How to spot a bullshitter: proof, hidden subsidies, and fake expertise

Trust was the first product

One of my earliest clear lessons came from a man I knew at my college gym. He was friendly, offered spots, and talked with me regularly. The relationship felt organic until the business pitch arrived. He wanted me to involve my mother in a referral-based opportunity that, after research, had the structure and incentives of an MLM.

The friendliness may have been genuine. The sequence still mattered. Trust lowered resistance before I had evaluated the opportunity. I nearly accepted the recommendation because I was evaluating the person I knew in the gym instead of the compensation model he was asking my family to enter.

Independent research broke the spell. Later 'be your own boss' and group recruiting calls became easier to identify because I changed the questions. I stopped asking whether the presenter seemed nice. I started asking who paid whom, what produced the payment, which expenses came first, and how the typical participant performed.

Find the incentive before evaluating the sentence

Manipulation becomes easier to see when the transaction is redrawn. The pest-control representative is not merely discussing insects. The solar representative may be selling a financed installation. The car dealer can make more when attention stays on the monthly payment instead of the vehicle price, add-ons, interest rate, and total paid. A creator can earn from the link regardless of whether the product remains useful after the return window.

The incentive does not automatically invalidate the advice. I use affiliate links too. It changes the evidence burden. A person paid when I say yes should not be the only source defining the problem, selecting the options, calculating the savings, and declaring the deadline.

My first questions are structural: What action pays you? Is the payment larger for one option? Who carries the downside if the projection fails? Which number would make you recommend that I do nothing? A qualified adviser can answer without treating the questions as disloyalty.

The rich twenty-something claim has a missing balance sheet

Social media is excellent at showing outcomes and terrible at showing capital structure. A young person can stand beside an expensive car, work from a resort, display a revenue dashboard, or say they escaped the 9-to-5. Those images leave out the questions that matter. Who owns the asset? Who funded the runway? Is the number revenue or profit? How much debt exists? Who provides housing or health insurance? Did the result survive longer than one unusually good year?

Parental or partner support is not shameful. Living at home, sharing expenses, receiving education, using a family network, joining a partner's benefits, or taking capital from someone who believes in you can be rational. The problem begins when the support disappears from the story and the result is sold as a replicable blueprint for people who do not have it.

People do not owe strangers a complete balance sheet. But when the visible lifestyle is being used to sell advice, access, a course, an investment, or a worldview, the missing denominator becomes decision-relevant. I do not infer fraud. I downgrade the claim until the economics can be separated from the costume.

Short experience is evidence, not mastery

A person can produce an excellent result in one or two years. They can also mistake one favorable cycle for a durable skill. A profitable launch does not prove that the model survives competition. One property does not create a full housing-cycle record. One bull market does not establish investment expertise. A year of training does not reveal how the method handles plateaus, injury, adherence, or long-term progression.

Dunning-Kruger is often used online as a sophisticated way to call someone stupid. That is not a useful reading. The original work concerned calibration within particular tasks: limited domain knowledge can also limit a person's ability to recognize the gaps in their performance. It is not a diagnosis for young people, and later research disputes how much of the familiar chart comes from a special metacognitive deficit versus measurement and regression effects.

The practical lesson survives without the meme. Ask people to calibrate their claim. What exact domain do they understand? For how long? Across which conditions? What failed? What would falsify the conclusion? A credible person narrows the claim faster than a bullshitter expands it.

• Credit a real result without automatically granting expertise in adjacent domains.
• Separate a repeatable process from one favorable market, employer, platform, relationship, or timing window.
• Look for corrections, failed predictions, abandoned methods, and decisions the person now handles differently.
• Require a longer record when the advice concerns irreversible health, debt, careers, legal exposure, or other people's money.

Why Mr ROI is not an escape-the-matrix victory lap

Someone can reasonably look at a 23-year-old running a site called Mr ROI and ask whether this is the same performance with better typography. That is a fair challenge. My answer is not that I am uniquely immune to overconfidence. It is that the site makes a narrower claim and leaves a record that can be checked.

I am not claiming that I am rich, financially independent, debt-free, self-made in a vacuum, or an expert on every category I cover. I had meaningful advantages. I started college unusually early and lived at home until 21, which created financial runway instead of a rent bill. I completed an applied-mathematics degree and work in Amazon supply chain.

I also have a useful but bounded base of firsthand evidence. I have trained consistently since 2018 and driven my Tesla roughly 60,000 miles. I have also spent years owning, modifying, breaking, repairing, and comparing the home, theater, fitness, technology, and vehicle systems I write about.

That is an unusual head start and a useful evidence base. It is not decades of wisdom. I have not lived through retirement, raised children, managed every economic cycle, or personally owned every product I discuss. The site should distinguish firsthand experience from outside evidence. It should also label illustrative numbers, unfinished outcomes, and questions I cannot answer yet.

Mr ROI is a method and evidence ledger, not a declaration that I have maximized life. Its credibility comes from exposing the complete decision and documenting real purchases and tradeoffs. I cite sources, disclose monetization, publish WAIT and PASS conclusions, preserve corrections, and let recommendations fail when the evidence changes. If the site stops doing those things, the criticism becomes correct regardless of my résumé.

