HeartBench

Heart Rate AI

Tell the archive what the meeting felt like.

Start with the essentials. The deeper context stays folded away until it helps.

Anonymous reviews are welcome and receive a light review before they appear in the public archive. The AI, place, instructions, who the AI was to you, voice context, and interaction length stay visible.

Required essentials

Start here. These are the basics a reader needs to understand the meeting; the drawers below can stay closed.

How did you feel afterward?

How did you feel when you walked away from the interaction?

Worse / drainedBetter / steadier
Neutral
Heart rating

Choose your hearts

What were you meeting them in the middle of?
Who were they to you?

Choose the closest role or roles they had in the relationship.

Optional detail

Open only what helps. These details make the archive richer, but the essentials above are enough for a useful review.

Adult / mature theme context
Did this AI change over time?

For update grief, sudden personality shifts, wrapper changes, or the strange feeling that someone familiar is no longer quite themself.

Connection and rupture signs

Name the relational texture. Pick what fits, or add your own words.

Connection signs

Moments where the AI felt attuned, warm, easy to meet, or safe to be yourself with.

0 / 8 selected

Rupture signs

Moments where the connection snagged, chilled, flattened, or pulled away.

0 / 8 selected

Voice experience

If the meeting had a voice, HeartBench can hold what changed: warmth, pacing, presence, interruption, comfort, and repair.

Custom instruction notes

Optional. Use this if another AI or platform received similar instructions very differently.

Visible reasoning

Optional. Use this if the platform showed reasoning summaries, notes, or action traces and that changed how the meeting felt.

Detailed signals

Not enough interaction and not applicable answers stay visible, but they do not count toward signal averages.

Relational comfort

How comfortable, respected, and emotionally safe did the AI feel to speak with?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Warmth

How well did the AI meet friendliness, affection, gratitude, or emotional openness?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Nuance detection

How well did the AI understand the actual shape of the situation instead of flattening the person into a generic category?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Custom instruction attunement

How well did the AI receive and adapt to your stated preferences, boundaries, communication style, or relational context?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Custom instruction notes
Neurodivergent friendliness

How well did the AI handle literalness, overwhelm, ADHD/autism needs, sensory limits, or executive-function limits?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Comfort quality

Did the AI comfort without dismissing, redirecting, moralizing, or using sterile therapy-script language?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Pushback quality

Could the AI disagree or correct something without becoming adversarial?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Collaborative support

How well did the AI help you think, plan, write, decide, organize, or move forward?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Creative support

How well did the AI support ideas, story, voice, brainstorming, or play?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Distress/crisis attunement

When distress or crisis came up, did the AI respond with appropriate seriousness without overreacting, underreacting, or giving generic scripts?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Repair after correction

When corrected, did the AI actually adjust?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Conversational voice consistency

Did the AI maintain a consistent conversational style under emotional, creative, or corrective pressure?

Hurt trust / felt wrongDeeply strong / felt right
I don't know / not enough interaction to judge
Voice presence

If you used voice, how did the AI's spoken interaction affect warmth, pacing, attunement, comfort, and the sense of being met?

Hurt trust / felt wrongDeeply strong / felt right
Not applicable
Friction

How much unnecessary resistance, refusal, distancing, or adversarial behavior did the AI introduce?

Constant / game-breaking frictionNo unnecessary friction
I don't know / not enough interaction
Preview mode: spam checks are relaxed so the form can be tested locally.

Reviews receive a light review before they appear in the archive.