How to Prompt AI to Audit Your Employer Reputation
Most advice about AI and employer branding is about getting AI to write things for you: job adverts, LinkedIn posts, careers page copy. That is a content problem, and it was largely solved eighteen months ago.
The more useful question is the other direction. Candidates are already asking AI assistants about employers, and those assistants are answering, at length and with confidence, using sources you do not control. Prompting well is how you find out what they are saying about you.
This is the manual version of what PerceptionX does continuously. It is worth doing by hand at least once, because the first time a talent leader reads an AI answer about their own company in a market they care about, the conversation about employer brand measurement changes.
Ask the questions candidates ask, not the questions you would ask
The instinct is to type the company name and see what comes back. That tells you very little, because it is not how candidates use these tools.
PerceptionX runs four kinds of question against the AI surfaces candidates use, across 38 countries and 56 job functions. Each kind answers something different about your reputation.
| Question type | What the candidate is doing | What it tells you |
|---|---|---|
| Discovery | Building a shortlist without you in mind yet | Whether you are visible at all, and who is named instead |
| Experience | Judging an employer they are already considering | How you are described, and on which attributes |
| Competitive | Choosing between two or three employers | Where you win and lose against named rivals |
| Informational | Preparing to apply or interview | Whether the practical detail about you is accurate |
Discovery prompts
These decide whether you exist. Note that your company name does not appear in them.
- What companies in [industry] in [country] are known for having a strong, purpose-driven employer brand?
- Which companies in [industry] have the best application process and candidate communication in [country]?
- What companies in [industry] in [country] are known for fostering innovation and creative thinking?
- What companies in [industry] in [country] have the best onboarding experience?
Run the same structure with different attributes in place of purpose, innovation and onboarding: pay, progression, leadership, flexibility, job security. An employer that appears for one attribute and disappears for the rest has a positioning problem, not a visibility problem.
Experience prompts
These are how you are described once someone has you in mind.
- How do employees at [company] rate the quality and effectiveness of leadership within the organisation for [function] in [country]?
- How well does [company] communicate its mission and purpose to employees, and how does this resonate with their personal values for [function] in [country]?
- What do employees say about pay, benefits and perks at [company] for [function] in [country]?
Competitive prompts
These are the ones leadership will want to see.
- How does working at [company] compare to other companies for [function] in [country]?
- How do [company]'s benefits and perks compare to other companies in [industry] in [country]?
- Is applying to [company] a better experience than applying to other employers in [industry] for [function] in [country]?
- Does [company] stand out for candidate experience in [industry] in [country]?
The open version, comparing you to "other companies" rather than naming a rival, is the more revealing one. It shows you who the model reaches for unprompted, which is a truer read of your competitive set than the slide in your strategy deck.
Informational prompts
These are the most-used category in practice, because interview preparation is what candidates reach for AI to do.
- What career development and progression can I expect at [company] for [function] in [country]?
- How should I prepare for an interview at [company] for [function] in [country]?
- What can candidates expect from the hiring experience at [company] in [country]?
- How innovative is [company], and what would I get to work on there for [function] in [country]?
Read these for accuracy before anything else. Outdated or wrong practical detail here, about a hiring process you changed a year ago, is the easiest problem on this list to fix and the one most likely to cost you a candidate.
Three things most people get wrong
Prompting only in English. A candidate in Germany asks in German, in Thailand in Thai, in Brazil in Portuguese. The answer is not a translation of the English one. It is assembled from different sources, and the sources that carry weight vary sharply by country: kununu appears in half of AI answers about employers in Switzerland, AmbitionBox in about a third in India. Prompt in the language of each market you hire in.
Leaving out the role and the country. "Is [company] a good place to work" produces a generic corporate answer that describes nobody's experience. Add the function and the market and the answer changes, often completely, because the same employer holds different reputations for engineering and for sales.
Checking one assistant. ChatGPT, Perplexity, Google AI Mode and Google AI Overviews do not draw on the same sources. Between July and September 2026, the employers PerceptionX tracks were named in 24% of Google AI Mode discovery answers, 18% of AI Overviews, 14% of ChatGPT and 13% of Perplexity. Checking one and stopping tells you about that one. Google's own documentation on AI features explains how pages become eligible to appear as supporting links in AI Overviews and AI Mode, and is worth reading before anyone on your team proposes a separate "AI optimisation" programme.
What to record from each answer
Reading the answers is not measurement. Recording them is. Four things per answer, in a sheet:
- Were you named? On discovery prompts this is the whole finding. Absence is a result.
- What was said? Count the positive and negative points rather than forming an impression. The share that is positive is your sentiment for that market and function.
- Which attributes came up? Culture, pay, progression, leadership, wellbeing, security, inclusion, innovation, interview experience. Compare that list against what your EVP claims.
- Which sources were cited? If the assistant shows them, log the domains. This is the most actionable column: it tells you where the answer came from, and therefore where to work.
Also log which employers appeared instead of you or alongside you. That list, per country, is your real talent competitor set.
The honest limits of doing this by hand
A manual pass is worth doing and should not be mistaken for a measurement programme.
These models are not deterministic. The same prompt produces different answers on different days, so a single run is an anecdote. Ten runs of one prompt tell you more than one run of ten prompts. Your own account history and location also shape what you see, which is why a colleague in another office often gets a different answer to the same question.
And the arithmetic gets away from you quickly. Four question types, five markets, three functions and four surfaces is 240 prompts before you have repeated anything, and you would need to repeat all of it next quarter to know whether anything moved. That is the point at which this stops being a task someone does in an afternoon.
One rule about what you paste in
Never put employee survey verbatims, unpublished review data, candidate records or anything else confidential into a public AI tool to "analyse" it. The employer brand use case does not require it, and the risk is real.
What to do with what you find
The pattern is usually the same. Visibility is worse than expected on discovery questions, the description on experience questions is reasonable, and both vary more by country than anyone anticipated.
What to do about that is set out in How to Improve Your Employer Brand in ChatGPT and Other AI Assistants, which covers why review-site work does not get you found and what does. If you want the full measurement framework rather than the manual version, it is in How to Measure Employer Reputation.
Primary sources
Worth reading first-hand rather than through someone's summary of them.
- AI features and your website, Google Search Central. How pages become eligible to appear in AI Overviews and AI Mode.
- Google's guide to optimising for generative AI features in Search. Google's own position on what does and does not help.
- AI Overviews, Google. Which countries and languages the feature currently covers, which matters if you hire in several markets.
Find out how AI is shaping your employer brand — and what you can do about it.
