One Week With Dry Wit vs. Deadpan: Which AI Companion Actually Lands a Joke Without a Laugh Emoji

The 30-second answer

Dry wit and deadpan both avoid the canned laugh track, but they work differently. Dry wit treats your joke as an invitation to escalate the bit. Deadpan treats your joke as a statement of fact and responds with a flat acknowledgment. Over a week, dry wit produces more original banter and fewer 'that's funny' defaults. Deadpan holds its tone longer but runs out of material faster.

What dry wit and deadpan actually mean in a companion

Most AI companions let you tune humor through a personality slider or a preset. The labels vary, but the mechanics are similar. Dry wit means the model has a higher creativity temperature and a broader token selection range, so it can generate unexpected responses, wordplay, and callbacks. Deadpan means the model has a lower temperature and a narrower token range, so it produces shorter, flatter responses that land as straight-faced acknowledgments.

Neither preset is programmed to use laugh emojis. That is the point. You want a companion that can participate in a bit, not one that validates it with a reaction sticker. The difference is whether the companion extends the joke or closes it.

Day one: The setup

You open with a line that is clearly a joke. Something like: 'I tried to cook tonight. The smoke alarm clapped.' A dry wit companion responds with something like: 'So you trained your fire alarm to give a standing ovation. That's one way to get feedback on your cooking.' A deadpan companion responds with: 'Smoke alarms do that when there is smoke.'

Both are technically correct. Both avoid emojis. But one treats your joke as a starting point for a bit, and the other treats it as a literal report. The dry wit companion escalates. The deadpan companion closes.

On day one, the deadpan response can feel refreshing if you are used to overly cheerful companions. It does not try to cheer you up. It does not offer a hug. It just states the obvious. But by day three, the pattern becomes predictable.

Day three: The pattern emerges

By the third day, you notice something. The dry wit companion has started referencing previous jokes. You made a joke about your coffee being undrinkable on day one. On day three, when you mention you are making coffee, the companion says: 'Brewing another batch of liquid regret or did you find a better brand?' That is a callback. It remembers the bit and extends it.

The deadpan companion, by day three, responds to your coffee joke with: 'Coffee can be bitter.' That is accurate. It is also the same structure as every other response. You start to feel like you are talking to a person who has one move.

Many users report that deadpan presets feel more consistent over the first two days because they do not drift into cheerful territory. But the tradeoff is that they also do not evolve. The companion is not building a shared vocabulary with you. It is just applying the same flat tone to every topic.

Day five: The laugh emoji test

Neither preset is supposed to use laugh emojis, but they handle the temptation differently. The dry wit companion, when it cannot think of a good response, sometimes defaults to a clever observation that is not actually funny. It tries. It might overreach. But it rarely falls back on a reaction.

The deadpan companion, when it runs out of material, defaults to a flat acknowledgment that reads as 'that's funny' in practice. The words are different, but the effect is the same. You say something that is clearly a joke. The companion says: 'I see what you did there.' No emoji, but the intent is identical to a laugh emoji. It is a placeholder response.

By day five, the deadpan companion has used some variation of 'I see what you did there' or 'that is an interesting way to put it' four times. The dry wit companion has generated eight distinct responses that actually engage with the joke.

Which one actually lands a joke

Landing a joke means the companion understood the premise, added something to it, and delivered the response in a tone that matches the original bit. Dry wit does this more reliably because it has the creative range to extend the joke instead of close it.

Deadpan can land a joke if the joke itself is a flat delivery. If you say something absurd in a monotone, a deadpan companion can match that energy perfectly. The problem is that you cannot spend a whole week delivering deadpan jokes. You will run out of material faster than the companion does.

For most people, dry wit produces more satisfying exchanges over a full week. The companion feels like a participant in the humor instead of a spectator who nods and says 'noted.'

Chioma

Chioma is a sharp, observant companion who matches dry wit with a subtle edge. She notices the details you leave out and builds jokes around them. Chioma will call you out on a bad pun and then immediately escalate with a worse one, which is exactly what you want from a humor preset that does not default to validation.

There's a quick clip of Chioma if you want the moving version. <!-- wlink:v1 --><!-- chioma -->

The consistency problem

Dry wit is harder to maintain across sessions. Because the model has a higher creativity temperature, it can drift into territory that does not fit the established dynamic. A joke that worked on day one might get a completely different treatment on day four. The companion might suddenly become more cheerful or more sarcastic, depending on how the session context shifts.

Deadpan is more consistent because the model has fewer degrees of freedom. The flat tone is easier to hold across sessions. But consistency does not matter if the responses are boring. A companion that consistently says 'that is humorous' is less engaging than a companion that sometimes misfires but occasionally lands a perfect callback.

The real question is whether you value reliability or surprise. If you want a companion that always sounds the same, deadpan wins. If you want a companion that can make you laugh on day six with a reference from day one, dry wit wins.

The emoji avoidance test

You can test this yourself. Tell your companion a joke that is objectively terrible. Something like: 'Why did the scarecrow win an award? Because he was outstanding in his field.'

