Employer brand in AI search results - What candidates hear when they ask AI about you

  • news
Recruiting tech talent

A service engineer thinking about a move no longer starts with Google by default. He asks ChatGPT instead: is this a good employer for an engineer, what does it pay, what is the shop floor actually like? Whatever comes back is your employer brand in AI search results. And you did not write a word of it.

Research by PerceptionX among 306 job seekers across seven countries, Germany included, found in May 2026 that 72 percent use AI before deciding whether to apply and that 82 percent say an AI answer has changed their view of a company as a place to work. At the same time, 58 percent had caught a model giving inaccurate information about an employer. Small sample, clear direction: you are being described to candidates whether you take part or not.

What does a candidate actually hear when they ask AI about you?

That is not a theoretical question. You can ask it yourself this afternoon. Take five questions a candidate really asks. Is this a good employer for a process operator? What does this company pay? What does the hiring process look like? Who are the alternatives in this region? Why do people leave?

In practice we see three kinds of answers. The first is emptiness: the model does not know you and names your competitor instead. The second is inaccuracy, such as an outdated collective agreement, a merger that closed two years ago or a vacancy that no longer exists. The third is the most dangerous, because it reads perfectly well: an average answer full of words like driven, ambitious and informal. Exactly the words every other employer in your sector uses. You do not claim a culture, you let people feel it, and a model can only retell what it has read somewhere.

Why do you not control your employer brand in AI search results?

Because most of the material an answer is built from is not yours. In early 2026 Omniscient Digital analysed more than 23,000 sources behind AI answers to 240 brand related prompts, put to ChatGPT, Perplexity and Gemini among others. Earned media accounted for 48 percent of the cited sources: independent media, forums and social platforms, review sites and directories. Content owned by the brand itself came in at 23 percent.

The most telling number is in the breakdown. For prompts about reviews, the earned media share rose to 82 percent. And a candidate question, what is it like to work there, is a review question by definition.

None of that is new. It is the Image pillar of the employee value proposition suddenly speaking a great deal louder. What the outside world says about you gets summarised and handed back as fact to someone who has never spoken to you.

What material does a model need to describe you well?

This is where our view of AI sits: AI is excellent for scale, people are essential for substance. A model cannot retell what nobody ever wrote down, filmed or reviewed. It will not invent the engineer who explains why he is still there after eleven years.

So the work sits at the front end. Employee stories with a name and a job title, not anonymous quotes. An honest description of shift work and what it means at home, because a job comes home with you. Video in which the shop floor speaks instead of the management team. Answers to the questions candidates really ask, in plain language, somewhere public and readable.

And the reverse holds too: synthetic content adds nothing here. A careers page written entirely by a model gets summarised by another model straight back into the same average. Why invent someone you could simply film? We wrote about that trade off earlier in our piece on AI in employer branding.

What belongs on your careers site?

The questions candidates put to AI should be answered literally on your own site. In the PerceptionX research the most common research topics were pay (53 percent), career opportunities (52 percent) and the interview experience (51 percent). Those are precisely the three subjects missing from most careers pages.

In practice that means a salary range instead of the word competitive, a real description of the roster, the number of steps in the process and how long someone waits for a reply. Per job family, on one findable page, not buried in a pdf or behind an application form. What else a strong careers website needs, we covered earlier.

Does this work differently in engineering, construction and logistics?

The Dutch market is less tight than it was two years ago. Statistics Netherlands counted 375 thousand open vacancies against 396 thousand unemployed in the second quarter of 2026, or 95 vacancies per 100 unemployed people. Trade had 70.5 thousand openings, manufacturing 29.4 thousand, construction 28.2 thousand and transport and storage 14.4 thousand. Less scarcity simply does not mean the right people walk in by themselves.

What really differs in these sectors is search behaviour. Glassdoor analysed twelve million job seekers and found mobile search highest in food services (64.9 percent), transport (63.1 percent) and retail (60.2 percent), against 35.3 percent in software development. That study dates from 2018 and 2019, so the absolute numbers are higher today, but the pattern has only become clearer.

Someone checking his phone between two jobs does not read three pages. One summarised answer is exactly what he wants. That makes the quality of that answer more important in blue collar sectors, not less.

Is your employer brand in AI search results also a filter?

It should be. You want to attract talent, but the right talent. If the answer is prettier than reality, you hire people who discover in week three that the commute is ninety minutes and that shift work really is shift work.

So it is no problem at all if an AI answer says the work is physically demanding or that you are outside in winter. That lets the wrong candidate drop out before he costs your planner a day. And the candidate journey does not stop at the signature: honest recruitment keeps people longer and hands you internal ambassadors in return.

Is this a campaign question or something else?

Usually something else. Not every question is a campaign, and this one is not. This is a content question and sometimes a structure question: what exists about you in findable places, is it still accurate, and can a candidate find the answer without already knowing your brand?

So start with the question and with a baseline. Draw up a fixed set of ten prompts, record what the models answer today and repeat it every quarter. Check your analytics for how much traffic your vacancy pages get from AI referrals. And simply ask new colleagues where they found you. Without a baseline you cannot show afterwards what it delivered, and without that evidence you will not get internal budget for the next step.

How does MonFri work on your employer brand in AI search results?

We start with your question. Are the service engineer vacancies not filling up, are you losing too many people in the first year, or does the market simply not know you exist? Then we look at what the outside world says about you, what your own people say and what your competitors for talent offer. Those three together form the employee value proposition, and that is the basis for everything a model will later retell about you.

After that we make the material that does not exist yet: stories from the shop floor, video and pages that genuinely answer candidate questions. We use AI for variants, tests and speed, never for the substance itself. Curious what that means for your organisation? Have a look at our services, call 020 23 38 587 or email hi@mon-fri.com.