I text my friend Marcus almost every day. Quick stuff — links, jokes, complaints about work. If you asked me two months ago, I'd have said we were close. Tight, even.
Then I looked at the actual data. Three months of our conversation history, analyzed for patterns. Here's what I found: 94% of our messages were logistical or reactive. Links forwarded without comment. One-word replies. The conversational equivalent of a head nod across a crowded room. We'd been texting constantly and saying almost nothing.
Meanwhile, my friend Sarah — who I text maybe once every two weeks — had conversations where 60% of the messages were substantive. Real questions. Real answers. Updates about things that actually mattered. Less frequent, but deeper by every measure.
I had the relationship backwards. The person I texted most wasn't my closest friend. They were just my most convenient one.
Your Texts Are a Dataset You've Never Looked At
You've sent thousands of text messages over the past year. Each one contains a tiny signal about the state of a relationship — who initiates, who responds, how quickly, how much effort goes into the reply, whether the conversation goes anywhere or just... stops.
Individually, none of these signals mean much. But in aggregate? They paint a picture that's often very different from the one in your head.
Research from Brigham Young University found that texting patterns — frequency, response times, and message length — predict relationship satisfaction with significant accuracy.[1] A study on linguistic style matching showed that how closely two people mirror each other's texting style correlates with their relational chemistry and long-term compatibility.[2]
This isn't just about romantic relationships. The same patterns show up in friendships — they're just harder to see because nobody's paying attention.
Five Patterns Hiding in Your Messages
After spending months analyzing my own conversation data (and talking to researchers who study digital communication), here are the patterns that revealed the most about my relationships:
The Initiation Ratio. Who texts first? In a healthy friendship, initiation roughly alternates. When one person starts 80% of the conversations, it usually means one of two things: the other person is going through something, or the relationship has become one-directional. I found three friendships where I initiated every single conversation for 90 days straight. Not because my friends didn't care — but because the pattern had become invisible to both of us.
The Response Depth Gradient. Compare the length and substance of your messages to theirs — and vice versa. When there's a persistent mismatch (you send paragraphs, they send "haha nice"), it doesn't necessarily mean the relationship is bad. Some people are just brief texters. But if the depth mismatch is new — if they used to match your energy and now they don't — that's a signal worth noticing.
The Conversation Completion Rate. How many of your conversations actually go somewhere versus fizzle out? A thread that's all one-liners and reactions is different from one where someone asks a follow-up question. I tracked this across my top 15 contacts and found that the friendships I rated as "strongest" had a conversation completion rate above 70%. The ones I rated "weakest"? Below 30%.
The Topic Ratio. What percentage of your conversations are logistical (scheduling, links, memes) versus personal (feelings, updates, questions about each other's lives)? There's no right ratio, but a shift matters. If a friendship that used to be 50/50 is now 90% logistics, the emotional bandwidth of the relationship has narrowed — even if the texting frequency hasn't changed.
The Gap Pattern. Not just how long the gaps between conversations are, but how they've changed over time. A friendship with a consistent 2-week rhythm is healthy. A friendship where the gap went from 3 days to 10 days to 3 weeks to 6 weeks? That's a drift in progress. And it's happening so slowly that neither person notices until it's been 4 months.
The Mirror Problem
Here's the thing that caught me off guard: the patterns didn't just reveal which friendships were fading. They revealed where I was the problem.
I'd been telling myself that certain friends had "gotten distant." The data showed a different story. In two cases, they were still reaching out at the same frequency — I was the one whose replies had gotten shorter and less frequent. My perception was that they'd pulled away. The reality was that I had.
Psychologists call this the "actor-observer asymmetry" — we attribute our own behavior to circumstances ("I've been busy") and other people's behavior to character ("they don't prioritize me").[3] Your text history is one of the few places where you can actually fact-check that story.
It's uncomfortable. The gap between who you think you are as a friend and who your data says you are — that's not a fun discovery. But it's a useful one. Because you can't fix a pattern you can't see.
What AI Can Actually See That You Can't
Here's where this gets interesting. A human reading through 3 months of text conversations would need days. They'd get bored, lose context, miss patterns. The signal-to-noise ratio in text messages is brutal — you'd have to wade through hundreds of "sounds good" and "lol" messages to find the moments that actually matter.
AI is uniquely good at this. Not because it understands relationships better than you do — it doesn't. But because it can process the volume. It can tell you that your response time to one friend has tripled over the past month. That you used to ask questions and now you mostly react. That someone mentioned something important 3 weeks ago and you never followed up.
These aren't insights that require emotional intelligence. They're pattern recognition at a scale that human attention can't match. The emotional intelligence part — deciding what to do with those patterns — that's still entirely yours.
What AI surfaces from your conversations isn't surveillance. It's a mirror. The kind of mirror that shows you the relationship patterns your brain is too busy to track and too biased to see clearly.
Start Noticing
You don't need AI to start paying attention to your texting patterns. Open your messages app right now and look at your recent conversations.
Who are you texting the most? Is that the person you'd say you're closest to — or just the person you happen to exchange the most logistics with?
Who used to text you regularly and doesn't anymore? When did that change?
When was the last time you asked someone a real question — not "what's up" but something that shows you remember what's happening in their life?
Your texts contain years of relationship data you've never examined. The patterns are already there. Most of us just never look.
References
[1] Brigham Young University. "Texting, Technoference, and Relationship Satisfaction." BYU Scholars Archive, 2016.
[2] James Pennebaker et al. "Language Style Matching Predicts Relationship Initiation and Stability." Psychological Science, 2011.
[3] Edward Jones and Richard Nisbett. "The Actor and the Observer: Divergent Perceptions of the Causes of Behavior." Attribution: Perceiving the Causes of Behavior, 1972.