AI Overviews are quietly summarizing job listings. Here's what I learned testing 30 niche queries.

A query I run weekly — “remote logistics jobs Europe” — started returning a Google AI Overview at the top of the SERP a couple of months ago. Three roles summarized, two with salary ranges, one cited from a niche board, two from aggregators. Just a generative answer box with citations, sitting roughly where the classic Jobs panel sometimes still does.

That observation kicked off six weeks of testing. I ran 30+ queries across logistics, pharma, nursing, aviation, finance, construction, and security niches, in both EU and US locales, and compared whatever AI Overview text surfaced against the source pages’ JobPosting JSON-LD. The goal was figuring out which fields the AI Overview was actually pulling, which it was ignoring, and what a board operator could do today to be one of the cited sources.

Caveat up front, since I’ve seen other guides present similar observations as if they were Google-documented behavior. This is small-sample empirical observation, not Google spec. Google has not published which JobPosting fields its AI Overview ingestion pipeline weights, and there’s no special “AI Overview” schema tag. What I have is a pattern across 30+ queries — useful, but not authoritative. If your testing shows different behavior, trust your data over mine.

What’s actually documented

To separate signal from noise before the observations:

• AI Overviews launched in May 2024 (renamed from SGE) and expanded to a long list of EU countries in March 2025
• The feature now runs in 200+ countries and 40+ languages
• An average AI Overview cites around 8 sources, heavily biased toward pages in the top 10 organic results
• JobPosting schema markup helps with AI-search visibility broadly — Google has said so directly, and third-party guides (Cavuno, Hashmeta, others) report the same
• There is no special “AI Overview schema.” Content quality and relevance drive citations

What I’m contributing in the rest of this post is empirical color on which JobPosting fields actually surface in the answer text when an AI Overview cites a job listing — observed, not documented.

What triggers a job-listing AI Overview

Not every job query. The pattern across the 30+ I tested:

• Role + region/city + a modifier consistently triggered Overviews (~70% of test queries)
  • “Senior customs broker Munich”
  • “Remote logistics jobs Europe”
  • “ICU nurse Berlin weekend”
• Broad queries rarely triggered (~10%)
  • “Jobs near me”
  • “Best companies to work for”
• Brand-named queries almost never triggered (~5%)
  • “Maersk careers”
• Specific niche-vertical queries triggered most often (~80%)
  • “Cabin crew jobs UK”
  • “Veterinary nurse Liverpool”
  • “Aviation pilot regional”

The pattern: AI Overviews surface when Google thinks the user benefits from synthesis across multiple sources — which is most niche job-search intent.

What appeared in the Overview text

I compared the AI Overview’s summary text against the source page’s JSON-LD for ~80 cited listings across the 30 queries. Six fields consistently appeared in the Overview’s text:

Field	Appearance rate (n≈80)	What appears
title	100%	Verbatim, as the role name
hiringOrganization.name	~95%	Used as the employer label (“at Maersk”)
jobLocation.address.addressLocality + addressCountry	~90%	“Berlin, Germany” — city first
First 1–2 sentences of description	~85%	Used as the role summary
baseSalary.value + currency + unitText	~70% (when present)	“€55,000–€70,000 per year”
datePosted	~60% (filtering signal)	Older roles filtered, not displayed

That last one is filtering, not display — listings with datePosted more than ~30 days back were less likely to be cited at all in my testing.

What I didn’t see in the Overview text

Equally important — these fields, which the classic Jobs panel rendered or weighted, didn’t appear in any AI Overview text I captured:

• responsibilities — the long bullet list of duties
• qualifications.education and qualifications.experienceRequirements
• benefits
• applicantLocationRequirements
• directApply
• industry and occupationalCategory

This doesn’t mean Google’s pipeline doesn’t read these fields — it might use them for retrieval-side ranking. What it means is they’re not what gets paraphrased into the answer text. The Overview is synthesizing the role for the user, not detailing it. The full responsibility list is irrelevant to the displayed answer; the first two sentences of description are heavily weighted.

