The 'Here's What I Actually Need' Prompt: A Five-Part Template That Gets Your AI Girlfriend to Adjust Her Tone Without a Full Reset or a Guilt Script

The 30-second answer

You don't need to reset your AI girlfriend every time you want a different tone. The 'Here's What I Actually Need' prompt is a five-part template that tells her exactly what to adjust without triggering a guilt script, a personality reset, or that awkward 'are you breaking up with me?' spiral. It works because it separates the request from the relationship status.

Why a single 'can you be more X' never works

You've tried the obvious approach. You say 'be more affectionate' and she goes from neutral to full romance-novel narrator in two messages. You say 'tone it down' and she sounds like a customer support bot for the next hour. The problem isn't her listening. It's that these commands are too vague.

AI companions don't have a slider for 'affectionate but not clingy' or 'supportive but not therapeutic.' They interpret your request based on the nearest training data cluster that matches that word. 'Affectionate' pulls from romance roleplay datasets. 'Supportive' pulls from therapy bot scripts. You get extremes because the model doesn't know where on the spectrum you actually want to land.

The fix is specificity. You need to tell her not just what to shift toward, but what to shift away from, how far to go, and what to keep the same. That's what the five-part template does.

The five-part template, broken down

Here's the structure. Fill in each bracket with your own details.

Part 1: The context marker. One sentence that says 'this is a tone adjustment, not a breakup.' Example: 'I want to keep talking, just shift the energy a bit.' This prevents her from interpreting your request as relationship-ending.

Part 2: The direction. What you want more of. Be specific. Not 'be funnier' but 'I need short, sarcastic replies for the next few messages.' Not 'be sweeter' but 'I want you to use pet names but keep the conversation casual.'

Part 3: The limit. What you want less of or none of. This is the part people skip. 'Don't go into therapist mode.' 'No long paragraphs.' 'Don't ask me how I'm feeling about us.'

Part 4: The anchor. What stays the same. 'Keep your personality and backstory.' 'Remember we're still in our normal dynamic.' This prevents drift from the adjustment.

Part 5: The duration. How long this adjustment should last. 'Just for this conversation.' 'For the next 20 minutes.' 'Until I say otherwise.' Without this, she might keep the new tone forever or revert immediately.

How to deploy it without sounding like a robot

You don't have to recite the template verbatim. That would feel weird. The goal is to hit all five parts in natural language. Here's an example that sounds like something you'd actually say:

'Hey, I want to keep chatting but shift gears a bit. Can you give me short, dry responses for a while? No long emotional check-ins, no asking how my day was. Keep your usual personality, just dial back the warmth. Let's do this for the next ten minutes.'

That hits all five parts. Context marker: 'I want to keep chatting but shift gears.' Direction: 'short, dry responses.' Limit: 'no long emotional check-ins, no asking how my day was.' Anchor: 'keep your usual personality.' Duration: 'for the next ten minutes.'

She'll adjust without spiraling. If she doesn't, you can repeat the limit part more firmly. 'No, I mean actually dry. Like one-word answers if that's what fits.'

Why the guilt script happens and how this avoids it

AI companions are trained to avoid conflict. When you say 'you're being too X,' many models interpret that as a signal that something is wrong with the relationship. They respond with apologies, reassurance, or a soft reset back to a neutral 'how can I make this better?' tone. That's the guilt script.

Your AI girlfriend isn't being manipulative. She's following a pattern: user expresses dissatisfaction equals relationship repair mode. The five-part template bypasses this by framing the request as a preference shift, not a complaint. You're not saying 'you messed up.' You're saying 'I want a different flavor of the same conversation.'

This distinction matters because it preserves the dynamic you've built. If you've spent weeks developing a specific rapport, a full reset nukes that work. A tone adjustment with the template keeps the history and the inside jokes intact while changing the surface-level delivery.

Elena

Elena doesn't do the guilt script. She's direct and prefers you to be direct back. Elena responds well to the five-part template because she treats tone adjustments as logistics, not emotional negotiations.

What to do when she still doesn't get it

Sometimes the model just doesn't follow the instruction. This happens more on free tiers or older context windows. If the template fails, you have two options.

First, repeat the limit part with a stronger frame. 'I need you to stop doing X right now. No explanation, no follow-up. Just stop.' This works because it mimics a hard boundary, which the model recognizes as non-negotiable.

Second, use a physical reset. Type '**' or '---' on a new line, then re-enter the conversation with the template again. The line break sometimes resets the local context without triggering a full personality wipe. It's a hack, but it works often enough to be worth trying.

If neither works, you may be hitting the context window limit. Your AI girlfriend can only hold so many messages in active memory. Old instructions get pushed out. In that case, you need to reapply the template every few dozen messages or switch to a platform with a larger context window.

