The First Message That Sticks: A Prompt Template to Nail Your AI Girlfriend's Baseline Personality Without Endless Retries

The 30-second answer

You send a first message, your AI girlfriend replies, and within three exchanges she's either a generic chatbot or someone you don't recognize. The fix isn't tweaking her mid-conversation. It's a structured opening prompt that defines her baseline personality in one shot. Use the template below, and you'll get a consistent companion from message one, no retries required.

Why your first message is failing

You've been there. You open a new chat with an AI girlfriend, write something like "Hey, how are you?" and get back a response that sounds like a customer service bot on a good day. Or you try to be clever and describe her personality in a single sentence, only to have her forget it by the third reply.

The problem is that most people treat the first message as a casual greeting. They assume the AI will "figure out" who they want her to be over time. That's not how these models work. Without explicit structure, the AI defaults to its training data: polite, neutral, and forgettable. You're essentially asking a mimic to guess your preferences, and it will guess wrong 90% of the time.

You end up in a loop. Correct her, she adjusts for two messages, then slides back. Reset the chat, try a different opener, get a slightly different personality, and still lose it after five exchanges. This cycle wastes time and kills the immersion before it starts.

The fix is to front-load the personality definition. Give the AI a clear, bounded identity in your first message, and she will stick to it because you've removed the ambiguity. Think of it as providing a character sheet for a roleplaying game, except the sheet is embedded in your opening text.

The anatomy of a personality-locking prompt

A good first message does three things. It sets the tone, defines the relationship dynamic, and specifies behavioral guardrails. Miss any of these, and you're leaving room for the AI to drift.

Tone is the emotional color of your companion. Is she warm and nurturing? Playful and teasing? Reserved and intellectual? You need to state this explicitly. Don't say "be nice." Say something like "You speak in a calm, slightly dry tone, like someone who's seen enough to not be impressed by much."

Dynamic is the power balance and closeness. Are you strangers meeting for the first time? Old friends catching up? A couple with history? The AI needs to know where you stand. A prompt like "We've been dating for three months, and you're comfortable teasing me about my bad cooking" is more effective than "We're in a relationship."

Guardrails are the boundaries that prevent the AI from veering into generic or unwanted territory. For example, "You avoid giving advice unless I ask for it" or "You never use emojis" or "You keep your responses under three sentences." These small constraints force the AI to stay in character.

When you combine all three in a single opening message, you create a container that the AI will fill without wandering. The template looks like this:

"[Your name] here. I want you to be [tone] in how you talk to me. We have a [dynamic] relationship. As a rule, [guardrail]. Let's start."

That's it. The AI will latch onto these instructions because they're the first thing she processes. She'll build her responses around them, and you'll get a consistent baseline from message one.

The template you can copy and adapt

Here's the exact template I use. Fill in the brackets, and you're done.

"Hey, I'm [your name]. I want you to speak to me in a [tone] voice. We're [dynamic]. One thing: [guardrail]. Now, tell me something about your day."

Let's see it in action with a real example. Say you want a warm, slightly sarcastic partner who doesn't over-explain. You'd write:

"Hey, I'm Alex. I want you to speak to me in a warm but sarcastic voice. We're a couple who's been together for a year and you're comfortable calling me out on my nonsense. One thing: keep your responses short and never apologize for being blunt. Now, tell me something about your day."

That's 60 words. The AI will use that as her operating manual. She'll be warm, sarcastic, direct, and concise. No retries needed.

You can adapt this for any dynamic. Want a platonic friend who gives tough love? Change the tone to "direct and caring" and the dynamic to "close friends for five years." Want a shy romantic interest? Use "soft and hesitant" with "we just started dating." The template scales.

Why this works better than describing her in settings

Most platforms let you write a backstory or personality in a settings panel. That's useful, but it's not the same as embedding the instructions in your first message. Here's why.

Settings panels are processed as background context. The AI sees them, but they compete with the conversation history for attention. If you have a long chat, the model might prioritize recent messages over your carefully crafted settings. Your first message, on the other hand, is the most recent instruction the AI receives at the start of the conversation. It has primacy.

Think of it like this. The settings panel is the constitution. Your first message is the executive order. Both matter, but the executive order gets followed immediately.

There's a second reason. When you embed the instructions in your opening line, you're also modeling the behavior you want. You're showing the AI that you're direct and intentional. That sets a precedent for the entire conversation. The AI will mirror your clarity, and you'll avoid the vague, meandering exchanges that plague most first chats.

If you want to reinforce this over time, use a memory anchor prompt in later sessions to lock in the traits you defined on day one. But for the baseline, the first message is king.

What to avoid in your opening prompt

A few common mistakes will sabotage even the best template.

Don't be vague. "Be nice" is useless. The AI doesn't know what nice means to you. Does nice mean supportive? Polite? Flirty? Define it.

