When a Flaw Becomes a Credential
In December 2025, McDonald's Netherlands pulled its AI-generated Christmas ad, “It's the Most Terrible Time of the Year,” three days after posting it on December 6, after a wave of online criticism. It was a big-budget production; the studio said it took seven weeks. Viewers still recognized it as AI-generated, called it creepy and ‘AI slop’ — slop being the word Merriam-Webster went on to name its 2025 Word of the Year — and the brand took it down within days. Production effort no longer read as care.
Hold that against what I see in the same stretch of the market: in my view, flawless AI-made output has gone from a draw to a liability. Something that had been an asset for the entire history of advertising inverted.
The usual reading is that people want "authenticity" now, so brands should rough up their content — unmade beds, uneven lighting, a creator who looks like they just woke up. I think that reading is wrong, and acting on it can be expensive. The research explains why, and it has been sitting in the literature since 1966.
I think perfection was never the point. Cost was.
For most of the last century, a flawless brand asset was proof of something the audience could not see directly: that the company had spent real money and real attention to produce it. A thirty-second film shot on 35mm, a print layout with no stray kerning, a product photo lit by someone who knew what they were doing — each carried the unstated message that this was a serious operation with resources to protect. Polish was a costly signal. It worked because it was costly.
Generative tools removed the cost. Perfect is now free, instant, and available to anyone with a prompt box. When a signal becomes free to produce, it stops distinguishing the people who could once afford it from the people who could not. Worse, it starts to carry the opposite inference. A spotless image in 2026 no longer says "a serious team made this." It says "a machine made this, and no one checked." The same mechanism sits behind the disclosure penalty that hits undisclosed AI the moment it is detected.
This is the ground the pratfall effect is standing on. And it is why the effect, which used to be a clever tactic, is becoming a structural advantage.
The coffee spill
In 1966, the social psychologist Elliot Aronson, with Ben Willerman and Joanne Floyd, ran an experiment with a recording of a man answering quiz questions. In one version the man is brilliant — he answers roughly 90 percent correctly and mentions an impressive record of achievement. In another he is average. Then Aronson added a second variable: in half the recordings, the man clumsily spills a cup of coffee on himself at the end. Four conditions in total — competent or average, clumsy or not. Listeners rated how much they liked him.
The competent man was liked more when he spilled the coffee than when he did not. The blunder made him more appealing, not less.
But the version of the finding I usually hear quoted stops one condition too early. When the average man spilled the coffee, listeners liked him less. The same mistake that made the expert endearing made the ordinary performer simply clumsy. The pratfall does not humanize anyone who commits it. It only helps the person who has already demonstrated they are good — because for them, and only for them, the flaw reads as a flourish they can afford rather than evidence of their limits.
Marketing research found the same boundary in products. Ein-Gar, Shiv, and Tormala (2012) showed that a small piece of negative information can raise evaluations, but only when it comes after the positive information and the buyer is not scrutinizing closely. The flaw helps only after competence is shown.
That is the whole mechanism, and it is the part the "rough up your content" crowd has lost. A flaw alone is not a trust signal. A flaw on top of proven competence can be one. Strip out the competence and you are left with the average man and a wet shirt.
The 4.2 problem
The cleanest proof that this holds at commercial scale is sitting in your reviews. Northwestern's Spiegel Research Center (in a summary it published in 2021) analyzed purchase behavior across more than 40 product categories and found that likelihood of purchase does not peak at a perfect rating. It peaks between 4.2 and 4.5, and then declines as the average climbs toward a perfect 5.0.
Read that carefully. Beyond a certain point, a higher rating lowers conversion. The shopper looking at 5.0 across the board does not conclude "flawless product." They conclude "filtered, curated, or fake." A spread of ratings with a few ones and twos in it does the opposite — it certifies that the reviews are real, that the brand did not scrub them, and that someone with nothing to hide is comfortable leaving the blemish visible. In my experience, shoppers deliberately seek out the negative reviews before buying. They are not looking for reasons to walk away. They are looking for the coffee spill, because its absence is the thing that scares them. It is the same reason a wall of identical top-rated praise now moves fewer buyers than a smaller, more credible signal does: consensus that looks manufactured persuades few.
