Clear insights, not vague advice
Every ChatMend feature is grounded in your conversation data and backed by attachment-informed research. Here is everything it surfaces.
Attachment Style Read
See patterns that may suggest secure, anxious, avoidant, or fearful-avoidant tendencies for each person — presented as signals, not diagnoses.
Relationship Pattern Map
Visualize pursue-withdraw cycles, emotional bids, and response patterns that define the dynamic between two people.
Emotional Triggers
Identify the moments where tone shifts — which topics cause distancing, defensiveness, or shutdown in the conversation.
Repair Capacity
See how often repair is attempted after conflict, who initiates it, how long disconnection lasts, and whether the relationship returns to safety.
Evidence-Based Insights
Every finding links to specific text patterns. No vague interpretations — see the evidence behind each signal.
What To Do Next
Get practical, grounded conversation guidance based on the actual pattern — not generic relationship advice.
What you get back
A structured, readable report — not a wall of text or a single mysterious score.
Relationship health read
An at-a-glance grade with emotional tone and overall state, backed by a plain-language summary.
Attachment map
A two-person radar across secure, anxious, avoidant, and fearful axes — so you can see the dynamic, not just labels.
The core pattern
The recurring cycle the conversation keeps returning to, broken into the steps that drive it.
Pattern evidence
Specific, quoted examples that show where each finding comes from in the thread.
Risks and strengths
What to watch for, rated by severity, alongside what is genuinely working in the relationship.
Timeline
How the dynamic changed over the span of the conversation, period by period.
Next steps and scripts
Tailored suggestions and a way to open the conversation — grounded in the actual pattern.
Ask follow-up questions
Go deeper on any part of your report in plain language, with answers tied to your conversation.
Not another compatibility quiz
Most relationship apps gamify your love life. ChatMend gives you evidence-based pattern analysis.
Why attachment, and not the personality quiz you already know
Most tools that promise to explain your relationship are built on a framework that was never designed to describe one. Here is how the common ones are regarded by the people who study them, and where ChatMend sits.
ECR-R
Experiences in Close Relationships — Revised
The most widely used self-report attachment measure in relationship research, and the one ChatMend's self-test is built on.
Measures
Attachment in close relationships, on two independent dimensions: attachment-related anxiety and attachment-related avoidance.
Shape of the answer
Two continuous scales. No types.
Built by item-response-theory analysis of existing attachment scales specifically to fix their measurement precision, and reported with high internal consistency across both dimensions.
Fraley, Waller & Brennan (2000), Journal of Personality and Social Psychology 78(2), 350–365
For reading a relationship: Designed for exactly this question. It is also relationship-specific — the same person can score differently with different partners, which is the whole premise of reading one conversation at a time.
AAI
Adult Attachment Interview
Widely treated as the reference standard for adult attachment — and almost impossible to scale.
Measures
State of mind with respect to attachment, coded from how coherently someone narrates their own attachment history.
Shape of the answer
Trained-coder classification.
Requires a roughly hour-long interview, transcribed, coded by someone who has completed a certified training. It measures discourse coherence rather than what a person says they feel, which is why it can diverge from self-report.
George, Kaplan & Main (1985; 3rd ed. 1996), University of California, Berkeley
For reading a relationship: The best instrument here, and unavailable to almost everyone. It cannot be self-administered, cannot be automated, and costs what a trained clinician costs.
Big Five
Five-Factor Model
The consensus model of personality structure in academic psychology, and the framework MBTI is usually re-analysed against.
Measures
General personality across five broad traits: openness, conscientiousness, extraversion, agreeableness, neuroticism.
Shape of the answer
Five continuous scales.
When MBTI responses were analysed against the five-factor model, four of the MBTI scales mapped onto four of the Big Five — the model was measuring recognisable traits, but as arbitrary dichotomies rather than as the continua they actually are.
McCrae & Costa (1989), Journal of Personality 57(1), 17–40
For reading a relationship: Sound, but aimed at the wrong target. It describes who you are in general, not how you behave with one particular person under stress — and it is the second of those that a conversation shows you.
MBTI
Myers-Briggs Type Indicator
Enormously popular, and long criticised by psychometricians on two specific grounds: the types are not stable, and the cut-points are arbitrary.
Measures
Assigns one of sixteen types from four either/or pairs (E/I, S/N, T/F, J/P).
Shape of the answer
Sixteen discrete categories.
Pittenger's review reports that a substantial share of people are assigned a different type when they retake it weeks later, and that scores on each dimension are distributed continuously rather than in two clumps — so the midpoint split that creates the type is a decision, not a finding.
Pittenger (2005), Consulting Psychology Journal: Practice and Research 57(3), 210–221
For reading a relationship: Not built to describe relationships and not stable enough to track change in one. If a measure can call you a different person after five weeks, it cannot tell you whether anything about your relationship has shifted.
Love Languages
The Five Love Languages
A useful vocabulary that outran its evidence. Relationship researchers reviewing it found the central claim unsupported.
Measures
Sorts people into a preferred way of giving and receiving affection.
Shape of the answer
Five discrete categories.
A review in relationship science concluded that the evidence does not support the three assumptions the framework rests on — including its core prediction that couples do better when partners match each other's language.
Impett, Park & Muise (2024), Current Directions in Psychological Science 33(2)
For reading a relationship: Good for starting a conversation about what someone wants. Not a measurement, and it makes no claim about conflict, repair, or what happens under threat.
Enneagram
Widely used in coaching and workplace settings; it does not have a research base comparable to the measures above.
Measures
Assigns one of nine personality types.
Shape of the answer
Nine discrete categories.
We could not find peer-reviewed validation of the nine-type structure in mainstream personality journals. That is an absence of evidence rather than evidence against it, and we are stating it as such.
No supporting publication identified
For reading a relationship: Nothing in it is designed to measure attachment, conflict behaviour, or change over time.
What ChatMend uses
OursFor the self-test
ECR-R
The attachment self-test in the ChatMend library is the ECR-R, scored on its two published dimensions — including the reverse-scored items, which is where most free online copies of it go wrong.
For reading a conversation
Two-dimensional attachment reading, plus Gottman conflict markers
The conversation engine reports the same two dimensions the ECR-R uses — anxiety and avoidance — as continuous scores rather than a type, alongside the four conflict behaviours Gottman's lab identified: criticism, contempt, defensiveness and stonewalling.
Attachment is the framework here because it is the one built to describe how people behave with each other under stress, rather than what they are like in general — and because it is measured on continua, which means it can show movement. A type cannot get better. A score can.
Where this stops
One thing this page should be straight about: reading a conversation is not the same as administering a validated questionnaire. The ECR-R has decades of psychometric work behind it. ChatMend's conversation analysis borrows its structure — two continuous dimensions instead of a type — but it is an AI reading your messages, not an instrument with published reliability figures. That is why every report carries a confidence level bounded by how much of the conversation was read, why the language is always "shows patterns associated with" rather than "you are", and why the self-test exists separately for anyone who wants the actual measure.
Sources last checked 25 August 2026. ChatMend is not a diagnostic tool and does not provide medical or psychological advice.
Everyday tools beyond the analysis
Practical tools to help you communicate better — on the roadmap.
Ask ChatMend
Ask follow-up questions about your analysis. Understand specific patterns in plain language.
Tone Check
Paste a message before you send it. Get feedback on how it may land based on the relationship pattern.
Repair Message Builder
Get a suggested repair message after conflict — tailored to the dynamic, not a generic template.
Boundary Builder
Draft a boundary statement grounded in the evidence from your conversation history.
Put these features to work
Bring a conversation and get an evidence-based read on the pattern.