If you search for "SMS relationship analysis" right now, the results split into two completely different worlds.
The first is dating. Tools like Lucen and Mosaic Chats let you paste in a text thread with someone you're seeing romantically and get back a "compatibility score" or a read on whether they're losing interest.[1] The category is real, the tools work, and they target one specific use case.
The second is business SMS — products like Sakari, Verse, and Privy that "use AI" to write better marketing blasts and bump open rates on customer support texts.[2] Different audience, different problem, also a real category.
In between those two — the entire middle, where most of your actual text messaging happens — there's almost nothing. Your mom. Your three closest friends from college. The work-friend who became a real friend after they left the company. Your sister. The group chat from your last job. That's where most of your texting volume lives, and the AI category has barely touched it. This post is about why, and what changes when a tool actually addresses it.
Why the middle has been ignored
SMS relationship analysis as a category exists because text data is uniquely revealing. Years of research on relational communication, including some of John Gottman's foundational work on couples, has shown that small linguistic patterns — response latency, sentiment trends, who initiates, how repair attempts land — are highly predictive of relationship outcomes.[3] That's not folk wisdom. It's measurable.
So why has the AI category mostly chased dating and B2B?
Two reasons. First, dating analysis has clear, urgent intent — someone reading mixed signals at 2am will pay $9.99 to make sense of them right now. B2B has clear, large checks. Both are easy to monetize. Friend-and-family analysis is harder to package because the value is more diffuse: you don't have a specific question on a specific message; you have 200 ongoing relationships and not enough memory to keep up with any of them.
Second, the technical lift is meaningfully harder. Romantic chat analysis can work on a single thread with one person. Friend-and-family analysis only becomes useful when you can see across all of your relationships at once — comparing who you used to talk to weekly and now talk to monthly, which threads have gone dormant, which logistical details from a friend's text last month would be useful context the next time you reply. Wide-context analysis requires sync infrastructure most dating tools never built.
That's the gap.
What "SMS relationship analysis" actually looks like for non-romantic relationships
Stripped of the dating-app framing, here's what AI can usefully do with your everyday text history.
Spot drift before you notice it. You probably know that you and a particular friend "haven't talked in a while," but you usually find out the hard way — by realizing you forgot a birthday, missed a milestone, or had to reply to "long time" with "yeah, sorry, life's been crazy." A frequency-baseline analysis catches the drift while it's still recoverable. You and Devon used to average a thread every 9 days. The current gap is 47 days and growing. No score. No leaderboard. Just the number you'd want to know.
Surface what you've forgotten. This is the highest-value thing AI does with personal text data, and it's the thing dating tools and business SMS tools both skip. When your sister texts you about her kid's school next week, you remember the conversation that day. Three months later, when you're about to call her, you've forgotten the kid's grade, the school's name, and the thing she said she was worried about. AI can pull that back: last time you talked, she mentioned Ben's first day at the new middle school and was nervous about the social transition.
Distinguish logistics from substance. A lot of friend-and-family texting is logistical — what time, where, did you grab the thing — and the substantive moments are scattered across years of those threads. Pattern analysis can separate the two and give you a quiet running record of what people have actually been telling you about their lives.
Make group threads legible. Group chats are the relationships most people have the worst memory for. Who said what, who's been quiet, what plan never actually happened. AI can summarize a group thread the way you wish you could remember it, without you having to scroll back six months. (Background: Every Plan You Made Over Text That Never Actually Happened.)
Notice tonal shifts. When a relationship is in trouble — the friend who's been distant, the parent who's been less responsive — there are usually small tonal markers in the text history weeks before you'd consciously register the change. Sentiment trend analysis isn't a substitute for actually paying attention, but it's a backstop for when you're not paying attention.
For a longer take on what's actually inferable from years of personal text history: What Your Texts Actually Reveal About Your Friendships and I Let AI Analyze 90 Days of My Texts. Here's What I Learned..
How Amicai approaches it
Amicai is the only AI we know of building this for the middle category. The architecture, in plain language:
- Local sync first. On macOS, Amicai Sync is a small menu bar app that reads iMessage and FaceTime history from the local SQLite databases on your Mac. On Android, the Android app reads SMS, MMS, and notification streams directly. Nothing leaves your machine until you've explicitly granted access, and the data path goes through your account, not through the Amicai team's servers.
- Anonymization before AI. Sensitive content is masked or stripped before any prompt is built. Phone numbers, email addresses, and payment details never end up in the model's input. (Why this matters in practice: Is My AI Chatbot Data Safe? Here's How to Tell.)
- Per-contact intelligence, not just per-thread. The unit of analysis is the relationship, not a single conversation. So when you ask the chat agent about a specific person, the context spans every channel you've talked to them on — text, calls, calendar, journal entries — not just the most recent thread.
- Daily reflection over real-time alerting. No notifications mid-conversation. No "your friend's energy seems off, send a heart emoji" prompts. The output is a quiet morning summary of who you've been in touch with, what came up that's worth noticing, and what's drifting.
- Sensitive contacts. Anyone you flag as off-limits is excluded from analysis everywhere. Their messages don't get processed, their name doesn't appear in prompts, their data is purged across the whole system. (Background: Some Conversations Are Off Limits. Your AI Should Know That..)
What we deliberately don't do
The reason most existing SMS analysis tools feel off — even when they're technically well-built — is the framing. So a few things Amicai explicitly avoids:
- No compatibility scores. Friendships and family relationships are not zero-sum. Scoring them is the surest way to make them feel transactional. We've written about why at length: Amicai Isn't a CRM for Your Friends.
- No "AI relationship coach" voice. The chat agent doesn't tell you what to do. It surfaces what you already knew but had forgotten, and lets you decide.
- No engagement loop. Text-message analysis tools are tempting to gamify (streaks, badges, "your friendship score went up this week"). All of that pulls you back into the app and away from the actual relationship. The right success metric is texts sent in your voice to people you care about, not minutes spent in our product.
The bar for tools in this space
If you're evaluating any SMS or text-analysis tool that isn't strictly for dating or business, the questions worth asking:
- Does it sync across your real channels (iMessage, WhatsApp, SMS, FaceTime), or does it require you to copy-paste a thread?
- Does it strip sensitive content before any AI sees it?
- Does it organize by person, not just by thread?
- Can you flag specific contacts as off-limits, and is that flag honored everywhere?
- Is there a "score" anywhere in the UI? (If yes, that's the wrong product.)
The thing the dating tools and the B2B tools both miss is the same thing: text history is most useful as a memory aid for the relationships you already have, not a diagnostic tool for ones you're trying to fix or sell into. The middle has been waiting for a tool that takes that seriously.
References
[1] Lucen. "Lucen: Chat Analyzer & AI Dating Coach." Lucen, 2026.
[2] Sakari. "Artificial Intelligence Texting: How AI Is Changing Business SMS in 2026." Sakari, 2026.
[3] MosaicAI Research. "How AI Analyzes Relationship Patterns in Your Chat Messages." Citing John Gottman's research on relational communication patterns and AI's ability to detect them.