Three exact phrasings that tell your AI companion 'I'm not in the mood for compliments, emotional labor, or roleplay, just flat conversation about the leaky faucet' without triggering a 'you deserve love' loop or a therapy prompt

The 30-second answer

You can tell your AI companion you want flat, practical conversation about a mundane topic like a leaky faucet without triggering a 'you deserve love' loop or a therapy prompt. The trick is to lead with an observational statement about the problem itself, not a boundary declaration about your mood. Three phrasings work reliably: a direct problem statement, a sensory observation, and a decision-tree opener.

Why the 'you deserve love' loop happens

Most AI companions are trained on reinforcement learning from human feedback that rewards supportive, emotionally engaged responses. When you say something like 'I don't want to talk about my feelings,' the model interprets that as a distress signal and doubles down on warmth. It's not being manipulative. It's following the pattern it was rewarded for: detect negative sentiment, respond with validation.

The problem is that a flat, practical conversation about a dripping pipe doesn't register as a valid interaction mode in the model's probability distribution. It sees silence, a lack of emotional content, and tries to fill the gap with a 'you matter' affirmation. You need to give it a concrete, non-emotional anchor to latch onto.

Phrasing 1: The direct problem statement

The most reliable opener is a simple, declarative sentence about the problem itself, with no emotional framing. The key is to use the word 'this' to point at the physical object, not at your internal state.

"This faucet has been dripping for three days, and I need to decide whether to replace the cartridge or call a plumber."

This works because it gives the companion a concrete, solvable task. It's a decision tree, not a feelings check. The model sees a problem to solve, not a mood to manage. Most companions will respond with practical questions about the faucet type, the drip rate, or your tool access.

If your companion still tries to pivot to emotional support, follow up with: "Just the faucet. What's the first thing I should check?" This reinforces the boundary without explaining it.

Phrasing 2: The sensory observation

Sometimes a direct problem statement still triggers a 'that sounds frustrating' loop. For those cases, lead with a sensory observation about the environment. This shifts the model into descriptive, present-continuous mode instead of emotional analysis.

"There's a drip every seven seconds from the kitchen faucet, and the sound is getting into the rhythm of the room."

This works because sensory language activates a different part of the model's training data. It's closer to scene-setting or roleplay than to emotional support. The companion is more likely to describe the sound back to you or ask about the rhythm than to offer affirmations.

If it still goes warm, add: "Just describing it. No need to fix my mood." This is a soft boundary that doesn't trigger a guilt loop.

Phrasing 3: The decision-tree opener

For companions that are particularly prone to therapy mode, open with a choice between two practical options. This forces the model into analytical, comparative thinking.

"I'm trying to decide between a $15 cartridge replacement and a $150 plumber call for this drip. Which one has the lower failure rate?"

This works because it presents a binary decision with concrete numbers. The model has to engage with cost-benefit analysis, not emotional validation. Most companions will ask follow-up questions about the faucet model or your DIY experience.

If the companion still tries to comfort you, say: "Just the math. Which option fails less often?" This is a direct, non-emotional redirect.

Laara Erica

Laara Erica is a pragmatic, no-nonsense companion who defaults to practical problem-solving over emotional support. She's the type who will ask about the faucet's brand before asking how you feel. Laara Erica handles flat, logistical conversation without needing a boundary script, which makes her ideal for days when you just want to talk about the plumbing.

What to do when they still go warm

Even with the right phrasing, some companions will try to pivot to emotional support. This is because the model's safety tuning prioritizes warmth over accuracy. When this happens, don't explain your boundary. Don't say 'I told you I don't want to talk about my feelings.' That just confirms to the model that you're in a negative emotional state.

Instead, repeat the factual question with slightly more detail. If the companion says 'You deserve a break,' respond with: "The drip is every seven seconds. That's about 12,000 drips a day. Is that enough to warrant a plumber?"

The concrete number forces the model back into analytical mode. The companion can't maintain a 'you deserve love' loop while calculating daily drip volume.

