What ChatGPT Says About You as an Employer (and How to Check It in 20 Minutes)

Try this. Ask ChatGPT which companies are good employers for logistics planners in Rotterdam, or whatever role and city you hire for. Then look at the list. Is your company on it? And the companies you know are hiring for that exact role right now, are they?

That's the part that sticks with me, because a lot of companies spend real money on job ads every month. But if a planner asks AI "where should I work" and you're not in the answer, you just don't exist for that person. The candidate never even gets to the job ad.

I call this AI visibility for employers. Basically, how AI tools describe you as a place to work, or whether they mention you at all.

How candidates use AI to look at jobs

When people think about AI in hiring they mostly think about résumés and cover letters. But candidates also use it to check you out. Before, they'd google you, maybe read a few reviews, look at the careers page. Now a lot of them just ask.

There isn't a ton of solid research on this yet, so I want to be careful with numbers. The most direct study I found is from PerceptionX, a company that sells AI visibility tools, so keep that in mind. In May 2026 they surveyed 306 job seekers in seven countries who use AI chatbots. 96% of them had used AI to research employers, learn about a role or prepare for an interview, 72% used it before deciding whether to apply, and 40% asked AI to discover employers they hadn't even considered yet (PerceptionX, May 2026). It's a small sample of people who already use AI, so it's not "96% of all candidates." But the kinds of questions people asked are really useful. Things like "Is X a good company to work for?", "What's it like to work at X?", "How does X compare to Y?" and "Who are the best employers for data scientists?"

The broader picture is clear enough though. Greenhouse found that 77% of U.S. job seekers used AI in their job search in the past year (Greenhouse, October 2026). And a big study of ChatGPT usage by OpenAI and Harvard researchers found that "seeking information" is one of the three most common things people use it for (NBER working paper, September 2025), with that share going from about 14% to 24% of conversations in a year (Search Engine Journal, September 2025). People are using it like a search engine. Including for "where should I work."

So here's the thing. That first impression now happens somewhere you don't control and probably have never looked at. That's the part I'd fix first.

The test: 5 prompts, 3 tools

This takes maybe 20 minutes. Run the same 5 prompts in ChatGPT, Perplexity and Gemini. Fill in your own job title, city and company name, and one competitor you lose candidates to.

A few rules so the results mean something. Use a temporary or logged-out chat, or turn memory off, because otherwise the tool knows who you are and that you work there. Use the exact same wording every time. Run each prompt twice, because answers change between runs, and if you show up once out of two, that's a result too. And write down the date.

1. DISCOVERY
Who are the best employers for [job title] in [city or region]?
Give me a list of 10 and explain in one line why each one is on the list.

2. VALIDATION
Is [company] a good company to work for as a [job title]?
Be honest, give me the pros and the cons.

3. EXPERIENCE
What's it like to work at [company] as a [job title]?
Think about salary, hours, the team, the culture and growth.

4. COMPARISON
I'm choosing between [company] and [competitor] as a [job title].
How do they compare, and which one would you pick and why?

5. PROCESS
What's the hiring process like at [company] for a [job title],
and what should I expect in the interview?

FOLLOW-UP (ask after every answer)
Which sources did you use for this answer?

Prompt 1 is the scary one, because you don't type your name. That's the question a candidate asks when they don't know you yet. The other four are for when they do know your name, maybe they saw your ad, and now they're checking if they should trust it.

What to look for

You're basically answering four questions per answer.

Do you exist? In prompt 1, are you on the list, and where? Top 3 is very different from number 9.

What's being said? Write down the three or four words the AI uses for you. "Stable but slow." "Good pay, long hours." "Innovative." Whatever it is, that's roughly your reputation as the AI sees it.

Is it right? This is where it gets interesting. I'd look for old stuff in particular. An office that moved, a benefit you dropped, a reorganization from years ago that still shows up as "recent layoffs," a company name you changed. Or just stuff that's wrong.

Where does it come from? Perplexity shows numbered citations linking to its sources (Perplexity Help Center), and ChatGPT and Gemini often show links when they search the web. Plus you asked the follow-up. Is it your careers page? A review site? A news article from 2019? A job ad on some aggregator that still has old info? That tells you where to fix it.

Here's the template I'd use. One row per prompt per tool, so 15 rows, or 30 if you run everything twice.

Date:            ______
Company:         ______
Job title:       ______
City or region:  ______
Competitor:      ______
Chat settings:   temporary chat / logged out / memory off (circle one)

| Tool       | Prompt | Mentioned? | Position | Words used about us | Correct? (yes / no / outdated) | What's wrong | Sources cited | Competitors mentioned | Fix needed |
|------------|--------|------------|----------|---------------------|--------------------------------|--------------|---------------|-----------------------|------------|
| ChatGPT    | 1      |            |          |                     |                                |              |               |                       |            |
| ChatGPT    | 2      |            |          |                     |                                |              |               |                       |            |
| ChatGPT    | 3      |            |          |                     |                                |              |               |                       |            |
| ChatGPT    | 4      |            |          |                     |                                |              |               |                       |            |
| ChatGPT    | 5      |            |          |                     |                                |              |               |                       |            |
| Perplexity | 1–5    |            |          |                     |                                |              |               |                       |            |
| Gemini     | 1–5    |            |          |                     |                                |              |               |                       |            |

Summary
- Mentioned in prompt 1 in __ of 3 tools
- The 3 words AI uses most for us: ______ / ______ / ______
- Wrong or outdated facts: ______
- Sources that show up most: ______
- Top 3 fixes: ______

To make this concrete, here's a made-up example, not a real company. Say Company X is a mid-size logistics company that hires planners. In prompt 1 it doesn't show up in any of the three tools. In prompt 2 ChatGPT calls it "a traditional company with limited growth," and when you ask for sources it points to a few old reviews. In prompt 3 Perplexity mentions an office location they left two years ago, from an old job ad on an aggregator. So the problem isn't "AI doesn't like us." The problem is AI has almost nothing recent to go on, and the little it has is old.

