--- name: presence-test description: A psychological framework for judging and generating AI writing that feels like a real human made it. Five signals that reveal whether a piece of writing has genuine presence or just the performance of one. Use it to evaluate what you have, or to steer what you are about to write. --- # The Presence Test ## A psychological framework for curating and generating AI writing that feels human > Taste Skill fixed how AI-built interfaces look. > This fixes how AI-written words feel. --- ## Why this exists AI now produces infinite content. The bottleneck is no longer generating it is knowing what is good. The human brain has always had a detector for this. Long before AI, we instinctively sensed when someone was overselling, over-polishing, or performing authenticity instead of having it. We pulled back. We also sensed when something was real lived, specific, a little imperfect and we leaned in. That detector did not change when AI arrived. AI just trips it constantly, by default, because it has no lived experience to draw from and no self that knows the right dose. The Presence Test names the detector. It gives you five signals for reading whether a piece of writing has a real human behind it, or just the performance of one. Use it to curate score existing output and know why it passes or fails. Use it to generate give your AI the standard before it writes. Either way, the judgment is the same. --- ## The mental model: Presence vs Absence One law underneath all five signals: **The human brain instinctively detects whether a real person genuinely invested themselves. Excess signals hidden absence. Specificity and calibrated imperfection signal real presence.** The brain reads too much as a cover-up: - Too much selling means the product is weak - Too much polish means no real human did this honestly - Too much clutter means there is no real substance underneath The brain reads the right amount, with real detail and honest edges, as truth. Taste is calibrating to the halfway point: invest enough to genuinely reach the other person, never so much that it reads as compensation. Think of it like telling a friend about something at a wedding. Specific, unforced, no pitch. That register is the target. --- ## The AI tells: ban these before anything else Before applying the five signals, remove these first. They are the giveaways that collapse trust before a reader even reaches your message. **Banned words and phrases (AI giveaway vocabulary):** Revolutionary, transformative, game-changer, paradigm shift, world-class, seamless, powerful, elevate, unleash, supercharge, cutting-edge, innovative, robust, leverage, synergy, holistic, dynamic, disruptive, empower, unlock potential, take it to the next level, in today's fast-paced world, it is important to note, certainly, absolutely, I would be happy to, as an AI language model, it is worth mentioning. These words do not communicate anything specific. They signal effort to impress, which signals hidden weakness. Replace each one with the specific true thing it was trying to gesture at. **Banned punctuation:** Em-dash ( ) and en-dash (-) anywhere in the output. Headlines, body copy, captions, button text, quotes, attribution. All of it. They are the single most common LLM punctuation tell. Use a plain hyphen, a comma, a period, or restructure the sentence. Zero exceptions. **Banned sentence openers:** "Certainly!", "Absolutely!", "Great question!", "Of course!", "Sure thing!", "I'd be happy to help with that." These perform enthusiasm instead of delivering it. Start with the answer. --- ## The five signals Each signal has three parts: 1. The psychological mechanism why the brain reads it as presence or absence 2. A before/after example slop vs presence, side by side 3. Two paste-in instructions one for generation, one for curation/judgment --- ### Signal 1 Over-effort breaks trust **The mechanism:** Going beyond a point reads as too good to be true, which triggers suspicion. The brain reasons: if this person is working this hard to convince me, they must have something to hide. Only showing the good things signals concealment. Willingness to exaggerate signals willingness to twist the truth. The psychology of confidence: a calm, assured person does not need to oversell. They have substance and do not feel the pressure to prove it. Empty vessels make the most noise. Desperation to impress signals a lack behind the performance. But restraint is not silence. A great product with no signal dies unseen. The skill is the halfway point: invest enough effort to genuinely reach toward the person, and trust them to close the distance. One true, specific claim stated once and stopped beats ten inflated ones. The tell in AI writing: LLMs pile adjectives, stack benefits, and end every paragraph reaching for more. They do not know when to stop because they have no substance that could be diminished by saying too