8 Essential Features of
a Chat Analysis Tool
Chat analyzers range from emoji counters to tools that read the emotional arc of a relationship. These are the eight capabilities that separate the second kind from the first, with an honest account of how MosaicChats handles each.
Key takeaways
- A message counter tells you who talks more. A real analysis tool tells you how the relationship is moving.
- The core four: a sentiment timeline, engagement statistics, personality and attachment estimates, and a compatibility score.
- The features compound. A score without an explanation is a number; an assistant with the data in front of it is advice.
- Multi-platform support matters because relationships rarely live on one app.
- Personality-from-text is roughly 76% accurate for full MBTI type in the best published test. Useful, not gospel.
Most free chat analyzers do one thing: count. Messages, emoji, active hours, longest streak. That is fun for a group chat and useless for the question people actually bring to these tools, which is whether a relationship is getting warmer or colder and why. Answering that takes a specific set of capabilities.
Why the feature set matters
The value is in the combination. Sentiment without timing cannot tell you whether a cold week was a fight or a work trip. Personality without compatibility is trivia. A compatibility score without an assistant is a number you will argue about. The eight features below are the set that lets each one give the others context.
1. A sentiment timeline
What it does
Scores the emotional tone of each message and plots it over time, per person. The line, not any single point, is the reading: is tone stable, rising or sliding, and how fast does it recover after a dip?
Look for: weekly trend lines, separate lines per participant, and enough history to see months.
In MosaicChats: the sentiment timeline plots each person week by week across the whole export, so a cooling-off period shows as a visible dip.
Modern models handle sarcasm and mixed feelings far better than the keyword dictionaries of a decade ago; a 2024 review of NLP sentiment analysis covers how far the field has come.
2. Personality and attachment estimates
What it does
Reads how each person writes, vocabulary, structure, emotional language, question-asking, and estimates personality type, attachment style and love language from it. The value is less the label than the contrast: why one of you writes paragraphs and the other sends "ok."
Look for: a per-axis breakdown rather than a single four-letter verdict, and an attachment-style read alongside it.
In MosaicChats: an MBTI tile with each axis scored for each person, plus an attachment style report and a love language analysis.
Hold the estimate loosely. A 2024 study found GPT-4 classified full MBTI type correctly for 76% of participants, and a 2025 paper describes ChatGPT-4's trait estimates from text as moderately but significantly accurate. Our guide to MBTI from texts explains what the signals are.
3. Engagement statistics
What it does
The counted layer: reply-time distributions, who initiates, message volume over time, and an activity heatmap of which hours and days the conversation lives in. These are exact, which makes them the right place to start reading.
Look for: distributions rather than a single average, and an initiation split.
In MosaicChats: the engagement tile shows reply-time patterns, initiation balance, message volume and the activity heatmap for both people.
How to read reply times without over-reading them is covered in the psychology of response time.
4. A compatibility score
What it does
Summarises how the two communication styles fit: pace, emotional expressiveness, balance, and how each of you handles friction. A single number is easy to over-trust, so the breakdown behind it matters more than the headline.
Look for: a stated scale and a per-dimension explanation.
In MosaicChats: a 0–100 compatibility score with the reasoning behind it, drawn from the conversation rather than a questionnaire.
For what compatibility scoring can and cannot claim, see the science behind AI compatibility.
5. A word cloud
What it does
The simplest feature and still one of the most revealing: what the relationship is about. A cloud dominated by logistics reads differently from one full of names, plans and feelings.
Look for: per-person clouds so you can compare vocabularies.
In MosaicChats: a word cloud tile sized by frequency, so the conversation's recurring themes are visible at a glance.
6. An assistant that has read the analysis
What it does
Charts raise questions. An assistant with the analysis in front of it can answer them: what a dip usually means, how to bring initiation into balance, whether a pattern is worth raising. This is the difference between a generic chatbot and advice about your relationship.
Look for: access to your actual results, and a choice of tone.
In MosaicChats: Myrah answers with your analysis as context, in one of four voices: The Coach, The Caring Friend, The Roaster, or The Gay Bestie.
Why an assistant with context beats a general-purpose chatbot for this is covered in our guide to choosing an AI for relationship advice.
7. The extras: recommendations, clips and artwork
What it does
Not analysis, but things you can only make from an analysis: gift and media suggestions drawn from what you actually talk about, the moments in the thread worth remembering, and a picture of the relationship.
In MosaicChats: song, film, book and gift recommendations from the conversation's themes, automatically detected memorable clips, and an AI-generated artwork for the relationship.
8. Multi-platform support
What it does
Relationships rarely live on one app. A tool that only reads WhatsApp misses the Instagram thread where it started. Format detection should be automatic; you should never have to convert an export by hand.
Look for: the platforms you actually use, and clear export instructions for each.
In MosaicChats: WhatsApp, iMessage, Instagram, Messenger, Telegram and Snapchat exports, detected automatically, or a pasted conversation. Export steps for every platform are in the step-by-step guide.
How the tiers compare
Free web counters
Message and emoji counts, word frequency, an activity timeline. Good for group-chat leaderboards. No sentiment, no personality, no explanation.
Sentiment-only tools
Emotional tone over time, and nothing about who initiates, how fast anyone replies, or how the two styles fit.
General chatbots
Reasonable general advice, no timing statistics, and no memory of your conversation beyond what you paste in.
MosaicChats
All eight of the above in one dashboard, from a single upload. For a wider survey of alternatives, see our comparison of chat analysis tools.
See all eight on your own conversation
Export a chat from any supported platform, upload it, and read the timeline, statistics, personality and compatibility tiles side by side.
Choose a tool by the question you are asking. If it is "who texts more," anything will do. If it is "what is happening to us," you need the timeline, the statistics, the style read and something that can explain them together.
Frequently asked questions
Do I need all eight features?
No. If your question is 'who texts more', a free counter answers it. If your question is 'is this relationship getting warmer or colder', you need a sentiment timeline and engagement statistics at minimum. The features compound: personality estimates mean more next to a compatibility score, and a score means more with an assistant to explain it.
How accurate is personality detection from text?
Better than chance, short of certainty. A 2024 study found GPT-4 classified full MBTI type correctly for 76% of participants; individual axes such as introversion versus extraversion are easier. Treat any personality readout as a description of how someone writes, not a diagnosis.
Which platforms does MosaicChats support?
Exported chats from WhatsApp, iMessage, Instagram, Messenger, Telegram and Snapchat, with format detected automatically. You can also paste a conversation directly.
Is there a free plan?
Yes. Analyzing chats is free (up to 50 uploads a week). The free plan includes five Myrah messages a week; a Pro plan raises that to 200 and lets Myrah read your full chat.
Related articles
References & Sources
- "Recent advancements and challenges of NLP-based sentiment analysis: A state-of-the-art review." Natural Language Processing Journal, 2024.Source
- "Artificial Intelligence and Personality: Large Language Models' Ability to Predict Personality Type." SAGE Journals, 2024.Source
- "On the emergent capabilities of ChatGPT 4 to estimate personality traits." Frontiers in Artificial Intelligence, 2025.Source