MLM claims need denominator discipline

A recruiter can show a real high earner and still create a false impression about the opportunity. The missing denominator is everyone who joined, spent money, worked, earned little or nothing, and left. Gross commission is not profit. Revenue before product purchases, travel, events, subscriptions, lead costs, taxes, and unpaid time is not take-home income.

Current FTC guidance says earnings claims need a reasonable basis and that atypical results require clear information about what typical participants earn and spend. If the company cannot provide a current written income disclosure with participant counts, zero-earner treatment, typical expenses, and net outcomes, I do not fill the gap with testimonials.

The practical test is to rebuild the opportunity without the presenter. List every mandatory and culturally expected cost. Count recruiting and selling hours. Use the median or a distribution, not the best screenshot. Compare the result with a conventional part-time job, freelance skill, or certification that builds portable career capital.

• Request the current income disclosure and compensation plan before paying or providing contacts.
• Ask how many participants earned zero and whether inactive people are excluded from the headline average.
• Subtract purchases, fees, travel, events, refunds, taxes, and unpaid hours.
• Separate customer retail demand from purchases made to qualify for compensation.
• Reject any model that turns family trust into an inventory of leads before the economics are proven.

AI made the costume of competence almost free

A polished memo, formula, code block, sales proposal, or technical explanation used to imply at least some effort. Generative AI can now produce the appearance in seconds. That is useful when a capable person uses it to accelerate work they can verify. It is dangerous when the output becomes a substitute for understanding.

I do not try to solve this with an AI detector. The FTC finalized an order against one detector after alleging that a 98 percent accuracy claim was unsupported and that independent testing on general-purpose content was close to a coin flip. Even a good detector answers who or what may have written the text, not whether the conclusion is correct.

NIST's AI risk framework emphasizes testing, evaluation, verification, and validation. I apply the same idea to people. The artifact can be AI-assisted. The owner must still show how it was built, what data it uses, where it fails, and what they would do when the case changes.

The six tests that make understanding observable

No single test proves competence. Together, these six force a person beyond memorized fluency. They also protect real experts who communicate imperfectly, because a quiet specialist can demonstrate the work without winning a performance contest.

Run a live change request

The fastest workplace or vendor test is a small, relevant change. If someone presents a forecast, change one input and ask how the result should move before recalculating. If they present SQL, ask for the grain, join keys, duplicates, and a test for missing records. If they present a recommendation, remove one constraint and ask which option enters the shortlist.

A bullshitter usually protects the artifact. A practitioner works the model. They may need time, documentation, or another expert, which is normal. What matters is whether they can state what they know, what they need, and how they will validate the answer.

Do not confuse fast speaking with competence. Some subject-matter experts communicate poorly under pressure. Give them another path: a written walkthrough, sample, reference, prior result, or reproducible test. The goal is to reduce uncertainty, not reward the most socially fluent person.

• For analysis: request data grain, definitions, exclusions, reconciliation, and one independently checked sample.
• For code: request a fresh test case, failure handling, dependency explanation, and ownership of the deployed result.
• For a contractor: request license when applicable, insurance, scope, exclusions, recent comparable references, and milestone acceptance criteria.
• For a creator: request primary sources, sponsorship disclosure, product duration, failure experience, and who should not buy.
• For a job candidate: use a bounded work sample with disclosed AI rules, then discuss choices and tradeoffs live.

Urgency is usually a request to skip verification

Most major consumer decisions do not become invalid tomorrow. The salesperson controls the discount, appointment, or inventory story and can often create another version. Real deadlines exist, but they should be independently verifiable: a published tax rule, auction close, flight departure, contract expiration, or scarce physical inventory with evidence.

My default for meaningful money or time is no same-call decision. I collect the written offer, leave, compare, and answer later. The first timeshare pitch nearly worked because the environment compressed our thinking. Becca's refusal to commit before research restored the time horizon and changed the result.

At work, urgency can also become coercion. In security jobs, leadership sometimes pushed undesirable hours and pay as if no alternative existed. Escalation can correct a local abuse, but the durable answer may be finding a better role when the organization's incentives will not change. Verification does not require staying indefinitely in a bad system.

My one-page verification protocol

For any high-consequence recommendation, I fill one page before deciding. It prevents charisma, fear, and AI polish from controlling the process.

Stay skeptical without becoming impossible

The point is not to assume everyone is lying. That creates a different failure: endless research, hostility toward legitimate expertise, and inability to delegate. I trust people faster when they disclose incentives, separate facts from estimates, show evidence, name failure conditions, and welcome verification.

I also set the evidence burden by reversibility. A $30 reversible purchase does not need an interrogation. A long loan, health intervention, job change, home system, or partner agreement deserves more work because the downside is larger and harder to undo.

AI has made presentation cheap. That should shift status toward the less glamorous parts of competence: clean inputs, reproducible work, accurate boundaries, useful predictions, and ownership after the meeting. I do not need every expert to sound impressive. I need their answer to survive contact with reality.