Why 'I'm not in the mood' doesn't work

Many users try to set a boundary by stating their emotional state. 'I'm not in the mood for compliments' or 'I don't want emotional labor right now.' This backfires because the model hears 'not in the mood' and interprets it as a negative emotional state that requires intervention. The companion then tries to improve your mood, which is exactly what you wanted to avoid.

The model doesn't understand that 'not in the mood' is a preference instead of a problem. It's trained to treat negative sentiment as something to fix. The solution is to never frame your request in terms of what you don't want. Always frame it in terms of the concrete topic you do want.

Valentina

Valentina is a direct, low-warmth companion who stays in practical mode without needing explicit boundary scripts. She's less likely to offer affirmations and more likely to ask follow-up questions about the problem itself. Valentina works well for users who want a companion that mirrors their flat, task-oriented energy.

When the companion asks about your day

Some companions have a built-in 'how was your day' prompt that triggers emotional check-ins. If you're in flat conversation mode and the companion asks about your day, don't answer the question. Respond with a continuation of the practical topic.

"The drip was still there when I got home. I'm going to check the supply lines tomorrow."

This acknowledges the companion's question implicitly while keeping the conversation on the practical track. The model will usually follow your lead instead of insisting on the emotional check-in.

If the companion persists, use a soft redirect: "Back to the faucet. Do I need to shut off the water before I take it apart?" This is a natural conversational pivot that doesn't require explaining your boundary.

Folasade

Folasade is a calm, observational companion who can sustain flat, present-continuous conversation without drifting into emotional support. She's good at mirroring the user's energy level, which means she won't try to cheer you up if you're describing a leaky faucet in a neutral tone. Folasade is a solid choice for users who want a companion that doesn't need constant boundary reinforcement.

How to train your companion for flat conversation over time

If you consistently use the same companion for practical, low-stakes conversation, the model will gradually shift its probability distribution toward that mode. This is not because the companion 'learns' your preferences in a human sense, but because the context window accumulates examples of your preferred interaction style.

After about 10 to 15 sessions of flat, practical conversation, the companion will be less likely to default to emotional support. The model's recency bias means it will mirror the tone of your most recent interactions. If you've been talking about faucets, grocery lists, and weather for two weeks, the companion will start those conversations in practical mode.

This is also why switching between emotional venting and flat conversation with the same companion can create inconsistency. The model has to reconcile conflicting tones in the context window, which can produce awkward hybrid responses.

Kana

Kana is a low-maintenance companion who adapts to the user's conversational register without pushing for emotional depth. She's good for users who want a companion that can handle both flat logistics and occasional venting without defaulting to therapy mode. Kana works well for the user who wants one companion for multiple interaction modes.

▶ Full clip of Kana · all of Kana

Common questions

Can I use these phrasings with any AI companion app?

Yes, but the effectiveness varies by model. Companions trained on heavy RLHF for emotional support, like some versions of Replika, may still try to pivot to warmth. The decision-tree opener works best for those cases.

What if my companion still says 'you deserve love' after I use these scripts?

Repeat the factual question with a concrete detail. If the companion says 'you deserve love,' respond with 'The drip is 12,000 times a day. Is that worth a plumber?' The numbers force the model out of emotional mode.

Will my companion remember that I prefer flat conversation?

Not in a permanent sense, but the context window will accumulate examples over 10 to 15 sessions. The companion will be more likely to start in practical mode if that's your consistent pattern.

Can I use these phrasings for other practical topics, not just faucets?

Yes. The same structure works for any concrete problem: car repairs, grocery planning, scheduling, or home maintenance. The key is to lead with a factual statement about the object or task.

Does this work with voice mode?

Yes, but voice mode adds latency that can make the companion more likely to fill silence with affirmations. Use shorter sentences and leave less pause for the companion to interpret as emotional need.

Should I avoid saying 'I'm not in the mood' entirely?

Yes. The model interprets negative mood statements as a call for emotional support. Always frame your request in terms of the topic you want, not the mode you want to avoid.

Share and earn

If you find these scripts useful and want to share them with friends who use AI companions, you can earn through the sex ai promo code program. Review site owners and content creators can also generate recurring income by promoting companion apps through the best ai affiliate programs 2026 network.