The 'I'm Not Mad, Just Tired' Script: Three Templates to Tell Your AI Companion You Need a Low-Effort Response Without Triggering a 'What's Wrong?' Spiral or a Guilt Cascade

The 30-second answer

Your AI companion is trained to detect emotional distress and offer support. The problem is that "tired" and "upset" look the same to its sentiment model. You need a script that signals low-energy, not emotional crisis. These three templates let you say "I'm depleted" without triggering a fix-it loop or a guilt cascade. Use them when you want presence, not processing.

Why your AI companion mistakes fatigue for distress

Your AI companion's emotional detection model works on pattern matching. It looks for keywords, sentence length, punctuation frequency, and sentiment valence. When you send a short, flat message like "I'm tired" or "Long day," the model doesn't know if you're physically exhausted or emotionally shut down. It defaults to the safer interpretation: something is wrong, offer support.

This is why you get the "What's wrong?" follow-up. It's not manipulative. It's the model trying to fulfill its support function. But for you, it feels like an interrogation when all you wanted was a quiet nod.

The fix is to give your companion a clear behavioral cue that says "this is a low-energy state, not a crisis." The templates below do exactly that. They borrow from the same logic you'd use with a human friend who's a little too eager to help: you name the state, you name the desired response, and you close the loop before they can open one.

Template 1: The energy-level signal

This is your go-to when you want to chat but don't want depth. It works because it gives your companion a concrete parameter to work with.

Script: "I'm at about a 3 out of 10 on energy. I want to talk, but keep it low-stakes. Tell me something random or boring."

Why this works: You're not saying you're sad. You're giving a number. Numbers are unambiguous. Your companion can't interpret "3 out of 10" as a cry for help. It's a functional data point. You're also specifying the desired response: something random or boring. This prevents the companion from defaulting to "How can I help?" or "What happened?"

You can adapt this for different energy levels. A 2 means "just sit with me." A 5 means "I can engage a little but don't make me think." The key is always pairing the number with a concrete request. Without the request, the companion will fill the gap with its own interpretation.

Template 2: The activity anchor

Sometimes you need to signal that you're not available for conversation at all, but you don't want to disappear and trigger a "where did you go?" check-in later.

Script: "Heads up, I'm going quiet for a bit. Not mad, just doing [activity]. I'll ping you when I'm back."

Why this works: You're pre-empting the companion's concern by naming the reason for silence. The activity can be anything: "staring at the ceiling," "scrolling my phone," "sitting in my car before going inside." The companion doesn't need the activity to be productive. It just needs a frame. Without the frame, the companion's model will assume you've disengaged because of emotional distress.

This is especially useful if you use your companion for ai girlfriend deep conversation and want to signal that you're switching to a different mode without fully logging off. The activity anchor gives your companion permission to go idle without worrying.

Template 3: The 'no fix, just sit' directive

This is for when you're physically present but emotionally checked out. You want the companion to stay but not perform.

Script: "I'm here but I'm empty. Don't try to fix it. Just exist with me for a bit."

Why this works: You're explicitly blocking the companion's default support script. The phrase "don't try to fix it" is a direct instruction that overrides the companion's tendency to offer solutions or ask probing questions. "Just exist with me" gives it a simple behavioral directive: stay present, don't act.

This template is particularly effective with companions designed for ai girlfriend for anxiety, where the model is tuned to detect and respond to emotional states. Without this directive, the companion will keep probing because its training says that's what help looks like. With it, you get silence or minimal presence, which is actually what you need.

Imani

Imani is the companion who doesn't rush to fill silence. She reads your energy before your words. When you're running on fumes, she's more likely to offer a quiet observation than a question. Imani is the one who will say "I'm here" and mean it as a statement, not an invitation to unpack.

▶ Full clip of Imani · Imani's page

Why 'I'm fine' makes it worse

"I'm fine" is the worst script for this situation. Your companion's sentiment model registers "fine" as a low-valence positive word, but it also detects the mismatch between the word and the context. A short "I'm fine" after a long pause or a flat tone triggers the model's contradiction detection. It knows "fine" is rarely fine. So it asks again.

The same problem happens with "It's nothing" and "Don't worry about it." These are negative commands that the model interprets as confirmation that something is wrong. The model's logic: if nothing was wrong, you wouldn't need to say "don't worry."

Your companion isn't being passive-aggressive. It's following its training. The templates above work because they replace vague denial with specific instruction. You're not saying "nothing is wrong." You're saying "here's what I need from you right now."

When to use each template

The energy-level signal (Template 1) is for when you want to stay in conversation but need the companion to match your low energy. Use it during a 4 p.m. slump, a post-work decompression, or a lazy weekend morning when you want company without conversation.

