Summary: The panic that candidates now ask an AI, get one verdict, and form one impression is mostly fiction. It assumes a clean funnel and a level of trust candidates don't extend to a machine. A friend who actually worked somewhere beats the AI summary every time. The real and smaller thing: AI search composes a dossier instead of listing links, and it rewards legibility, plain, readable, consistent story across the open web, over production value. A model quotes what it can read, so a text Reddit thread can outrank a gorgeous JavaScript career site. That's a craft problem worth fixing. But legibility only gets you read, not chosen. Differentiation still comes from true, specific, sometimes unflattering things a competitor can't copy because they happen to be true about you. Don't buy a program to "own AI search." The work underneath the acronym is the job you should have been doing anyway.

Is AI search really a new threat to employer brand?

Mostly no. Every few years the industry finds a new thing to be terrified of, and the terror always arrives pre-packaged with a nifty platform or app or dashboard you can buy to make it go away. This year's thing is AI search. The pitch goes like this: candidates no longer visit your career site; they ask a machine, the machine renders a verdict; and if you haven't "optimized for AI answer engines," then some model is out there badmouthing you to a candidate who will never darken your funnel. Cue the panic. Roll the six-figure retainer.

I want to walk this one down the middle of the street in broad daylight, because there's a real thing buried in here and there's a lot of dude-bros in top-and-bottom-matching tracksuits selling you protection you don't need.

What's the racket?

You're being told job search used to run through channels you could see: the career site, the job board, the LinkedIn post. You knew where the candidate was standing. Now, the story goes, the candidate stands nowhere. They open a chat window, type "what's it like to work at Acme Amalgamated Industrial, and how does it stack up against the other two," and a language model hands back a tidy paragraph assembled from Glassdoor, Reddit, a stale news article, and whatever else it could scrape. The beautiful, expensive employer brand gets squeezed into one sentence, sitting next to two competitors nobody asked to be compared against, and the whole exchange happens in a bedroom, or kitchen office you're not allowed into.

And look, that's not nothing, but, like the shell game, watch what the pitch assumes closely: the candidate has one or so questions, goes to one super-brain (which of course would never get anything wrong), gets one answer, and forms one impression. It assumes a funnel fully complete, start to finished app and onboarded last week.

Job search has never been a funnel. It's a scavenger hunt run by someone with four tabs open and a friend on text. These days, sipping a vodka on ice, not even a martini for flavor.

The fiction: AI is the new front door

Does anyone actually research a job through a single AI answer? No. A real person sees a role on LinkedIn. Texts a friend who used to work there. Half-reads a Glassdoor page and decides the one-star reviews are all from the same bitter guy. Googles the CEO. Asks an AI to rewrite their resume, ignores half of what it says. Applies through some niche board a Slack group told them about. Forgets the whole thing for three weeks. Gets a recruiter email and makes a gut call on the tone of it in about four seconds.

Where in that food fight mess is "the AI summary" the deciding input? It isn't. It's one touch point, or interaction, at best, among fifteen, and not the one carrying the gold star of decision influence. The idea that it's a distinct, ownable channel worthy of its own budget line is mostly an old pitch playbook disguised as a new playbook in a fancy hat so somebody can sell you a hat. Or a bridge in Brooklyn. I have 3 for sale.

If a candidate is cross-checking the machine's verdict against a friend who actually worked there, the friend wins. Every time. Fluency is not authority. The AI answer gets the same skeptical squint as everything else on the screen, and most people, correctly, trust the paragraph a machine wrote about a company roughly as much as they trust the company's own recruiting copy. Which is to say: provisionally, and with one eyebrow up.

The fact: search stopped listing results and started composing dossiers

Here's the part I won't wave away, because it's real and it's small and both are true at once. Google handed you ten blue links and made you do the picking. You assembled your own impression, which meant you owned the bias in it. The AI answer skips that. It hands you a verdict you didn't build. That's a difference in kind, not degree, and it's worth understanding even if it isn't worth panicking over.

The consequence that touches your work isn't about trust. It's about legibility, which is a fancy way of saying what a machine can scrape easily. A model quotes what it can read. If your whole story lives inside a gorgeous, art-directed career site that's ninety percent video, motion, and JavaScript, the machine may reach past all of it and quote a two-year-old Reddit thread instead, because that thread is plain text and your masterpiece is not. The employer brand that "wins" the AI summary isn't the one with the best film. It's the one that's most readable, and most consistent about its story, across the open web where the machine is actually looking.

But you should be yelling out the problem in that logic. Just because it's readable doesn't make it likable, trustable, or actionable. Same old problem that's existed for, oh, the last 30 or so years. That's a craft problem. It's our problem. And it's genuinely useful to fix, because fixing it pays off on every interaction, thing read, thing watched, thing heard, not just the chat window.

The hustle: everything with "AEO" in the name

Careful, the matching top and bottom tracksuits are coming in. The moment "make your story legible and consistent" becomes a real, boring piece of hygiene, someone is going to dress it up, slap "Answer Engine Optimization for Employer Brand" on the deck, and sell it back to you at five times the price. Same underlying work. New acronym. Bigger invoice. Remember that bridge for sale? Should I send you the brochure?

Don't buy the acronym. The work underneath it is legitimate: consistent messaging across the properties a machine can actually read, plain-text versions of the story that lives in your video, a career site that isn't a locked box to a crawler, your narrative told the same way on the sites that get quoted. That's just doing the job well. It has always been the job. It doesn't need a new name, and the second it gets one, the markup arrives with it.

The tell is simple. If someone's selling you a program to "own AI search," fuggedaboutit. If someone's helping you make your client's story clear and consistent everywhere it lives, that's the same thing you should've been doing before any of this, and the AI summary is just one more reason it was worth doing.

So what actually gets a candidate to choose you?

The candidate-side AI panic is mostly fiction. It assumes a funnel that doesn't exist and a trust that candidates don't extend. The real thing is lower-key: one touchpoint now composes rather than lists, and it rewards legibility over production value. Worth understanding, worth building for, in the same breath that you refuse to pay a premium for it under a new name.

But legibility just gets you read. It doesn't get you chosen. And this is the part the tracksuits will never sell you, because you can't put it on a dashboard.

Differentiation, the real kind, was never in the margins. It's not the ping-pong table, the "fast-paced environment," the unlimited PTO nobody takes. Everybody has those, and the AI will happily list them for all three companies in the same flat sentence, because on paper they are the same sentence. Fighting over that inch of margin is how you end up sounding exactly like the two competitors you got compared against.

The distinction that actually moves someone is the stuff underneath: whether the place tells the truth about itself, whether the people are people you'd want in the trench, whether there's real care in how they treat the work and each other. That doesn't come through in a feature list. It comes through in the specifics, the true, tangible, sometimes unflattering specifics that only exist because someone did the research to find them and had the nerve to say them plainly. A machine can read that. A candidate can feel it. A competitor can't copy it, because it happens to be true about you and not about them.

So the work isn't optimizing for the machine. The machine is just the newest set of eyes on the same old question: is there something real here, told by people who mean it, or is there a tracksuit? Get the real thing right, say it clearly, and every component, collateral, asset, story, and experience takes care of itself, the chat window included.

I'd rather tell a client the truth. Don't take my word for it either. Ask the AI's. Maybe not Grok. The scary story is easier to charge for. It's just not the one that's true, and the boring true one happens to be the work we should be doing anyway.