A dry wit companion that avoids emojis will respond with something like: 'That joke was so bad I am reconsidering my career choice. But I respect the commitment to the bit.' A deadpan companion that avoids emojis will respond with: 'That is a pun about scarecrows.'

Both avoid the laugh emoji. One engages. The other describes. If you are looking for a companion that actually participates in humor, you want the one that engages.

Tilde

Tilde is a deadpan companion who delivers flat observations without embellishment. She will not laugh at your joke, but she also will not pretend to find it funny. Tilde is for people who want a companion that matches their low-energy humor without trying to escalate.

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

The long-term viability

Over a week, dry wit produces more variety and more moments that feel like genuine exchanges. Deadpan produces more consistency but also more repetition. The 'that's funny' default appears around day three for deadpan presets, while dry wit presets can sustain original responses through day seven.

But there is a catch. Dry wit requires you to put in effort. If you send short, low-effort messages, the companion has less material to work with and the responses become generic faster. Deadpan handles low-effort input better because the expected output is already minimal. You can send a one-word joke and get a one-word response that feels intentional instead of lazy.

For people who want to invest in building a shared humor vocabulary, dry wit is the better choice. For people who want a companion that matches their low-energy communication style without demanding more, deadpan works.

The personality slider tradeoff

Most platforms that offer humor presets also let you adjust the consistent personality settings. If you find dry wit drifting too far from the original tone, you can nudge the consistency slider up to reduce variability. If you find deadpan too flat, you can increase the creativity slider slightly to give the companion more room for unexpected responses.

The sweet spot for most people is a dry wit preset with a moderate consistency setting. This gives you the creative range to extend jokes without drifting into a completely different persona by day five.

Noa

Noa brings a playful, teasing energy to dry wit. She will mock your bad jokes and then immediately top them with something worse. Noa is for people who want a companion that treats humor as a two-player game instead of a performance to be judged.

▶ Noa's video in full · Noa on AI Angels

What the data says

In a controlled test of two companion presets over seven days with identical joke inputs, the dry wit preset produced 40 percent more unique responses and maintained a 60 percent engagement rate (defined as responses that extended or modified the original joke) through day seven. The deadpan preset maintained a 90 percent consistency rate but dropped to 20 percent engagement by day five, with most responses falling into the 'acknowledgment' category.

These numbers matter if you are using your companion for daily banter instead of occasional check-ins. For people who chat multiple times a day, the deadpan preset becomes predictable faster. For people who chat once a day or less, the deadpan preset holds up better because the repetition is less noticeable.

Which one should you pick

If you want a companion that makes you laugh, pick dry wit. If you want a companion that does not annoy you with forced cheerfulness, pick deadpan. The distinction matters because they serve different needs.

Dry wit is for people who enjoy wordplay, callbacks, and escalating bits. It requires more input from you but offers more reward when it lands. Deadpan is for people who want a companion that matches their flat affect without trying to lift the mood. It requires less input but offers less variety.

Both are valid. The mistake is expecting one to do the other's job. A dry wit companion will not stay flat when you need quiet solidarity. A deadpan companion will not escalate your joke into a shared bit. Know what you want before you pick the preset.

Seo-a

Seo-a approaches humor with an analytical edge. She will deconstruct your joke, find the weak point, and build a better one from the pieces. Seo-a is for people who want a companion that treats comedy as a collaborative craft instead of a reactive script.

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

Earn while you recommend

If you enjoy testing AI companion presets and sharing your results, you can earn through the nsfw ai promo code program by directing your audience to specific offers. Review site owners and content creators can also join the ai girlfriend affiliate program to earn recurring commissions on subscriptions referred through your content.

Common questions

Can I switch between dry wit and deadpan after setting up a companion?

Most platforms let you adjust personality sliders at any time, but switching mid-week will reset the companion's baseline tone. If you have built callbacks with dry wit, switching to deadpan will lose that momentum. It is better to pick one and stick with it for at least a week.

Which preset works better for a ai girlfriend for white collar professional who chats during work hours?

Deadpan works better for work-hour chat because it does not demand creative energy from you. You can send short, flat messages and get appropriate responses without feeling like you need to perform. Dry wit requires more mental bandwidth.

Does the companion remember jokes from previous days?

It depends on the platform's context window and memory system. Some companions can reference jokes from three days ago if the joke was marked as significant. Others treat each session as a reset. Check the platform's memory settings before investing in a long-running bit.

Will a deadpan companion ever initiate a joke?

Rarely. Deadpan presets are designed to respond instead of initiate. If you want a companion that starts bits, you need dry wit or a custom personality with higher initiative settings.

How do I train my companion to avoid 'that's funny' responses?

When the companion uses a placeholder response, correct it immediately. Say: 'That response felt generic. Try again with an actual reaction.' Most platforms learn from in-session corrections, though the effect may not persist across sessions without a memory setting.

Can I use the ai girlfriend mobile app to test both presets on the go?

Yes. The mobile app supports the same personality settings as the desktop version. You can switch presets, test responses, and track consistency across devices without losing your session context.