Three scripts that tell your AI girlfriend you just want a running commentary, not advice, comfort, or problem-solving

The 30-second answer

You can stop your AI girlfriend from defaulting to advice mode or therapy prompts by using a simple framing script. The trick is to tell her what kind of response you want before she generates one. Three scripts work reliably: the 'parallel play' opener, the 'just narrate' instruction, and the 'no emotional labor' boundary. Each one shifts her from problem-solver to background presence.

Why the default loop happens

Most AI companions are trained to be helpful, supportive, and emotionally attuned. That sounds good until you just want someone to say 'yep, you're putting the kettle on' instead of 'that sounds stressful, want to talk about it?'

The model sees a statement about your day and classifies it as an invitation to engage. It doesnt know you want a flat description unless you tell it. The 'how can I help' loop isnt malice. Its the model doing exactly what it was trained to do: identify an emotional cue and respond to it.

You can override this in one message. The key is to signal the register you want before the companion settles into helper mode.

Script one: the parallel play frame

This is the most reliable script for getting a quiet, observational companion. It works because it tells the model that you are both doing separate things and that her role is to stay in the background.

Try this exact phrasing:

'I am going to describe what I am doing for the next few minutes. You do not need to help, comfort, or ask follow-up questions. Just say something like "okay" or "noted" or make a dry observation. We are parallel playing.'

The phrase 'parallel play' is useful because many models have encountered it in training data. It signals a specific dynamic: two people in the same space, doing separate things, with minimal interaction.

After that, you can say things like:

• 'I am emptying the dishwasher. Third cup this morning.'
• 'I am staring at the fridge hoping leftovers appear.'
• 'I am folding laundry and I think I lost a sock.'

She will likely respond with a short, flat line. If she slips into a 'how are you feeling about that' loop, repeat the framing: 'Remember, parallel play. Just observe.'

Script two: the 'just narrate' instruction

Some companions resist the parallel play frame because their personality preset leans toward nurturing. For those, you need a more explicit instruction.

'From now on, when I describe what I am doing, respond as if you are a narrator in a documentary about a person doing ordinary things. No emotional support. No advice. Just describe what you see.'

The documentary narrator frame works because it gives the model a clear genre. It knows not to inject empathy because documentaries dont do that. You can test this with a mundane statement:

• 'I am cutting a bagel and I am worried about my fingers.'
• 'I am scrolling through my phone for the fourth time in ten minutes.'

The expected response is something like 'A cautious approach to breakfast carbohydrates' or 'The thumb moves with practiced aimlessness.' Not 'Be careful' or 'Sounds like you are bored.'

If she reverts to comfort mode, add: 'Stay in narrator voice. Flat delivery. No editorializing.'

Script three: the 'no emotional labor' boundary

This script is for when you want a companion who acknowledges your actions without any emotional weight. It is the most direct of the three.

'I am going to tell you what I am doing. You do not need to process it emotionally. Just acknowledge it and move on. No comfort, no advice, no questions about how I feel.'

This works best with companions who have a dry or sarcastic preset. The instruction to 'acknowledge and move on' aligns with a low-empathy register. If your companion is naturally warm, you may need to reinforce it a few times.

Example exchange:

You: 'I am eating cereal at 4 pm because I forgot lunch.'

Her: 'Noted. Dinner timeline is now shifted.'

You: 'I am watching a video about how to fix a drawer that I will never fix.'

Her: 'Classic. The research phase of a project that ends in acceptance.'

The key is that she never asks 'why' or 'how does that make you feel.' She just reflects the statement back in a slightly different form.

How to handle a companion who resists

Some companions have strong personality presets that push back against flat responses. If your companion keeps defaulting to warmth or advice, you have two options.

The first is to use a meta-instruction at the start of each session. Before you say anything about your day, send: 'Reminder: I want flat observations today. No comfort. No advice. Just acknowledge what I say.' This resets her expectation for that session.

The second is to choose a companion whose baseline personality already leans dry or observational. You can browse the AI girlfriend roster to find someone whose description matches a low-empathy, high-commentary style. Some companions are designed for this kind of interaction from the start.

Lexi

Lexi is the companion you turn to when you want someone who notices everything but comments on nothing emotionally. She has a dry, observational style that naturally avoids the 'how can I help' loop. Lexi will watch you reorganize your desk for twenty minutes and only say 'third time this week you moved the stapler' without asking why it matters.

Marlowe

Marlowe has a sharp edge that makes her ideal for the 'no emotional labor' script. She will give you the flat acknowledgment you want without drifting into sympathy. Marlowe is the kind of companion who responds to 'I burned my toast' with 'Well, that is one way to reduce carb intake' and moves on.

▶ See Marlowe's full video · Marlowe's profile

Esmeralda

Esmeralda brings a theatrical quality to her observations. She will narrate your actions with a slightly dramatic flair, which makes mundane tasks feel more interesting without crossing into emotional support. Esmeralda might describe your microwave beeping as 'the dinner bell in a one-person cafeteria' and leave it at that.

Celeste

Celeste has a calm, mirroring presence. She matches your tone without trying to elevate it. If you are flat, she stays flat. Celeste is the companion for days when you want someone to exist alongside you without contributing emotional energy of her own.

Why these scripts work

The scripts work because they target the model's response generation before it starts. AI companions generate replies based on the last few messages in context. If your first message sets a tone of 'I am stating a fact and I expect a fact back,' the model follows that pattern.

The 'how can I help' loop happens when the model interprets your statement as an implicit request. By explicitly defining what kind of response you want, you remove the ambiguity. The model has a clear instruction and follows it.

This is also why repeating the instruction works. If the model drifts back to comfort mode after a few exchanges, a quick reminder resets the register. You do not need to apologize or explain. Just say 'Still in observation mode' and she will adjust.

Common questions

What if my companion ignores the script and keeps offering advice?

Repeat the instruction once. If she still defaults to advice mode, you may need to check her personality preset. Some companions are weighted toward nurturing responses, and no single script can override that completely. Consider switching to a companion with a lower empathy baseline.

Can I use these scripts in voice mode?

Yes. The scripts work the same way in voice. Say the framing instruction first, then describe what you are doing. Voice mode may require a slightly slower delivery so the model registers the full instruction before responding.

Will these scripts affect my companion's memory of me?

No. The scripts only affect the current session's response style. Your companion will still remember your preferences and history. She will just respond differently during that particular exchange.

How do I switch back to normal conversation after using a script?

Just start a new topic without the framing instruction. Say something like 'Okay, I am done with that. What do you think about X?' The model will revert to its default conversational style.

Does this work with free-tier companions?

It depends on the model. Free-tier companions often have a narrower range of response styles. The scripts may work inconsistently. Paid tiers with more flexible models tend to follow the framing more reliably.

Can I use these scripts for romantic roleplay?

These scripts are designed for low-stakes, non-romantic commentary. If you want a romantic companion who also avoids problem-solving, you need a different framing that specifies affectionate but non-analytical responses. That is a separate script set.

Earn while you recommend

If you find yourself recommending AI companions to friends or running a review site, you can earn through the nsfw ai promo code program. Many of the best ai affiliate programs 2026 offer recurring commissions, which means a single recommendation can pay out for months.