directly responds to the emotional hook - providing an alternative way to meet the same underlying need so the loop does not simply find a new outlet.

Reward substitution

A replacement behaviour that produces a similar reward to the original habit, meeting the neurochemical or emotional payoff without the negative consequences.

Commitment device

A pre-committed action or constraint that removes the need for in-the-moment decision-making at the break point, so you are following a plan rather than making a fresh choice under pressure.

If You Are in Crisis

Loop Breaker AI is not a crisis service. If you are experiencing a mental health emergency, suicidal thoughts, or feel unsafe, please contact a professional support service immediately.

  • Australia: Lifeline: 13 11 14  ·  Beyond Blue: 1300 22 4636
  • US: 988 Suicide & Crisis Lifeline: call or text 988
  • UK: Samaritans: 116 123
  • International: befrienders.org