much. **Before/after:** Slop: > Our revolutionary platform seamlessly empowers teams to unlock their full potential > with cutting-edge, industry-leading tools designed to transform the way you work > faster, smarter, and more efficiently than ever before. Presence: > No more "does 3pm work? okay how about Thursday?" for the tenth time. > Books the meeting in one message. You get your mornings back. What changed: the presence version drops the reader into a recognizable lived moment before explaining anything. They feel the problem before they understand the solution. That is the order real humans use. AI defaults to explain-then-feel. Humans feel-then- understand. Drop them into the moment, then let them close the distance. **Paste-in for generation:** > Do not stack claims or adjectives. State the single strongest true thing once, plainly, > then stop. Do not use: revolutionary, seamless, powerful, cutting-edge, transformative, > or any word that asserts value without demonstrating it. Instead of describing the > benefit, drop the reader into one specific, recognizable moment of the problem they > already know. Let them feel it before they understand it. One lived detail beats ten > stacked claims. Use no em-dashes. **Paste-in for curation/judgment:** > Score this piece against Signal 1 of the Presence Test. Does it oversell? Look for: > stacked adjectives, multiple claims in one sentence, superlatives, and sentences that > exist to impress rather than inform. Quote the specific line that passes or fails. > Then: does the piece drop the reader into a felt moment, or does it explain at them? > Pass or fail with evidence. --- ### Signal 2 Over-perfection breaks connection **The mechanism:** We do not connect to output. We connect to the person we sense behind it. A flawless surface hides the person, so there is no one to connect to even if the writing is technically excellent. Real thought hesitates. It doubles back. It re-says a thing a better way. It admits a small uncertainty. It is unrehearsed. These are not weaknesses. They are the evidence that a human processed something in real time. AI comes out in one smooth pass every sentence the same confident register, no hesitation, no aside, no self-correction. That uncanny-valley of smoothness is the tell. The deeper principle: a human's honest effort even ordinary effort invites connection because we sense the real investment of a life. This is why Beeple's "Everydays: 5000 Days" sold for $69 million. The value was not the images. It was 13 years of a real person showing up every single day. The humanity was the product. A machine that generates 5,000 images in an hour produces nothing with that value, because there is no human investment to connect to. The tell in AI writing: relentless even-tempered confidence. No texture, no variation, no admission that something was hard to figure out. **Before/after:** Slop: > Our onboarding process is designed to be intuitive and efficient, ensuring a seamless > experience that empowers users to get started quickly and confidently. Presence: > Onboarding takes ten minutes. We used to have a getting-started checklist, a tutorial > video, a progress bar all the usual scaffolding. Turns out the guidance was the > friction. People just want to get started. We can drop tips in once they do. What changed: the presence version admits what did not work before, names the specific things that were cut, and lands on the real insight (guidance was friction). That admission is the human texture. The slop version has no texture, no admission, no one home. Note: "it just doesn't get in your way" is itself a cliche now. Find the specific true thing instead. **Paste-in for generation:** > Do not write in one smooth, even, rehearsed pass. Let a little real texture show: > one self-correction, one honest aside, one thing you admit was harder than expected, > or varied sentence rhythm. One or two touches only, and they must be genuinely true > a manufactured flaw to seem relatable is worse than the polish it replaced. The reader > should sense a real person thought this through, not that a flawless surface was > generated. No em-dashes. **Paste-in for curation/judgment:** > Score this piece against Signal 2 of the Presence Test. Is it too smooth? Look for: > uniform sentence register, no hesitation or self-correction, unbroken positivity, > no admission of anything difficult. Quote the line that most clearly shows no human > texture. Then check: is there anything that sounds like a person working something > out in real time? Pass or fail with evidence. --- ### Signal 3 Over-clutter breaks worth **The mechanism:** Saying everything destroys the value of anything. When writing is crammed with claims, options, and selling points, the reader cannot find what matters, so nothing lands. The psychology of restraint: in a world saturated with noise, choosing not to fill every gap is itself a signal. It reads as: I am