We've all hired somebody we thought was fantastic. Smooth, articulate, polished. And then three months in, we looked at them and thought: oh my gosh, what happened to you? This person is a nightmare.

So how did they get all the way through our interview process?

Here's the uncomfortable part. When we asked hiring managers about it afterward, 86% said that in hindsight, there were clues. Words the candidate used in the interview that, looking back, perfectly predicted this person was going to be a terror to deal with.

They heard the clues. They just didn't know what they were listening for.

There's a reason for that, and it's worth sixty seconds before we get to the interesting part.

And once you think about it, it's obvious why. Testing skills is easy. Hand a programmer thirty lines of code and have them find the errors. Hand a nurse a fetal monitor. Skills are black and white.

Testing attitude is the hard part. There's no laptop you can hand someone to measure coachability. So most companies don't really test it at all — they take a gut read and hope.

Which brings us to the interesting part.

Part One — Read this answer

Before I explain anything, try it yourself

I want to show you something rather than tell you about it. Below is a real interview question and a real candidate's answer — an actual response, not something I made up to prove a point.

Read it, and rate the candidate on a scale of one to seven. One means terrible fit. Seven means great fit. Be honest with yourself about your first instinct.

When I ask people what they liked about that answer, the responses are consistent and reasonable. They say the candidate has a learning mindset — they were "always learning on the fly." They mention the problem-solving ability. They note the optimism: there's usually an answer if you look hard enough. Somebody who doesn't believe in failure.

All perfectly sensible. And all of it is wrong.

Part Two — The science

Attitude leaves a fingerprint in the words

Here's where this stops being an art and becomes something you can actually measure.

We asked 1,427 professionals to answer 15 open-ended interview questions as though they were applying for a job. After statistical validation, we had 20,572 answers. Then we had a panel of hiring managers and HR executives grade every one of them: is this a high performer answer, a low performer answer, or somewhere in between?

Then we did something nobody had done before. We took the high performer answers and the low performer answers and ran a linguistic analysis on both — comparing the actual words, the pronouns, the verb tenses.

The differences were staggering.

Notice what the second one does not tell you: whether they ever actually called a customer. It only tells you they know what the best practice would be. That's not an answer. That's a book report.

One more you probably haven't thought about since high school

There's a category of grammar called modal verbs — auxiliary verbs that imply probability and possibility rather than fact. Could. Should. Would. May. Might.

Notice what they're not saying. They're not saying "I did this thing." They're saying "I could do this. I would do this. You should do this." Possibility instead of history. Once you start hearing modals in an interview answer, you'll notice how much of the answer is describing a world that never actually happened.

The 13 words to listen for

Put all that research together and you get a cheat sheet. These are the 13 words that, once you know why they're there, become almost impossible to un-hear:

Absolutes · second-person pronouns · modal verbs · negation · negative emotions · adverbs

If you take nothing else from this page, take this: listen for "I did" versus "you should." That single distinction will change more interviews than anything else here.

Part Three — Back to the answer

Now let's go back to that candidate

Remember the answer you rated? The one 60% of hiring managers approve of? Let's walk through it with what you now know. Here it is again:

And remember one more thing about that answer: this is as good as this person gets. This is their Sunday best. They're trying to get the job. Are people more polished when they're trying to impress you, or six months after they've got the job?

"Toxic" isn't one thing

Worth pausing here, because that candidate raises a question: toxic to whom?

Yes, some people are toxic anywhere — the liars, the cheats. Those exist. But far more often, you're dealing with someone who is toxic in your particular culture and would be perfectly fine somewhere else.

If you have a fast-moving, constantly-changing culture and you hire someone change-resistant, that person will likely become toxic on your team. Put them in a stable, methodical organization and they're great. If you have a deeply collaborative culture and you hire someone highly individualistic, they'll probably become toxic in your environment — and thrive at a company built for individual contributors.

They're not bad. You're not bad. You're just bad together.

This is why you can't copy someone else's hiring. Southwest and Google are as different as night and day. You can't hire like Southwest unless you are Southwest. Which means the first real question isn't "how do I spot toxic people" — it's "what does toxic look like here?" Answer that, and everything else on this page has something to aim at.