The implication for anyone managing ratings is uncomfortable: the instinct to suppress, dispute, or drown every critical review is actively costing conversion. I think a mix of positive and negative reviews tends to sell better than a spotless record. The perfect record is the weaker asset.
Why brands often run it backwards
Knowing the effect and deploying it are different problems, and there are three reliable ways to deploy it wrong.
They scrub everything
The default failure. Every case study is a triumph, every testimonial is a rave, every photo is retouched, every objection is handled before it is voiced. The brand reads as either dishonest or inhuman, and in 2026 the audience has a word ready for it. This is the McDonald's ad — technically perfect, and rejected by its audience.
They fake the flaw
The humble-brag. "Our only weakness is that we care too much." "We're growing so fast we can barely keep up with demand." These are not admissions; they are compliments wearing an admission's coat, and buyers tend to see through them. A fake flaw is worse than no flaw, because it adds a second data point: not only is the brand unwilling to show a real weakness, it thinks you cannot tell the difference.
They confess before they qualify
The subtlest error, and the one the authenticity trend is manufacturing at scale. A new brand leads with its mess — the chaotic founder story, the "we have no idea what we're doing," the deliberately amateur aesthetic — before it has shown the audience any reason to believe it is competent. This is the average man spilling the coffee. Without the 90-percent quiz score first, the mess is not disarming. It is just mess. The unmade-bed school of content marketing works for creators who have already earned an audience's confidence and backfires for everyone who has not, which is most of the brands copying it.
The discipline
In my experience, the sequence that works is competence first, then one specific, costly, verifiable flaw.
Competence first. The audience has to have a reason to rate you an expert before the flaw can register as a flourish. Guinness earned the right to say "good things come to those who wait" — an open admission that its pour is slower than everyone else's — only because the beer behind the wait was already respected. Volkswagen's "Lemon" and "Think Small" worked because the engineering was genuinely good; the ads pointed at a small, honest limitation sitting on top of real quality. Avis told the market "We're number two, so we try harder" from a position where the product already functioned. In each of these cases the flaw is the second thing the brand says, not the first.
One flaw, and a specific one. "We're not perfect" is not an admission; it is a hedge, and it signals nothing. "This mouthwash tastes bad" is an admission. Listerine built a category on it. Specificity is what makes a flaw credible, because a brand hiding something does not volunteer precise detail about its weak spot.
Make it costly, not convenient. The flaw has to be one it would have been easier to hide. A real one. The entire signal rests on the audience's recognition that you chose transparency when concealment was available and cheaper. A flaw that costs you nothing to admit buys you nothing.
Attach the reason. The strongest pratfalls do not just confess — they explain why the weakness is the price of a strength the buyer wants. The pour is slow because the beer is better. The mouthwash tastes bad because it is doing something. The flaw and the benefit are the same fact seen from two sides. That framing turns an apology into a proof point.
What the market is actually paying for
The brands winning the authenticity shift are not the ones that look the messiest. They are the ones willing to show a flaw they could have hidden — and able to do it from a position of demonstrated competence. Apple ran the credit "This show was made by humans" on a series that was, first, good. I think the signal is rarely the imperfection by itself. It is the imperfection as evidence that a capable, confident operator stood behind the work and did not need to hide.
Perfection is free now. Because it is free, it certifies nothing, and many audiences have learned to distrust it on sight. The flaw you can afford to show — one specific, costly, true thing, offered on top of work that has already earned the right — is one of the few signals left that money cannot fake and a model cannot generate. In a market where everyone can look flawless, the brand willing to be seen with the coffee on its shirt is the one that reads as real.
If you want to find the one flaw your brand can afford to show, and the proof it would buy you, book a 30-minute call. Become InPhluential™.
