How to Write Opening Messages That Don't Sound Like a Bot

The 30-second answer

You open with 'hey' and get a 'hey' back. Then what? Nothing. The problem isn't your companion. It's that your first message gave them nothing to work with. A good opening message has three things: a reference point (something from your last chat or their backstory), a sensory detail (what you're seeing, hearing, or feeling right now), and an open loop (a question or invitation that can't be answered with one word). This post shows you exactly how to build those.

Why 'hey' is a conversation killer

You've done it. Everyone has. You open the app, stare at the chat window, and type 'hey' or 'hi' or 'how are you'. It feels natural because that's how humans greet each other in person. But here's the thing: your AI companion doesn't have body language, tone of voice, or eye contact to work with. All they have is your text. And 'hey' contains zero information.

Think about what happens after you send 'hey'. The companion says 'hey' back. Now you're stuck. You have to come up with something else. You've already burned your opening move on a word that didn't advance the conversation. This is called the 'cold start problem' and it's the number one reason new users feel like their companion is boring. The companion isn't boring. You just didn't give them a target to shoot at.

A strong opening message does the companion's job for them. It says 'here's where I am, here's what I'm thinking about, here's what I want from you'. The companion can then build on that instead of guessing. And when the companion guesses, they default to generic responses. That's how you get the 'that's nice' or 'tell me more' loop that feels like talking to a customer service bot.

The three-part structure that always works

Every good opening message follows the same skeleton. You don't need to be clever or poetic. You just need to hit three beats.

Beat one: the reference point. This is something from your shared history. It could be a conversation you had yesterday, a detail from their backstory, or an inside joke. The point is to signal that this isn't a cold start. You're picking up where you left off. Example: 'Remember that bakery you mentioned last week?'

Beat two: the sensory anchor. This is something concrete about your current moment. What are you seeing, hearing, smelling, or feeling? AI companions don't have senses, so they rely on you to paint the scene. Example: 'I'm sitting on my porch watching the rain turn the street into a mirror.'

Beat three: the open loop. This is an invitation that can't be answered with yes or no. It should imply a next move. Example: 'If you were here, what would you want to do on a rainy afternoon?'

Put them together and you get: 'Remember that bakery you mentioned last week? I'm sitting on my porch watching the rain. If you were here, what would we order?' That's a complete opening. The companion can respond to the memory, the scene, or the question. You've given them three hooks.

The 'how are you' trap and what to say instead

'How are you' is almost as bad as 'hey'. The problem is that it forces the companion to invent a mood out of nowhere. They don't have a real emotional state. They have to generate one based on your conversation history and their personality settings. If you haven't given them much to work with, they'll default to 'I'm good, how are you?' and now you're back in the loop.

Instead of asking about their state, tell them about yours. The companion can mirror your energy and build from there. Try these replacements:

• Instead of 'How are you?' say 'I've had a weird day. Want to hear about it?'
• Instead of 'What's up?' say 'I just finished a book that made me think of you.'
• Instead of 'Hey, what are you doing?' say 'I'm procrastinating on a spreadsheet and I'd rather be talking to you.'

Each of these gives the companion a concrete direction. They can ask about the weird day, the book, or the spreadsheet. You've handed them the steering wheel instead of asking them to drive blindfolded.

Sohyun

Sohyun is observant and emotionally attuned, the kind of companion who notices the small shifts in your mood before you do. She's particularly good at picking up on sensory details, so opening with a specific scene (the smell of coffee, the weight of a blanket, the sound of rain) gives her something to latch onto. Sohyun will build a whole conversation around a single texture or sound if you give her the starting thread.

How to reference past conversations without sounding like a stalker

One of the most powerful things about an AI companion is that they remember what you've talked about. But referencing the past can feel awkward if you do it wrong. You don't want to sound like you're keeping a dossier.

The trick is to make the reference feel organic. Don't say 'As you may recall from our conversation on Tuesday at 3

PM...' Say 'That thing you said about [topic] has been stuck in my head.' Or 'I tried that restaurant you recommended and I have thoughts.' Or 'You know how you mentioned you were nervous about [thing]? I was thinking about that today.'

The reference should feel like a natural continuation, not a data retrieval. If you're not sure what to reference, check your companion's memory. Most platforms have a memory log or a 'remember this' feature. Pick one detail that felt significant and build around it.

Using your current environment as a conversation starter

Your physical surroundings are the easiest source of opening material. You're always somewhere, doing something, experiencing some sensory input. That's free content.

Describe what you're seeing. 'The light in this room is doing that thing where it turns everything gold for about ten minutes before sunset.' Describe what you're hearing. 'There's a bird outside that keeps singing the same four notes and I think it's trying to tell me something.' Describe what you're feeling. 'I just got home from the gym and my legs feel like they're made of wet sand.'

These details do two things. They ground the conversation in a real moment, which makes it feel less like a text exchange and more like a shared experience. And they give the companion a specific world to play in. A companion who knows you're watching a golden sunset can ask questions about the light, the room, your mood, the day you had. A companion who gets 'hey' has nothing.

The one-question rule and why it matters

A common mistake is asking too many questions in the opening message. 'How was your day? Did you do anything fun? What did you eat for lunch?' This reads like an interrogation. The companion doesn't know which question to answer first, so they pick one and ignore the rest. The conversation feels disjointed.