Part Four — Your questions

You're giving away the answers

Everything above assumes you asked a question worth answering. Most interviewers don't — and the way they ruin their questions is so small you'd never notice it.

First, the questions to stop asking entirely

Tell me about yourself. What are your strengths? What are your weaknesses? Those are the three most common interview questions in North America, and every candidate has them rehearsed.

"Sure — I'm a motivated self-starter. I love individual accountability, but you know what, I also love working on teams."

"Weaknesses? Well, I don't like to share this too much, but sometimes I care too much. Sometimes I work too hard, and it can be a little intimidating to other people because I'm just so dedicated."

When you ask those questions, you are not testing attitude. You're testing acting ability and memory. They're not evil questions — they're just a complete waste of the twenty minutes you have.

Same goes for hypotheticals. What would you do if the server crashed at 2 a.m.? — "I'd instantly assemble a SWAT team, triage the customer needs, and bring the servers back in stages." Lovely. Now: will they actually answer their phone at 2 a.m.? You have no idea. When you ask a hypothetical question, you get a hypothetical answer.

And the leading question is the worst of all. "We're an incredibly collaborative culture — you're comfortable with that, right?" You just told them the answer and asked them to repeat it back.

Now the mistake that even good interviewers make

Here's the frame that fixes this. There are two kinds of people in the world: problem bringers and problem solvers.

Ask a problem solver about a problem, and they'll tell you about the problem — and then, automatically, with no prompting from you, they'll tell you how they solved it. They can't stop themselves. It's who they are.

Ask a problem bringer about that same problem, and they'll tell you about the problem. And stop.

That difference is pure gold. And you only get to see it if you stay out of the way.

Watch how easily it gets ruined:

Two things went wrong in that first one. The obvious one is "and what did you do?" — you've just handed them the ending and told them to skip to the part where they win. The subtler one is the word adapt. It presumes they adapted. Maybe they didn't. Maybe they sat there paralyzed — and wouldn't you want to know that before you hired them?

Again, two problems. "And did so successfully" tells the candidate: only tell me about your wins. And balance smuggles in the answer — I don't want to know about a time you balanced priorities, I want to know about a time you faced them, and then I need you to tell me whether you balanced them or dropped every one of them on the floor.

The magic words are faced and experienced. They're neutral. They presume nothing.

Part Five — The part nobody does

Interviewing is a test. Where's your answer key?

Here's the thing that ties all of this together, and it's the step almost every company skips.

Interviewing is a test. You're giving candidates a test. So what's the point of administering a test if you don't have an answer key that lets you grade it?

An answer key — we call them Answer Guidelines — is simply a set of real-life good and bad answers to your interview questions, with an explanation of what makes each one good or bad. That's it. That's the whole idea. And it's the difference between hiring managers who guess and hiring managers who know.

Because here's what happens without one. Go back to that candidate you rated at the top of this page. In a live session with hundreds of hiring managers, the ratings for that answer spread across the entire scale. Some people give it a 2. Plenty give it a 6.

The fix is what I call a calibration exercise, and it's the single most valuable thing you can do with your hiring team this month. Take a real answer. Have everyone rate it independently. Then show the spread and talk through it together.

The first time a team does this, ratings are all over the map. By the second or third or fourth round, everyone lands within a point of each other. That's the goal. Not perfect agreement with me — internal consistency, so that the same candidate gets the same evaluation regardless of who happens to be in the room.

There's no other way to train this that works. You have to look at real candidate answers, and you have to make people evaluate them out loud.

One more reason this matters now

Candidates are using AI to game your interviews. That's not a prediction — it's happening right now.

So how do you stop it? You do what teachers do. If I give you a take-home assignment, I cannot prevent you from using AI — detectors are borderline useless and at worst actively biased. But if I get you in a room and hand you one of those old blue exam books, we're back to reality.

Which means the asynchronous video interview — where the candidate gets time to compose an answer off-camera — is now the worst tool you own. You've given them five quiet minutes with ChatGPT and a script to read back to you.

The counterintuitive conclusion: as AI gets better, live human interviewing skill becomes more valuable, not less. The ability to hear a rehearsed answer in real time is about to be one of the most important skills in your organization.