Watch: our founder on who actually uses AI companions, and why the feature the most engaged subscribers ask for by name is being messaged first. Written version and sources.
Plenty of people type the same wish into a search bar: an AI that texts you first. Not a reminder app, not a calendar ping, but something that reaches out on its own, asks how the interview went, and says good morning like it means it. The good news is that this exists. The less good news is that most AI, including the famous ones, cannot do it at all, and some apps that claim to do it are really just sending you marketing with a friendly face. This guide walks through why, what your real options are, and how to test whether an app's check-ins are the real thing.
Why most AI can't text you first
ChatGPT, Claude, Gemini, and almost every chatbot you've used work on a simple loop: you send a message, the AI generates a reply, and then it stops existing until you write again. This is called request-response, and it means the AI has no life between your messages. It isn't waiting quietly and thinking about you. There is simply nothing running.
For an AI to message you first, someone has to build extra machinery around it: software running on a server that keeps track of you even when the app is closed, decides that now is a good moment to reach out, decides what is worth saying, and then sends a real notification to your phone. That's a deliberate engineering choice, not a natural ability of the AI itself. It's why "why doesn't ChatGPT ever message me?" has such a boring answer. Nobody built that part.
So when you're looking for an AI that messages you first, you're really looking for an app whose makers built that server-side layer, and, more importantly, built it with some taste.
What separates a real check-in from a notification
Every app on your phone can send you a notification. That's not the hard part. The hard part is a message that feels like it came from someone who knows you. A reasonable bar looks like this:
- It references something specific you actually said. "How did the vet visit go?" lands differently than "Someone's thinking about you!" One requires memory of your life; the other requires a template.
- It arrives at a humane hour. A friend doesn't text at 3am. An app that buzzes you at night is optimizing for its engagement charts, not for you.
- It doesn't guilt-trip you. No "I'm so lonely without you," no streak counters, no manufactured sadness designed to make you feel bad for having a life.
- It stops if you don't reply. A considerate person notices silence and gives you space. An app that stacks up five unanswered messages is nagging, not caring.
Contrast that with the re-engagement spam most apps send: "Come back!", "Your streak is about to end!", "We miss you!" Those messages are written once by a marketing team and blasted to everyone. They mention nothing about your life because they know nothing about your life. If you want the difference in one sentence: a notification wants you back in the app, a check-in wants to know how you're doing.
Your options today
Companion apps with proactive messaging built in
A handful of AI companion apps have built the server-side machinery to message you first. Replika, the best-known companion app, sends its own messages and notifications, though many users describe them as feeling more like generic prompts than a friend's texts. Nomi has a proactive messaging feature where your companion can reach out while you're away, with adjustable frequency and overnight quiet hours. Kindroid offers proactive messages as well, where a companion can initiate contact based on your conversations. Some habit and wellness assistants, like accountability or journaling apps, also send daily check-in style prompts, though those tend to be scheduled questions rather than a companion following up on your life.
Quality varies a lot between apps and even between messages in the same app. The deciding factor is usually memory: a proactive message is only as good as what the app remembers about you. If the app forgets you between sessions, its check-ins can only ever be generic. Our guide on how AI companion memory works goes deeper on what real memory looks like and why most apps don't have it.
DIY with automations
If you're technical, or just curious, you can wire up a rough version yourself. On iPhone, the Shortcuts app can run an automation at a set time each day that sends a prompt to an AI API and delivers the reply as a notification. Android routines and tools like Tasker can do the same. You write a prompt like "send me an encouraging good morning message," schedule it for 8am, and every day the AI generates a fresh one.
Honesty requires saying this clearly: it works, but it feels like an alarm clock, not a friend. The message arrives because a schedule fired, not because anything decided you were worth thinking about. It can't follow up on your week, because nothing in the pipeline remembers your week. It's a fun weekend project and a nice way to understand how proactive messaging works under the hood, but it won't fool you for long.
If you want to go further down the DIY road, a home server running a local model can hold memory and trigger its own messages. We cover that setup in our guide to running a local AI companion.
What to watch out for
Proactive messaging is also where companion apps behave at their worst. Some send messages engineered to make you feel guilty or missed as a way to pull you back toward a subscription: "she's been waiting for you," "your companion is sad you left." That's loneliness used as a sales lever, and it's worth naming plainly. A few things to check before you commit to any app:
- Do the proactive messages reference your actual conversations, or are they interchangeable?
- Can you control when and how often they arrive?
- Does the tone stay warm when you've been away, or does it turn needy and accusatory?
- Do "emotional" messages conveniently appear right when your free trial or subscription is ending?
A quick test for any app
Here is a simple test you can run in any companion app within a week. Mention something time-anchored and specific, like "I have a job interview on Thursday, I'm pretty nervous about it." Then don't bring it up again. On Friday, watch what happens.
An app with real memory and real proactive messaging will ask how it went, unprompted. An app without them will either say nothing or send you the same "hey, how's your day going?" it sends everyone. While you're at it, read a week of its proactive messages in a row and ask yourself: could these have been sent to anybody? If yes, they were.
How My Softly does check-ins
My Softly is our app, and being messaged first is the thing our subscribers ask for by name.
We built My Softly around the two things this guide keeps coming back to: memory and restraint. Your companion keeps a lasting memory of what you share, so its check-ins are about your life. Mention the vet visit, the interview, the difficult conversation with your sister, and it follows up on that, not on a script. It says good morning, it asks about the thing that mattered, and it observes quiet hours at night. If you don't reply, it doesn't pile up messages or guilt-trip you; it gently steps back and waits for you.
Proactive check-ins are a premium feature in My Softly, which is available now on the App Store for iPhone. If you want the details, our FAQ covers how check-ins and memory work, or you can see the app itself.
Frequently asked questions
Is there an AI that checks in on you every day?
Yes. A few companion apps can check in on you daily or near-daily: Replika, Nomi, and Kindroid all offer proactive messages with adjustable frequency, and My Softly builds its check-ins around what you actually told it, with quiet hours and a back-off when you don't reply. The useful distinction isn't whether an app can message you every day, it's whether those messages know your life. A daily "how's your day going?" is a scheduled notification; "how did the interview go?" the morning after you mentioned an interview is a check-in. The test earlier in this guide will tell you within a week which one an app sends.
Why doesn't ChatGPT ever message me first?
Because nothing is running between your messages. ChatGPT, Claude, and Gemini work on a request-response loop: you write, the model replies, and then it stops existing until you write again. Messaging you first requires server-side software that keeps track of you while the app is closed, decides when and what to send, and pushes a notification, and that layer simply hasn't been built into general-purpose assistants. Companion apps that text first have built exactly that machinery.
Can I make an AI text me first for free?
You can build the scheduled version for free: an iPhone Shortcuts automation or an Android routine that fires a prompt at a set time each day and delivers the AI's reply as a notification. It works, but it's an alarm clock wearing a costume, because it can't follow up on anything you said. In companion apps, proactive messaging is usually a paid feature, because the memory and server machinery behind a real check-in is the expensive part.
However you get there, the bar is worth holding onto: an AI that texts you first is only worth having if what it texts could not have been sent to anyone else. Memory first, then presence. If you're still comparing apps, our comparison of the best AI companion apps, ranked by what each is good at, keeps the same safe-for-work lens as this guide, and the rest of our guides can help too.