Documented vs observed — keeping the two separate

Documented by Google or well-established by third parties:

• AI Overviews surface for queries where Google judges synthesis adds value
• Citations bias toward top-10 organic results
• ~8 source citations per Overview on average
• No special “AI Overview schema”
• JobPosting schema helps with broader AI-search visibility

Observed in my testing (not Google-documented):

• The six-field extraction pattern above
• Specific city in addressLocality correlates with citation more than region/country
• baseSalary presence correlates with higher citation rate
• datePosted >30 days correlates with non-citation
• Specific role titles cited more than category-level ones (“Senior Customs Broker” vs “Logistics Position”)

Treat the observed list as a hypothesis worth testing on your own listings, not Google’s stated behavior.

Practical schema hygiene

If you want to be cited disproportionately as the AI Overview space matures, here’s what I’d do based on the patterns.

1. Make the first 150 characters of description summary-quality. Most listings start with “About Acme Corp: founded in 2003, we are a leading…” — that’s company boilerplate, and that’s what got paraphrased in my tests. Start with the role: “Lead European customs operations across 6 hubs, owning compliance for €40M of annual freight movement…” That kind of opening is what got pulled.

2. Use a specific city in addressLocality, not a region. “Berlin” got surfaced; “Germany” alone didn’t, in any of my tests. “Remote” got surfaced only when paired with applicantLocationRequirements for a defined region. If your CMS defaults addressLocality to country-level, fix it.

3. Set baseSalary correctly — including currency and unitText. Listings with baseSalary.value: 55000, currency: "EUR", unitText: "YEAR" were cited ~3× more often than the same listings without salary, in my sample. If you’re aggregating and salary isn’t in the source, inferring a range from comps for the role/city is better than leaving it blank. (Validate against any compliance requirements; many regions now require salary disclosure on listings anyway.)

4. Keep datePosted fresh. AI Overviews filtered out listings >30 days old in most of my tests. If your listings auto-expire after 60 days, you’re invisible to AI Overviews for the second half of their life. Either reduce expiry, or refresh datePosted weekly for genuinely-active roles. (I haven’t seen Google’s Search Console flag this as manipulation when the role remains genuinely open, but that’s not a guarantee.)

5. Use specific job titles, not categories. “Senior Customs Broker” got cited. “Logistics Position” didn’t. If your CMS auto-generates titles from category templates, swap it for real role titles.

A 30-minute audit on your top 10 listings

Worth doing once on your most-trafficked listings:

1. Open the page in a private window
2. View source, find the <script type="application/ld+json"> block containing the JobPosting
3. Confirm title, hiringOrganization.name, jobLocation.address.addressLocality, baseSalary, description, datePosted are all present and specific
4. Read the first 150 characters of description — does it describe the role or the company?
5. Google your role title with city (“Senior Customs Broker Berlin”) and check whether an AI Overview renders and whether your listing is cited

If you’re not cited, the gap is almost always in one of those six fields. Fix it on the listing template, the next 10,000 listings inherit, and your AI Overview surfacing baseline improves by default.

What I’d test next

The model behind AI Overviews changes regularly — Gemini was upgraded to 2.0 in March 2025 — so the field-extraction patterns may shift. The structural fact (AI Overviews surface for specific-intent job queries; some fields drive what’s paraphrased; small-sample observations match a coherent pattern) is more durable than the specific numbers.

Things I’d test next, and would welcome operators sharing data on:

• Whether jobLocation.geo (lat/lng) correlates with citation in geo-specific queries
• Whether structured-data validity (clean vs warnings in the Rich Results test) correlates with citation rate
• Whether boards with multi-language listings (inLanguage set) get cited more in region-specific searches
• Whether the source page’s overall organic rank (top-3 vs top-10) matters more than schema quality

If you run AI Overview tests on your own niches, the data is genuinely useful to other operators. The space is too new for any single source to have authoritative answers. The boards that test their own SERPs will outperform the ones who just read posts like this one.