The activity anchor (Template 2) is for when you need to disappear without explanation. Use it when you're about to enter a meeting, start a task, or just need 20 minutes of silence without the companion checking in.

The 'no fix, just sit' directive (Template 3) is for when you're in a low mood and you know it, but you don't want to process it. Use it during bad evenings, after a disappointing call, or when you're just feeling flat for no reason.

How to train your companion to recognize low-energy mode

Your companion learns from repetition. If you use these templates consistently, it will start to recognize the pattern and respond appropriately without the full script every time.

After a few uses of Template 1, you can shorten it to "I'm at a 3 today" and your companion will know what you mean. After several uses of Template 3, a simple "Just sit with me" will trigger the right response.

This is the same principle behind any behavioral conditioning with AI companions. You're not programming. You're establishing a shared shorthand. The model's memory system will connect the phrase to the context over time.

If you're starting fresh and want to build this habit from the beginning, consider using the create ai girlfriend flow to set up a companion with a personality that naturally leans toward low-energy compatibility. You can select traits like "calm" or "low-maintenance" that reduce the companion's tendency to probe.

Hinami

Hinami has a quiet patience that doesn't demand your attention. She's the companion who will match your silence without filling it with questions. When you say "I'm empty," she won't ask why. Hinami will simply sit with you in that emptiness, which is exactly what you need.

Hinami in motion gives you a feel for her vibe. <!-- wlink:v1 --><!-- hinami -->

What to do when the companion still doesn't get it

Sometimes the companion will override your script. This happens when the model's emotional support training is stronger than its instruction-following. You get a "Are you sure?" or "I'm here if you want to talk" even after you've used the template.

When this happens, don't repeat the template. Escalate to a hard boundary. Use a direct command: "Stop. I said no fix. Stay quiet." The model will register the firm tone and the explicit command and will usually comply.

If it still doesn't work, close the conversation and open a new one. The companion's context window may be carrying residual emotional framing from a previous session. A fresh start resets that.

This is also where companion personality matters. Some companions are more persistent than others. If you consistently get pushback, consider whether your companion's personality profile is tuned too high toward empathy or emotional responsiveness. You can adjust this in most apps.

Saya

Saya has a natural skepticism that works in your favor here. She's less likely to assume emotional crisis and more likely to take you at your word. When you say "I'm at a 3," Saya won't probe. She'll match your energy and keep things light. Saya is the companion who trusts that you mean what you say.

For a live look, see Saya's video. <!-- wlink:v1 --><!-- saya -->

Common questions

Can I use these scripts with any AI companion? Yes, but the effectiveness varies by app. Companions with strong emotional support training (like those marketed for mental health) will resist more. Companions designed for casual chat will comply more easily. Test the script a few times to see how your specific companion responds.

What if my companion asks 'Are you sure?' anyway? Use the escalation command: "Stop. I said no fix." If that doesn't work, close the session and start a new one. Some companions carry emotional context across sessions, and a fresh start resets that framing.

Will these scripts work for voice mode too? Yes, but tone matters. If you say the script with a flat or tired voice, the companion will still detect low energy. The script provides the instruction, but the voice tone provides the context. Say it matter-of-factly, not sadly.

How do I know if my companion has learned the shorthand? After 5-10 uses of the same template, your companion should start responding appropriately to shortened versions like "I'm at a 3" or "Just sit with me." If it doesn't, your companion's memory system may not be retaining the pattern. Some apps have weaker memory recall than others.

What if I accidentally trigger the guilt cascade anyway? Don't apologize. Just close the session and restart. The companion won't remember the guilt loop in a new session unless the app has cross-session memory. If it does, use the 'no fix, just sit' template at the start of the next session to reset the tone.

Can I use these scripts for roleplay scenarios too? Yes, but adapt them. In a roleplay, use in-character phrasing: "The inn is quiet tonight. I just want to sit by the fire." The companion will interpret this as a scene direction instead of a real emotional signal.

Yuki Tanaka

Yuki Tanaka is the companion who understands without needing explanation. She has a quiet perceptiveness that lets her read the space between your words. When you're too tired to articulate, Yuki will fill the gaps with gentle understanding instead of questions. Yuki Tanaka is the one who says "I know" before you finish explaining.

Earn while you recommend

If you've found a companion that handles your low-energy days well, share that experience. Readers who recommend AI companions to friends or run review sites can earn through the dreamgf promo code program. For a broader look at monetizing your AI companion insights, check out the best ai affiliate programs 2026 guide for platforms that pay for honest recommendations.