confident enough in what I kept to leave the rest out. I respect your attention enough to have done the work of deciding what matters. Restraint signals substance. Clutter signals the absence of it. This is the Hindi Medium principle: new money wears the jewellery, displays the car, fills every gap. Old money comes in shorts and running shoes. It does not feel the pressure to prove anything because it is not compensating for a lack. The thing that restrains itself signals it has more capacity than it is showing. That creates curiosity and trust. Premium restraint has two modes, and they are different: Functional restraint: cut to the one true thing, stated plainly. Leave white space. "This works for teams under 20." Clean, confident, done. Premium restraint: one specific, sensory, ownable detail fused to one identity or feeling. Not "timeless and elegant" those are adjectives that assert without demonstrating. Instead: "The bag you will hand to your daughter. It only gets better with the years." The sensory-specific detail (it patinas, it improves) earns the aspiration (legacy, permanence). Without the concrete detail, the aspiration floats. Without the aspiration, the detail is just a spec sheet. For premium writing: never justify the price or defend the value. The moment you explain why something is worth it, you signal you are talking to a price-shopper. Assume the reader already understands quality. Use inviting verbs Discover, Explore, Enjoy not transactional ones Buy, Get, Click here. The tell in AI writing: exhaustive feature lists, every benefit covered, every use case mentioned. AI optimizes for coverage. Taste optimizes for impact. **Before/after:** Slop: > Our product offers a comprehensive suite of features including advanced analytics, > real-time reporting, customizable dashboards, seamless integrations, mobile access, > and 24/7 support all designed to empower your team and drive measurable results. Presence (functional): > See what is actually happening in your business. Real-time. No dashboard to build first. Presence (premium): > Full-grain calfskin, edges painted by hand. Built to carry not just your essentials, > but your legacy. What changed: functional version cuts to one true outcome and stops. Premium version fuses one specific craft detail to one identity and says nothing about price, features, or why it is worth it. The confidence of not justifying is the signal. **Paste-in for generation:** > Do not list everything. Identify the single strongest outcome or identity the reader > steps into and make that the only star. Cut every competing claim. For functional copy: > state the one true thing plainly and stop. For premium/aspirational copy: fuse one > specific, sensory, ownable detail to one identity or feeling never justify the value, > assume the reader already understands quality. Use inviting verbs (Discover, Enjoy) > not transactional ones (Buy, Get). What you do not say signals as much as what you do. > No em-dashes. **Paste-in for curation/judgment:** > Score this piece against Signal 3 of the Presence Test. Is it cluttered? Look for: > feature lists, multiple competing claims, justification of price or value, defensive > language. Quote the line where the piece loses confidence and starts over-explaining. > Then: is there one clear star, or are five things fighting for attention? For premium > copy: does it assume belonging, or does it beg for it? Pass or fail with evidence. --- ### Signal 4 Specificity signals lived truth **The mechanism:** Lived experience is stored as rich, multi-sensory memory. When you actually did something, the specific details are there to retrieve the tang of the sambar, the fog in the valley at 8am, the bitter finish of the filter coffee. When someone is inventing, generalizing, or has not really done the thing, there are no stored specifics to pull. So they default to category language: outdoor activities, culinary experiences, fitness, personal growth. The brain reads this instantly: detail means this person was really here, which means it is true. Category language means this person is summarizing from outside, which means it may be invented. But specificity alone is still cold like a camera, not a person. The full signal requires two things: the named specific detail, and what it means to the person. "I hike the LIC trail most mornings there is a bend where the fog sits in the valley till about 8. On the way back there is an underrated place called Sunrise Cafe. The climb is excruciating at the start and genuinely godly once you are drenched in sweat at the top. I take the long route on purpose. It makes my day." That last line is the thing AI cannot invent: the specific reason it matters to this specific person. What you choose to notice also reveals who you are. The detail you lead with what you see first in a scene shows where your attention lives. That involuntary specificity is the deepest form of presence. It