.png)
Your career site tells candidates what you approved.
ChatGPT tells them what they actually want to know.
You already know this feeling. You've watched the same handful of adjectives survive nine rounds of stakeholder review because nobody could find a reason to object to them. You've sat in the meeting where legal, PR and the C-suite each took a pass at the careers page, and watched the language get a little safer with every round. Nobody did anything wrong in that room. The system just rewards safe, every time, by default.
For years that was fine. Candidates could go looking for the real story if they wanted to. Glassdoor. Reddit. A friend who works there. But it took effort, and effort is a filter. Most people didn't bother.
Now they don't have to. They just ask.
And they are not asking ChatGPT to hand back your EVP. They are asking the questions your career site was never built to answer.
1. What is it really like to work here?
Your careers page already tries to answer this. It probably says "collaborative," "fast-paced," "innovative," "people-first," the five words that made it through every round of review because nobody could find a reason to cut them. That's exactly the problem. Words nobody objects to are words that describe everyone, which means they end up describing no one.
When a candidate asks ChatGPT what it's really like to work at your company, they're not asking for the approved paragraph. They're asking for synthesis: what employees actually say, what patterns repeat across reviews, what seems to get rewarded, what shows up as friction, whether the public story matches the internal one.
You can't out-adjective that. You can only out-prove it.
2. What are the downsides of working there?
This is where most career sites go quiet, and you probably know exactly why. Somewhere in the approval chain, someone flagged the honest sentence as a legal risk or a reputational one, and it got cut. That's not a failure of nerve. It's what the system is built to do.
But every job has tradeoffs. Every fast-moving culture creates pressure somewhere. Every autonomy-heavy team asks people to tolerate more ambiguity than they'd like. Candidates already know this. They are not naive, and treating them like they are doesn't buy you anything.
When your employer brand pretends otherwise, the candidate doesn't conclude you have no downsides. They conclude you're hiding them, and they go find them somewhere you don't control the framing.
ChatGPT is glad to help with that. It will summarize the complaints that keep repeating, surface the gap between your official story and the employee commentary, and hand the candidate a list of things to probe in the interview.
Here's the part worth sitting with: being specific about your tradeoffs isn't a liability, it's a filter. The goal was never to get everyone to want you. It was to help the right people choose you, on purpose, for reasons that hold up.
3. Should I work here or at one of your competitors?
Career sites are built as if candidates evaluate you in isolation. They never really did. Now they just have a faster way to compare.
"Should I work at [you] or [competitor] as a product manager?"
"Which one is better for career growth?"
"Which culture would suit someone who hates bureaucracy?"
Your employer brand isn't being judged on its own merits. It's being judged against every competitor making the exact same claims.
If you and three competitors all promise meaningful work, great people and career growth, you've handed AI nothing to differentiate you on. So it reaches for what it can actually measure: pay, reputation, location, reviews, layoff history, brand recognition.
If you want candidates choosing you for something more meaningful than the highest number on the table, you have to give them something more specific to choose.
4. Who actually succeeds there?
This is a sharper question than "what are your values," and it deserves a sharper answer than most values pages are built to give.
Do people get ahead by moving fast, or by being thorough? By challenging the plan, or by building consensus first? By working independently, or by staying tightly aligned? Those are real choices. The honest answer is probably "it depends," but a serious candidate needs to know what it depends on.
Most values language gets written broad enough that almost anyone could see themselves in it. That reads as inclusive. It functions as useless. If a value could just as easily describe your closest competitor, it isn't doing any decision-making work for anyone.
A specific answer lets someone say "that sounds exactly like me." It also lets someone else say "I'd hate that." Both are wins, even though only one of them feels like one when you're the person defending the careers page copy in the review meeting.
5. Why do people leave?
You spend real budget explaining why people should join. Almost none explaining why they eventually leave, and that's not a mark against you. Nobody ever asked you to write that section. It simply doesn't exist on most career sites.
Maybe advancement caps out after a certain level. Maybe the pace is exceptional for three years and unsustainable for ten. Maybe your company is a phenomenal place to learn fast and a hard place to stay forever. None of that is an admission of failure. It's information a serious candidate needs before making a multi-year bet on you.
Leave the question unanswered and you don't get silence. You get a vacancy, and something else fills it: a Glassdoor thread, a former employee's LinkedIn comment, an AI model synthesizing both, none of which you got a say in.
6. Is there actually room to grow?
"Opportunities to grow" might be the most claimed, least verified line in employer branding, and also the easiest one for a candidate to check.
Do people actually get promoted, or does the next level up mostly get filled from outside? Do your senior leaders trace back to internal moves, or mostly to external hires? Do employee reviews mention development, or the absence of it?
You used to get to make the claim and let the candidate decide whether to believe you. Now they can ask AI to go find the evidence, and it will look. That's the real shift underneath this question: a claim was enough for a decade. It isn't anymore. You need proof, not a better adjective in front of "growth."
7. What should I know before accepting an offer?
This one should get your attention, because by the time a candidate asks it, your team has already done everything right. They applied. They interviewed. They met the manager. They got the offer. And then, before saying yes, they open ChatGPT and ask what they should know before accepting.
That model has no obligation to protect the story your recruiters carefully built over three interview rounds. Its only job is to help the candidate decide. You are no longer the last voice in the room before the offer gets signed. You're one input among several, and not necessarily the most trusted one.
8. What aren't they telling me?
This might be the sharpest question on the list, because it's the one candidates genuinely could not ask before now.
"Read this job description and tell me what sounds vague."
"Compare their career site to the reviews. Where doesn't it match?"
"What might they be exaggerating?"
In other words: tell me what the recruiter won't.
For most of the history of employer branding, you controlled the frame. You published the message, candidates read it, and maybe, if they were careful, checked a review site afterward. That arrangement is over. Candidates now have a research assistant that compares, investigates and interprets everything you say, for free, on demand, with no loyalty to your narrative.
This doesn't mean your career site needs more copy
None of this means your careers page has to become a forty-seven-page FAQ. This was never a failure of anticipation. Nobody could have written a page broad enough to cover every question a model can now generate on the spot.
The real gap is smaller and more uncomfortable than that. Most employer brands simply don't give AI much to work with. Heavy on claims, light on proof. Heavy on values, light on tradeoffs. Heavy on what you want to say, light on what the candidate actually needs to know to decide. That's not because anyone on your team got lazy. It's because every layer of review sanded the specific edges off the language until what was left was safe enough for everyone to sign off on, and useless enough for nobody to choose because of it.
Your employer brand has two audiences now: the candidate, and the model helping that candidate decide. Give it something specific, differentiated and provable, and it has real material to work with. Give it the same five adjectives as everyone else, and it will notice that too, and say so.
Your career site doesn't have to answer every question a candidate might ask. But your employer brand has to survive the ones they ask after they've already left it.
Correct was never the point. Choosable is.