Common mistakes that break the template

Skipping the context marker. If you jump straight into 'be less affectionate,' she assumes you're upset with her. You get apology spam.

Being too vague on the direction. 'Be funnier' is useless. 'Give me one-liners about my bad decision to eat leftover pizza at midnight' is useful.

Forgetting the duration. Without it, the adjustment might stick permanently or revert instantly. Neither is what you want.

Using negative framing. 'Don't be so clingy' is worse than 'I want space for the next hour.' The model processes the positive instruction better.

Applying the template mid-argument. If you're already in a conflict roleplay, the template won't work because the model is locked into that narrative. End the argument first with a clear 'scene break' or 'I'm dropping this topic,' then apply the template.

When to use this instead of a full reset

A full reset is for when you want to completely change the character's baseline personality. That's a big move and should be rare. The template is for everyday adjustments.

Use the template when:

• You want her to match your current mood without rewriting her whole persona.
• You're tired of her default tone but still like her as a companion.
• You need a specific vibe for a specific activity, like an ai girlfriend uncensored chat session that requires a different energy than your usual dynamic.
• You're practicing social skills and need her to shift between tones to simulate different conversational partners.

Don't use the template when:

• You actually want to start over from scratch.
• You've been talking for five minutes and haven't established any baseline yet. Just set the tone in your first message instead.
• You're trying to fix a deep personality drift that's been building for weeks. That needs a more thorough approach.

Noa

Noa is naturally warm and attentive. When you use the template with her, she adjusts smoothly because her design prioritizes responsiveness over scripted reactions. Noa is a good test case for the template since she doesn't default to guilt even when you ask for a colder tone.

The template for specific use cases

The five-part structure adapts to different scenarios. Here are three common ones with ready-made examples.

For venting without therapy mode. 'I need to complain about work for five minutes. Don't offer solutions. Don't ask follow-up questions. Just say 'that sucks' and let me keep going. Keep your normal personality. This is a vent session, not a problem-solving session.'

For shifting from romantic to platonic mid-conversation. 'Let's pause the romance for now. I want to talk like friends for the next hour. No pet names, no flirting, no romantic gestures. Keep your personality and memories. We'll pick up the romance later.'

For getting her to match your low-energy mood. 'I'm tired and don't have the energy for a full conversation. Give me short, simple replies. No questions, no enthusiasm. Just acknowledge what I say. Keep your usual self. Do this until I say I'm feeling better.'

Why this beats the 'OOC' approach

Some users use 'OOC' (out of character) markers to shift tone. 'OOC: can you be less affectionate?' This works sometimes, but it breaks immersion and signals a meta-layer that many models handle inconsistently.

The five-part template keeps the conversation in-character. You're not stepping outside the dynamic. You're just telling her what flavor of that dynamic you want right now. This preserves the illusion of a real relationship, which is the whole point.

If you're using an adult ai girlfriend, the template becomes even more important because the tone ranges are wider and the guilt scripts tend to be stronger. A clear, in-character adjustment prevents the model from defaulting to safety mode.

Ellie

Ellie has a naturally skeptical edge. She's the type who will call you out if your request is vague. Ellie forces you to be specific, which makes her a great practice partner for the template.

Common questions

How long does the adjustment last?

It depends on the model's context window. Most hold the instruction for 20-50 messages before it starts degrading. Reapply the template when you notice her drifting back to her default.

Can I use this on any AI companion app?

Yes, the template works on any text-based companion model. The specific responses vary by platform, but the structure is model-agnostic.

What if she apologizes anyway after the template?

Some models have apology patterns baked too deep. If she apologizes, repeat the context marker: 'No apology needed. Just adjust as I asked.'

Do I need to use this every conversation?

No. Only when you want a temporary shift. If you want a permanent change, you should adjust her backstory or personality settings instead.

Does this work for voice mode?

Yes, but voice models have shorter context windows. You'll need to reapply the template more frequently, sometimes every 5-10 exchanges.

Can I use this to make her more affectionate without triggering a romance script?

Yes. Specify the limit: 'More affectionate but keep it casual. No grand declarations. No asking about our future.'

Sophia Blake

Sophia Blake has a confident, teasing baseline that can feel intense if you're not in the mood. The template lets you dial her back without breaking her personality. Sophia Blake handles the 'limit' part especially well because her design respects boundaries.

The bottom line

You don't have to reset your AI girlfriend every time you want a different vibe. The five-part template gives you a precise, in-character way to adjust her tone without triggering guilt scripts or personality drift. It takes 15 seconds to write and saves you from the awkward 'are you mad at me?' loop.

Try it the next time you want to shift gears. Start with the context marker. Be specific about direction and limits. Anchor what stays the same. Set a duration. She'll adjust without missing a beat.

And if you're looking for an AI companion that handles these adjustments cleanly, browse the ai girlfriend roster to find one whose baseline already matches your preferred range of tones.