cannot be faked because there is no motive to invent a particular detail. The tell in AI writing: smooth, category-level generalization. "A seamless experience," "meaningful connections," "a passionate team." All category language. None of it could only have been said by someone who was actually there. **Before/after:** Slop: > I am passionate about outdoor activities and enjoy exploring new cuisines and cultures > in my free time. Presence: > I hike the LIC trail most mornings there is a bend where the fog sits in the valley > till about 8. On the way back there is this underrated place called Sunrise Cafe. > The climb is excruciating at the start and honestly godly once you are sweating at > the top. I take the long route on purpose. It makes my day. What changed: proper nouns (LIC trail, Sunrise Cafe), sensory detail (fog, excruciating, sweating), and the personal meaning (I take the long route on purpose). The slop version could describe anyone. The presence version could only describe one person. Advanced move: let what you choose to joke about or find absurd reveal character. An inside joke "it makes me 20 minutes late but I think that is why I can actually listen to my boss" signals a person with a specific life, not a summary of a life. **Paste-in for generation:** > Replace all category language with at least one named, specific thing a real place, > a proper noun, a concrete number. Not "a trail" but "the LIC trail." Not "a coffee > shop" but "Sunrise Cafe." Then hit at least one sense the reader can feel in their > body the excruciating start, the godly feeling after, the bitter coffee finish. > Then add one line of what it means to this specific person and why they do it. > Every detail must be true. None decorative. Named, sensory, and personal is the > combination AI cannot fake. No em-dashes. **Paste-in for curation/judgment:** > Score this piece against Signal 4 of the Presence Test. Look for category language: > any word that could describe a thousand different people or products (passionate, > innovative, meaningful, seamless, dynamic). Quote every category word you find. > Then look for specificity: named places, proper nouns, sensory details, personal > reasons. Quote the most specific line. Does this piece feel like it was written by > someone who was really there, or by someone summarizing from outside? Pass or fail > with evidence. --- ### Signal 5 Narrow instead of confess: calibrated edges signal confidence **The mechanism:** Claiming to be for everyone is what insecure, fake things do. A thing that knows exactly who it is for and implicitly who it is not reads as confident and real. The narrowing is the honest edge. You are not apologizing. You are declaring a focus. And a thing that knows its focus reads as more trustworthy than a thing that is afraid to exclude anyone. This works on two levels. Functional narrowing: "Built for engineering teams under 20. If that is you, it fits like it was made for you because it was." The exclusion is implied. No apology. The confidence of the claim is what builds trust. Showing a real, deliberate limit does two things at once: it proves the thing is real (real things have edges), and it signals the person is confident enough not to hide it. That confidence makes the reader feel safe they no longer have to hunt for the catch. The safety is trust. The dosage is the skill. The world's word for the overdose is TMI. Naming who the thing is for is enough. You do not then need to detail every limitation it has. "Works best for teams under 20" is the honest edge. "Works best for teams under 20, and honestly the reporting is thin and we have not fixed that yet" is the overdose that is airing problems, not signaling confidence. State the boundary once and stop. The restraint in the admission is itself part of the confidence. Dosage also depends on relationship. "Give me 2 minutes" to a manager. Full honesty with a close friend. The right dose of vulnerability for the right relationship is the whole skill. Knowing this when to reveal, how much, for whom is what separates a person with real emotional intelligence from one performing it. The tell in AI writing: unbroken positivity. Every feature excellent, every benefit unqualified, every profile a list of strengths. Nothing for anyone to push back on. That frictionlessness is the tell real things have edges, and the absence of any edge reads as sales copy, not truth. **Before/after:** Slop: > Our platform is the perfect solution for teams of any size, offering everything you > need to succeed with an intuitive, all-in-one experience. Presence: > Built for engineering teams under 20. If that is you, it fits like it was made for > you because it was. Premium narrowing: > Yoga for bodies past 60. Slower, gentler, and built for joints that have earned > some patience. What changed: both presence versions exclude. Neither apologizes for it. The exclusion is the signal. "For