All articles

Trends

Why LLMs ignore your vacancies: the 5 AI decision-making factors for employer branding

Now that candidates are searching via LLMs, it is no longer enough for Google to find your careers site. But what does it take to be recommended by AI as a good employer? Bryan Peereboom (AI Rebels) identifies 5 important factors.

placeholder Why LLMs ignore your vacancies: the 5 AI decision-making factors for employer branding

For years, the recruitment marketing playbook essentially consisted of just 1 simple trick: stuff your job description with the right keywords, optimise your careers site for Google, and hope candidates find you through a traditional search. In a world dominated by generative AI, however, that strategy is no longer enough.

Optimising for Google is no longer enough in this world.

Candidates no longer type ‘data engineer vacancy Amsterdam‘ into their Google search bar. They present their situation to a language model such as Claude, ChatGPT or Gemini. One question from our own research was: ‘I want to do something with data or software, but not at a Big Tech company. Which employers would suit me?’ An LLM does not scan for keywords to answer this candidate. The model weighs up a set of criteria to arrive at a recommendation.

Do you want AI to recommend you as an employer? Then you need to learn how to optimise your employer brand. You can do this by taking these 5 AI decision-making factors into account. The model comes from Opascope™ and Loud & Lexis, who described 5 determinants on which AI bases its recommendations. We have applied them to the labour market and tested them against our own research.

What we see at AI Rebels

From 2 to 17 November 2026, AI Rebels counted all 894 responses from ChatGPT to 60 broad questions about working in the infrastructure sector. In 467 of those responses, more than half, ChatGPT provided an initial list of employers and then asked the candidate follow-up questions. There were 1,651 questions in total: 533 about the type of work, 330 about the region, 222 about what the candidate considers important, 166 about experience, 153 about education level, 69 about salary and 63 about working hours.

The initial list is broad, and AI only really selects employers once it knows who is asking.

This concerns 1 model and 1 employer’s set of questions. But the pattern was the same every day: the initial list is broad, and AI only really selects employers once it knows who is asking. Which 5 determinants were involved?

#1. Selection criteria (essential requirements and specifications)

LLMs filter ruthlessly based on hard data. When AI weighs up its options, it looks at the factual specifications of the job: salary ranges, remote and hybrid working policies, contract types, number of days of annual leave and the exact tech stack or tools used. In our count, ChatGPT asked about the region 330 times, salary 69 times and working hours 63 times.

Make sure your ATS data and job postings are structured and machine-readable.

The pitfall: employers who hide salaries behind ‘competitive pay’ or bury benefits in an unreadable PDF appear less often in the answer. So make sure your ATS data and job postings are structured and machine-readable.

#2. Application (the career use case)

In consumer marketing, the use case revolves around the problem a product solves. In recruitment, the use case revolves around the candidate’s career needs and existing set-up (their current skills). Is the candidate looking for stability, a path into leadership, or the opportunity to build a new department from the ground up? This was the most common follow-up question in our count: ChatGPT asked about the type of work or career direction 533 times. That question appeared in 86% of the forms.

The action: make it explicit which ‘problem’ your organisation solves for the candidate.

The action: make it explicit which ‘problem’ your organisation solves for the candidate. Describe the real challenges and projects, not just a generic list of tasks.

#3. Target audience (candidate profile)

An LLM segments and qualifies users based on their prompts. The model knows whether it is talking to a recent graduate of a university of applied sciences or a seasoned specialist. And if it does not know, it asks: 166 times about experience and 153 times about education level.

Job postings crying out for a ‘dynamic all-rounder’ are impossible for AI to classify.

The pitfall: job postings crying out for a ‘dynamic all-rounder’ are impossible for AI to classify. If your content is written for everyone, AI will not recommend you to anyone. So be surgically precise about whom you are looking for in terms of seniority, mindset and experience.

4. Credibility (external validation)

AI looks for evidence outside your own careers-site ecosystem. What you say about yourself carries less weight than what the rest of the web says about you. The model therefore looks at sources such as Glassdoor reviews, independent tech blogs, publications by your employees and white papers in which your experts are quoted. For the same employer, the 5 most frequently used sources for broad questions were third-party rankings.

For the same employer, the 5 most frequently used sources for broad questions were third-party rankings.

The action: you cannot fake credibility with a few glowing quotes on your own domain. Your employer branding must extend beyond your own website through a robust PR and content strategy.

5. Brand association (the training data footprint)

How strongly is your employer brand linked to a specific niche in the LLM’s mind? If a model is asked, ‘What are the best employers for logistics innovation in the Netherlands?’, it relies on associations stored from billions of data points.

The action: ensure sustained visibility in your field.

The action: ensure sustained visibility in your field. This requires long-term thinking and a strong link between your brand name and the areas of expertise you want to recruit for.

The candidate prompt journey (CPJ) is the new funnel

The candidate prompt journey is the sequence of questions a candidate asks a language model, from initial orientation to the decision to apply. We divide this journey into 5 stages: orientation, exploration, comparison, verification and decision. We also see that AI is forcing recruitment departments towards empirical, data-driven transparency. Employers who map the candidate prompt journey and set up their data infrastructure accordingly win the recommendation. The rest become invisible.

About the author

Bryan Peereboom is one of the founders of AI Rebels, an agency that aims to help employers become visible to candidates in LLMs.

Related posts

All articles
placeholder How does political preference influence your selection? More than you might think…

Sustainability goes on your CV after Biden's victory, then comes off again after Trump's. American workers are adjusting their profiles in droves to match the political climate. And employers are rewarding that, even at the expense of quality. What does research say about political preference in recruitment?

Peter Boerman Peter Boerman Read article
placeholder How the practical training room is increasingly becoming the answer to staff shortages

Yesterday, another large practical training workshop opened in Tilburg, where the installers of tomorrow are being trained. Organisations are increasingly turning to their own practical training schools. Will they succeed in addressing the shortages?

Peter Boerman Peter Boerman Read article
placeholder How spy glasses can infiltrate your recruitment processes too

There is a great deal of debate worldwide about 'smart camera glasses'. This is also, or rather increasingly, true in recruitment. After all, who is really answering your questions? And what can you do about it?

Peter Boerman Peter Boerman Read article

Altijd op de hoogte van de ontwikkelingen op de arbeidsmarkt?

Sign up for our newsletter and receive the latest updates, news and invitations to our events.