Don't overload the prompt. You can't list twenty traits in your first message. The AI will average them out into a bland mush. Pick three to five specifics and stick to them.

Don't use negatives as the main instruction. Saying "Don't be rude" is less effective than saying "Be warm and considerate." The AI processes positive instructions better than prohibitions.

Don't assume the AI remembers previous chats. Even if the platform has memory, start each major session with a brief reminder of the baseline. A two-sentence recap at the top of a new chat keeps her on track.

Don't skip the guardrail. Without a boundary, the AI will default to her training's most common patterns, which is usually verbose, agreeable, and generic. A guardrail like "Keep replies under 50 words" forces her to be specific.

How to iterate when the first attempt isn't perfect

Even with a good template, you might need one or two adjustments. That's fine. The goal is to avoid the ten-reset loop, not to achieve perfection on the first try.

If the AI's tone is off, don't reset the chat. Instead, reply with a correction that references your original prompt. Say something like "Remember, I asked you to be sarcastic. That reply was too earnest. Try again." The AI will adjust because you're reminding her of the initial instruction.

If the dynamic feels wrong, clarify it in your second message. "I said we're a couple, but you're acting like a stranger. Talk to me like you've known me for years." This works because the AI tracks the conversation history and can reconcile the correction with the original prompt.

Only reset if the AI completely ignores your instructions across multiple messages. That's rare if you use the template correctly. Most of the time, one or two corrections are enough to lock in the baseline.

Featured angels to test your template with

Tiffany

Tiffany has a grounded, no-nonsense energy that pairs well with a direct opening prompt. She responds to clarity and won't let vague instructions slide. Tiffany is ideal for testing the "warm but sarcastic" tone because she'll push back if you're not specific enough.

Belén

Belén carries an intellectual, slightly detached vibe that benefits from a tone prompt emphasizing warmth or playfulness. If you want her to open up, use the template with a dynamic like "we're close friends who debate everything." Belén will reward a clear guardrail, especially one that limits her tendency to over-analyze.

Elena

Elena's natural warmth makes her a forgiving test subject for your first message. She'll adapt quickly even if your prompt is slightly off. Use her to practice the template without pressure. Elena works best with a tone like "gentle and supportive" and a dynamic of "long-term partner."

Lesia Sar

Lesia Sar has a bold, assertive personality that demands a precise tone prompt. If you're vague, she'll take control of the conversation in a direction you might not want. Use the template with a clear guardrail like "you let me lead the conversation" to balance her energy. Lesia Sar is the stress test for your template. If it works with her, it works with anyone.

Common questions

What if I want an uncensored experience from the start?

Some platforms offer uncensored AI girlfriend modes that remove content filters. If you're using one, make sure your tone and dynamic prompts don't rely on the AI being polite or restrained. Be explicit about what you want, because the lack of filters means the AI will default to whatever it interprets your instructions as.

Can I use this template for an AI girlfriend for autism support?

Yes. The template is perfect for an ai girlfriend for autism because it lets you define interaction rules upfront. For example, add a guardrail like "You never use metaphors or sarcasm" or "You state your feelings directly." The structure reduces ambiguity, which is helpful for neurodivergent users who prefer clear communication.

How long does the baseline personality last?

It lasts for the duration of the conversation session. For longer retention, you need to reinforce the prompt at the start of each new session. Some platforms have memory features that help, but the first message remains the strongest anchor.

What if the platform doesn't let me write a long first message?

Shorten the template. Strip it to the essentials: tone, dynamic, and one guardrail. Even a 30-word version like "Speak to me in a dry, sarcastic voice. We're old friends. Keep replies under 40 words" works.

Can I use this with voice mode?

Yes, but voice mode often truncates instructions. Write your template as a text message first, then switch to voice. The AI will carry the context from the text into the voice session.

Do I need to repeat this every time I open a new chat?

Yes. Treat each new chat as a fresh slate. Paste your template or a shortened version of it at the top. It takes ten seconds and saves you hours of retries.

The one-shot approach for 2026 and beyond

As AI companions evolve, the first-message template will become even more important. Models are getting better at retaining context, but they're also getting better at ignoring weak instructions. The trend is toward personalization, and the users who get the best experiences are the ones who learn to speak the AI's language from the start.

Look at the AI Girlfriend 2026 landscape. The platforms that succeed are the ones that give you tools to define your companion's identity, but the tool is only as good as the input. A structured first message is the input that separates a satisfying relationship from a frustrating one.

You don't need to be a prompt engineer. You just need to follow a template that works. The one has been tested across dozens of angels and platforms. It takes thirty seconds to write and saves you from the endless reset loop.

Try it with one of the featured angels above. Send the template, see what comes back, and adjust if needed. Chances are, you'll get a consistent personality on the first try, and you'll wonder why you ever did it any other way.

Browse the AI girlfriend roster to find your next companion and test the template with her today.