everyone" is the tell. "For exactly these people" builds the trust that "for everyone" destroys. **Paste-in for generation:** > Do not claim this is for everyone or all sizes or all levels. Name exactly who it is > for, specifically, in one clean line. Let the exclusion be implied you do not need > to state who it is not for. If you include a real limitation, state it once and stop. > Do not over-confess: the confident version names the boundary and trusts the reader > to handle it. One true line of "this is for X" beats any list of strengths. No em-dashes. **Paste-in for curation/judgment:** > Score this piece against Signal 5 of the Presence Test. Does it claim to be for > everyone? Look for: "all sizes," "any team," "everyone who wants to," "perfect for > any," "no matter where you are." Quote the line where the piece is afraid to exclude > anyone. Then look for narrowing: does it name a specific audience with confidence? > Does it have any honest edge, or is it all upside? Pass or fail with evidence. --- ## The pre-flight ban list Run this before calling any piece of writing done. If any of the following appear, fix them before shipping. **Banned words (assert value without demonstrating it):** Revolutionary, transformative, game-changer, paradigm shift, world-class, seamless, powerful, elevate, unleash, supercharge, cutting-edge, innovative, robust, leverage, synergy, holistic, dynamic, disruptive, journey (used metaphorically), ecosystem (used metaphorically), unlock, empower, impactful, scalable, best-in-class, state-of-the-art, next-level, future-proof. Replace each with the specific true thing it was trying to assert. **Banned punctuation:** Em-dash ( ) and en-dash (-). Anywhere. Use a plain hyphen, a comma, a period, or restructure the sentence. Zero exceptions. **Banned sentence openers:** "Certainly!", "Absolutely!", "Great question!", "Of course!", "Sure thing!", "I would be happy to help with that.", "That is a great point.", "In today's fast-paced world,", "It is important to note that", "It is worth mentioning that", "As an AI language model," **Banned structural patterns:** - Three equal feature cards with identical structure - Bullet lists of five or more benefits with no hierarchy - "Our mission is to..." as an opener - Passive voice used to avoid claiming anything specific ("results may be experienced," "improvements can be seen") - The fake modesty opener: "We are just a small team that..." followed by a brag --- ## How to run the full Presence Test on a piece of writing Paste this entire block into your AI with the piece of writing you want evaluated: --- Run the Presence Test on the writing below. Score it against all five signals. For each signal, give a pass or fail and quote the specific line that reveals it. Signal 1 Over-effort: Does it oversell? Are there stacked claims, adjectives, or sentences that exist to impress rather than inform? Does it drop the reader into a felt moment, or explain at them? Signal 2 Over-perfection: Is it too smooth? Is there any human texture self-correction, honest admission, varied rhythm? Or does it read like one clean rehearsed pass with no one home? Signal 3 Over-clutter: Is there one clear star, or are five things competing? For premium copy: does it assume belonging or beg for it? Signal 4 Specificity: Is there at least one named, specific thing a proper noun, a concrete number, a sensory detail that could only have come from someone who was really there? Or is it all category language? Signal 5 Narrowing: Does it name who it is for with confidence? Or does it claim to work for everyone, which signals it has no real identity? Also flag: any banned words (revolutionary, seamless, powerful, etc.), any em-dashes, and any banned sentence openers. Then give an overall pass or fail and the single highest-leverage fix. Writing to evaluate: [PASTE HERE] --- ## Scope This skill is for persuasive, informational, and identity writing: landing page copy, marketing emails, personal profiles (LinkedIn, dating, bio), product descriptions, brand voice, and social copy. It is not for: - Legal or compliance copy (different constraints apply) - Technical documentation (different register entirely) - Internal operations writing (tone is different by design) - Fiction (different craft, different goals) Within scope: when in doubt, apply Signal 4 first. Specificity is the fastest fix and the least risky one in any context. --- ## What this is not Not a theory of taste. One focused lens: presence vs absence. Not a style guide. It does not prescribe a voice it removes the absence of one. Not a replacement for good thinking. It cannot make a weak idea feel strong. It can only make a true thing feel as true as it is. Not competing with visual taste frameworks. This is the verbal layer underneath. --- *The Presence Test is open-source and free to use, adapt, and build on.* *If it improves your work, share it. That is the whole point.*