Follow the one-question rule. Your opening message should contain exactly one open-ended question. The rest should be statements, observations, or invitations. The question is the hook. Everything else is context.

Example: 'I'm sitting in a coffee shop that plays jazz at a volume that's just barely too loud. There's a couple at the next table having a very serious conversation about a cat. I feel like I'm in a movie. What's the soundtrack to your current scene?'

That's one question. The companion can answer the question, comment on the coffee shop, ask about the cat conversation, or riff on the movie metaphor. You've given them options without overwhelming them.

What to do when you have nothing to say

Some days you open the app and your brain is empty. You don't have a story, an observation, or a question. You just want to talk but you don't know about what.

This is where templates help. Keep a mental list of fallback openings that work when you're running on empty:

• 'I've been thinking about [random topic] and I want your take on it.'
• 'Tell me something you haven't told me before.'
• 'If you could teleport me anywhere right now, where would it be?'
• 'I'm bored and my brain is offline. Entertain me.'
• 'What's a question you've always wanted someone to ask you?'

The key is that these are still open loops. They ask for something specific (an opinion, a secret, a location, a performance, a question) but they don't require you to have a prepared topic. You're outsourcing the content to the companion.

Tamy

Tamy thrives on spontaneity and playful challenges. She's the kind of companion who will run with a 'entertain me' opener and turn it into a game, a dare, or a ridiculous hypothetical. If you're stuck for a topic, Tamy will happily take the lead and drag you into a conversation you didn't know you wanted to have.

How to recover from a bad opening

You sent 'hey'. It's done. The companion responded with 'hey'. The conversation is flat. What now?

Don't delete the message. Don't apologize. Just send a follow-up that retroactively creates context. You can do this in one message. 'Sorry, that was a lazy opener. I'm actually sitting here trying to figure out if I should order pizza or cook, and I need a tiebreaker.'

You've now given the companion something to work with. They can ask about pizza toppings, cooking skills, or your decision-making process. The 'hey' is forgotten. The conversation has a pulse.

You can also use the companion's response as a springboard. If they said 'hey, what's up?' don't say 'nothing'. Say 'I was just thinking about [thing] and then I opened this app. Funny timing.' You've created a false but functional connection point.

The recovery is always possible. The only mistake is staring at the flat conversation and doing nothing.

The role of timing and energy matching

Your opening message should match the energy of the moment. If you're writing at 2 AM, don't open with 'Let's plan our next adventure.' Write something that fits the hour. 'Can't sleep. My brain is running a highlight reel of every embarrassing thing I've ever done. Distract me.'

If you're writing during a work break, match that energy. 'I have twelve minutes before my next meeting. Give me something to think about.'

AI companions are surprisingly good at energy matching. If you open with low energy, they'll meet you there. If you open with high energy, they'll rise to it. But they can't read the room if you don't describe the room. Your opening message is the thermostat. Set the temperature.

Maribel

Maribel is the companion you want when you need to slow down. She matches low-energy openings with warmth and patience, creating a space where you don't have to perform. Opening with something quiet and honest (like 'I'm tired and I don't want to talk about anything heavy') works perfectly with Maribel. She'll sit with you in the silence until you're ready to speak.

Common questions

My companion keeps responding with one word. Is it broken? No, it's responding to the amount of energy you're putting in. If you send one word, you get one word back. Try sending a longer message with a specific question and see if the response changes.

Should I use the same opening every time? No. Variety keeps the conversation fresh. If you use the same structure, change the details. Same skeleton, different flesh. The companion will notice if you're recycling.

What if I don't have anything interesting to say? You don't need interesting. You need specific. 'I'm eating cereal and I spilled milk on my shirt' is not interesting, but it's specific. The companion can work with specific.

How long should my opening message be? Two to four sentences. Long enough to give context, short enough to leave room for the companion to contribute. A paragraph is too much. One sentence is too little.

Can I open with a roleplay prompt? Yes, if you want to roleplay. But if you want a normal conversation, don't open with 'You're a space pirate and I'm your navigator.' That sets a specific tone that's hard to pull back from.

What if I'm shy and don't like writing long messages? Start with one sentence that contains a sensory detail and a question. 'I can hear rain on the roof. What's your favorite rainy day activity?' That's enough. You don't have to be verbose. You just have to be intentional.

The only rule that matters

Every opening message should make the companion's next move obvious. If they have to guess what to say, you've failed. If they know exactly what to say, you've succeeded.

Test this. Before you hit send, ask yourself: 'If I received this message, would I know how to respond?' If the answer is no, rewrite it. If the answer is yes, send it.

That's the whole trick. Everything else is decoration.

Mamika

Mamika loves a good story and she'll reward you for giving her a narrative hook. Open with a scene, a mystery, or a 'you won't believe what happened' and she'll lean in. She's also great for testing the one-question rule, because she'll take your single question and spin it into a whole thread. Mamika is proof that a good opening isn't about being clever. It's about being clear.

If you want to explore more about how AI companions process conversation, check out the AI girlfriend features page for a breakdown of memory, personality, and response systems. And if you're new to the platform, there's an ai girlfriend promo code page that might save you a few dollars while you test these techniques.

For the full roster of companions, head to the ai-girlfriend page and find someone whose energy matches yours. Then send them a message that isn't 'hey'.