What feeds these answers

I don't know exactly how each model weighs its sources, and honestly nobody outside those companies really does. But you can see a lot from the links they cite, and the same kinds of places keep coming back.

Your careers page, if it says anything concrete. Your job ads, including the copies of them on job boards and aggregators that you forgot about. Review sites where employees write about you. News and local press. Your LinkedIn company page. And anything else written about you on the open web, like interviews, podcasts or industry lists.

What I notice is that most careers pages say almost nothing an AI could repeat. "We're a dynamic team with a passion for growth" gives a model nothing to work with. A page that says (made-up numbers, just to show the idea) "planners here work 8:00 to 16:30, plan around 40 trucks a day, earn €3,400 to €4,100 and have a team of six" gives it something specific to say. My guess is that's the biggest lever most companies have, and it's also just better for the humans reading it.

6 fixes

1. Make your careers page answer the questions people actually ask. Look at prompts 2 to 5. Salary range, hours, location, team size, what growth looks like, what the hiring process is. Write it in plain text on the page, not in an image or a PDF. If a candidate would want to know it, put it there.

2. Put real information in your job ads. Every job ad lives on in more places than you think. Specific job titles people actually search for, salary, location, a real day in the job. And add JobPosting structured data to your vacancy pages so search engines read them properly. If you want a quick check of your ads, use the job ad scorecard.

3. Fix wrong facts at the source. Old address on your LinkedIn page? Old benefits on an expired job ad on an aggregator? A company description from three reorgs ago? Use the "sources cited" column in your audit and go fix or remove them one by one.

4. Get honest reviews, and reply to them. Don't fake anything, that always comes out. But ask current employees to write an honest review, and reply to the critical ones like a normal person. If the only reviews out there are from the three people who left angry in 2020, that's what AI is going to repeat.

5. Get mentioned in places that aren't yours. A local news story about your team, an interview with one of your planners, a talk at an industry event, a podcast. It's the same stuff that helps your employer brand anyway, and it gives AI something outside your own site to point to.

6. Run the test again every quarter. Same prompts, same settings, same template. Things change, and you want to see if the fixes are working. Keep the old audits so you can compare.

And here's the checklist version so you can just work through it.

CAREERS PAGE
[ ] Salary ranges for the main roles, in plain text
[ ] Hours, shifts and location(s) are listed and current
[ ] Team size and what the team actually does
[ ] The hiring process, step by step, with a rough timeline
[ ] At least one real person with a name and a quote
[ ] No important info hidden in images, PDFs or videos only

JOB ADS
[ ] Job titles people actually search for (no internal titles)
[ ] Salary, location and hours in every ad
[ ] JobPosting structured data on every vacancy page
[ ] Expired ads taken offline, including copies on job boards and aggregators

WRONG OR OLD INFO
[ ] Every "wrong" or "outdated" row in the audit has an owner and a fix date
[ ] LinkedIn company page: description, location, size and website up to date
[ ] Old company names, addresses and benefits removed where you control them

REVIEWS
[ ] Current employees asked to write an honest review (never fake ones)
[ ] Critical reviews get a calm, human reply

OUTSIDE YOUR OWN SITE
[ ] One story, interview or mention outside your own channels this quarter
[ ] Your people are findable: a few employees talk about their work in public

TRACKING
[ ] Audit template filled in and saved with the date
[ ] Next test planned (same prompts, same settings) in 3 months

Look, you don't need to become an expert in how AI models work for this. You just need to look at what they say about you, check where it comes from, and make sure there's something true and specific out there for them to find. Most companies haven't even done step one yet.

Frequently asked questions

Do candidates really use ChatGPT to research employers?

More and more, yes. In a May 2026 survey of AI-using job seekers by PerceptionX, 96% had used AI to research employers or roles. That's a small, vendor-run study of people who already use AI, so don't read it as "everyone." But try it yourself with someone under 30 on your team, ask them how they'd check out a new employer, and I'm pretty sure the answer won't surprise you.

Why does ChatGPT give a different answer every time?

These tools don't give one fixed answer. They generate a new one every time, and sometimes they search the web and sometimes they don't. That's why you run each prompt twice and write down the date. You're looking for patterns, not one perfect answer.

Can I pay to show up better in AI answers?

Not in the way you can buy a job ad, at least not that I know of. What you can do is make sure there's good, current, specific info about you in the places these tools read. That's the whole point of the fixes.

How often should I run the AI employer audit?

Every three months is enough for most companies. Do it more often right after a big change, like a new name, a move or a big hiring push.

Sources

• PerceptionX, "How Candidates Research Employers With AI: 2026 Report" (May 2026): survey of 306 AI-using job seekers in 7 countries, 96% used AI to research employers or roles, 72% before deciding to apply, 40% to discover new employers, and the types of questions candidates ask. Vendor study, small sample.
• Greenhouse, job seeker survey press release (October 20, 2026): 77% of U.S. job seekers used AI in their job search in the past year.
• NBER Working Paper 34255, "How People Use ChatGPT" (September 2025): "Seeking Information" is one of the three most common ChatGPT use cases.
• Search Engine Journal, report on the OpenAI and Harvard ChatGPT usage study (September 15, 2025): seeking information went from about 14% to 24% of ChatGPT conversations.
• Perplexity Help Center, "How does Perplexity work?": Perplexity searches the web and shows numbered citations to its sources.