About
How the analysis works
BareFlag reads conversations the way a trained eye might read a poem. It looks for specific patterns that psychology research has documented. Here's exactly how.
- Extracts and parses the text (using OCR for screenshots)
- Evaluates the text against twelve research-based construct definitions
- Identifies specific linguistic markers that match each construct
- Assigns a confidence level and overall risk tier
- Returns citations to the underlying research
The twelve constructs
Gaslighting
Reality denial, trivialization, and discrediting that makes someone question their own memory and perception.
Coercive Control
Isolation tactics, microregulation, economic control, degradation, and intimidation that restrict autonomy.
DARVO
A three-stage sequence: Deny, Attack, Reverse Victim and Offender. Used to deflect accountability.
Love-Bombing
Excessive affection and grand declarations disproportionate to the relationship stage, designed to create rapid attachment.
Intermittent Reinforcement
Unpredictable alternation between warmth and cruelty that creates addictive attachment patterns.
Future-Faking
Elaborate promises about a shared future that are disproportionate to the relationship or contradict actual behavior.
Breadcrumbing
Sporadic, non-committal messages designed to maintain interest without genuine commitment.
Negging
Backhanded compliments or strategic insults that undermine confidence while maintaining plausible deniability.
Narcissistic Abuse Cycle
Phases of idealization, devaluation, discard, and hoovering that form a repeating pattern of emotional manipulation.
Dark Triad Linguistic Markers
Language patterns associated with grandiose narcissism, Machiavellianism, and psychopathy traits.
Romance Scam Indicators
Scripted messages, fabricated emergencies, financial requests, and identity inconsistencies.
Grooming Patterns
Gradual trust-building, boundary-pushing, secrecy creation, and isolation from support systems.
What the risk tiers mean
Low
No significant indicators detected. Communication appears within normal range.
Medium
Some concerning patterns warrant reflection and monitoring over time.
High
Multiple strong indicators of manipulation tactics across the conversation.
Limitations
- Analysis reflects patterns in the submitted text only, not the full relationship context
- Normal conflict and manipulation can look similar in isolated messages
- The AI can make mistakes, especially on ambiguous content
- Cultural and neurodivergent communication styles may not be well-represented in the underlying research
- This isn't therapy, diagnosis, or legal evidence
What to do with results
- A high tier isn't a call to action. It's information for reflection.
- Consider sharing the analysis with a trusted friend, therapist, or DV advocate
- If you're in danger, prioritize your safety using the resources below
Support
If you'd like to talk to someone
National Domestic Violence Hotline. Twenty-four hours. Call 1-800-799-7233.
Crisis Text Line. Text HOME to 741741.
thehotline.org. Online chat available.