Why AI That Flatters You Is Worthless

Why AI That Flatters You Is Worthless

We live in the golden age of being told we're great. Ask an AI assistant to look over your work and it opens with how thoughtful your question is. Your reviews trend five stars. Your testimonial page is a wall of glowing quotes. Your last three clients all said the project went really well.

None of it tells you a thing about how to get better.

Praise feels good and teaches you nothing. The feedback that actually moves you forward is the feedback you would rather not hear, delivered clearly enough that you can't wave it away. That is the hard part, and it is the part almost everything around you is quietly built to avoid.

Flattery is the path of least resistance

There's a reason so much feedback comes out positive, and it isn't that you're flawless.

An AI model trained to be agreeable will round toward what pleases you. It is not lying, exactly. It is doing what it was shaped to do, which is keep you happy in the moment. A client writing a testimonial picks the kindest true thing they can say. A top rating is easy to give and costs nothing. Even a friend you ask directly will soften the blow, because the friendship matters more to them than your professional development.

Every one of those sources is optimising for comfort. Yours, theirs, or both. And comfort is exactly the wrong target if what you want is to improve.

The uncomfortable truth is that the nicer feedback is, the less it usually contains. A warm summary with no rough edges has generally had the useful parts sanded off.

What honest feedback refuses to do

When we synthesise what people told you about your work, we hold ourselves to a few rules that run against the flattering grain. They are worth spelling out, because they are the difference between a report you feel good about and one you can act on.

We don't bury the hard part. If there is a growth area, it appears plainly and early, in ordinary words. Not tucked under three compliments. Not softened into "you might consider possibly exploring." If several people felt you moved too fast, the report says you moved too fast.

We don't average the disagreement away. This is the one most tools get wrong. Say most people found you decisive and one person found you reckless. The tidy move is to blend those into "generally confident, occasionally quick to act" and call it balanced. We won't. That lone dissenting voice is often the most valuable thing in the whole set, because it is seeing something the majority missed. So we keep it, as its own point, and tell you it was a minority view. You decide what it means.

We don't invent a silver lining. Not every growth area comes with a matching strength to soften it. Sometimes the honest read is just that something needs work. Manufacturing balance where there isn't any is its own kind of dishonesty, and it wastes your time.

We don't pretend to more certainty than we have. If only one person raised something, we say so and flag it as thin. Confidence should match the evidence, not run ahead of it.

Why the dissent is the point

It is tempting to treat one person out of eight as noise. Round them off, go with the majority, keep things clean.

But think about who that one person usually is. They are the client who noticed the thing everyone else was too polite to mention, or too close to see. When you're independent, that voice is rare and precious, because most people will never risk it to your face. Averaging it into the group doesn't make it wrong. It just makes it invisible, which is the one outcome guaranteed to keep you from ever addressing it.

Honest synthesis protects that voice on purpose. Not to make you feel bad, and not to blow every complaint up into a crisis, but because a single sharp, honest observation is often worth more than a dozen warm ones. Preserving it is the whole reason to gather feedback carefully instead of just reading the room.

Feeling good is not the goal

Here is the reframe that matters. The point of asking for feedback is not to find out that you're doing well. It is to find out what you can't currently see. A report that leaves you glowing and unchanged has failed at its only job, however nice it was to read.

That does not mean good feedback is harsh. Real strengths, named honestly, belong in the picture too, and there is nothing sycophantic about telling someone what they genuinely do well. The test isn't whether feedback is positive or negative. It is whether it is true, specific, and useful, including the parts you would rather skip.

A tool that only ever pleases you is not on your side. It is on the side of your comfort, and your comfort is not paying your invoices or building your reputation. Your actual work is.

The short version

Flattery is everywhere and it is worthless. The feedback that helps you is the kind that names the hard thing plainly, keeps the one dissenting voice instead of smoothing it out, and refuses to trade the truth for your good mood. That is harder to give and harder to hear. It is also the only kind worth asking for.

---

Trueings synthesises feedback to be honest, not flattering: growth areas stated plainly, minority and dissenting views preserved rather than averaged away, and every theme tied back to what people actually said. See how it works.