Can Google and AI find agency jobs? What we found on 1,195 job pages
We tested one live job page on each of 1,195 agency job boards in October 2026, and two gaps stood out. 45% had no JobPosting data, so those jobs cannot appear in Google's job search. 17% of the 1,094 we could read had no job text in the page as it arrives, so AI crawlers that skip scripts find no job. The findings come from The Staffing Career Site Report 2027.
Daniel BartakovicsCo-Founder & CDO10 min read
Ask AI

45%
Job pages with no JobPosting data
In the report: 45% of 1,195 agency job pages we tested had no JobPosting structured data, which Google uses to list jobs in its job search.
The Staffing Career Site Report 2027, page 6
The short version
45% of 1,195 agency job pages we tested in October 2026 had no JobPosting data, so those jobs cannot appear in Google's job search.
17% of the 1,094 job pages we could check had no job text in the HTML the server sends, so AI crawlers that skip scripts find none.
Few boards block crawlers in robots.txt: it kept Googlebot out of 3% of 1,585 agency job boards and the AI search crawlers out of 4%.
The setup decides most of the result: JobPosting data on 80% of staffing website vendor boards, against 23% on ATS-hosted boards.
How does a job get into Google and AI search?
A Bullhorn job reaches a candidate through Google or an AI assistant only if a crawler can read the job page first. Search engines and AI companies each send crawlers, programs that read web pages. Letting one read the job comes before any recruitment SEO work on keywords. The report sets out the rules each route follows, from the documentation of the companies that run them:
- Google for Jobs is the job search box Google puts above its normal results, and it fills from JobPosting data. It is a hidden label on the job page, written for machines, that gives Google the title, pay, location, employer and closing date. Required fields include title, description, date posted, employer and location; salary, expiry date and a direct-apply flag are recommended (Google Search Central).
- Every job needs its own address. Google generally ignores the part of an address that follows a
#(Google Search Central). A job that only exists atagency.com/#/jobs/123is unlikely to be indexed as its own page. - Google runs JavaScript, the scripts that build a page in the browser, with delays (Vercel and MERJ, 2024). In a separate December 2024 study by Vercel and MERJ, the crawlers used by OpenAI, Anthropic and Perplexity did not run them. ChatGPT search also needs OpenAI's search crawler, OAI-SearchBot, allowed in robots.txt, the short file that tells crawlers which parts of a site they may read (OpenAI).
- Google's AI features draw on its normal index. Google says no special markup or AI text file is needed (Google Search Central).
Every route starts with a page a machine can read: its own address, the job text in the page, and robots.txt that lets the crawler in. The five-layer agent-ready framework builds on that base.
What did we find on 1,195 job pages?
Most agency job pages pass the basic checks; the gaps sit in the JobPosting data and in pages that are built in the browser. Figure 6 of the report shows the share of job pages that met each check, and the biggest gaps also made the report's ten headline numbers.
Share of agency job pages meeting each check
What most pages get right
Risks
| Check | Share of job pages |
|---|---|
| Job readable in the page HTML, no JavaScript needed | 83% of 1,094 pages |
| Sitemap found | 79% of 1,087 pages |
| Canonical tag present (a line naming the job's main address) | 70% |
| JobPosting structured data | 55% |
| Hash-based job URL, with a # in the address (a risk) | 11% |
| robots.txt blocks an AI search crawler (a risk) | 4% of 1,585 boards |
| robots.txt blocks Googlebot (a risk) | 3% of 1,585 boards |
| Job page set to noindex, a tag that keeps it out of search (a risk) | 3% |
Readable in the page HTML means the job text is already in the HTML the server sends. It is in the page when it arrives, not added afterwards by a script. A sitemap, the list of a site's pages that helps search engines find new jobs faster, was found for 79% of the 1,087 pages we could check.
The readable and sitemap checks use fewer pages because bot protection stopped the check on some sites. Those pages count as unknown, not as failures, so the shares are not padded with guesses.
17% of the 1,094 pages we could check had no job text in the HTML the server sends. That side of the readable check is the one AI search depends on. For Google that means waiting until it renders the page.
For an AI crawler that does not run scripts, the job is not there. And of the pages with a canonical tag, 86% pointed to the job itself. The rest pointed somewhere else, such as the job list or the homepage, which asks Google to treat the job page as a copy of another page.
Why can 45% of job pages not appear in Google's job search?
45% of the 1,195 job pages we tested had no JobPosting data, and Google for Jobs lists jobs only from that data. Those jobs cannot show up in the job search box that candidates use when they type "forklift jobs near me".
For an agency, this is the gap that hides best. A Bullhorn job can be live on the site, look complete to a candidate on a phone at lunch, and still never reach Google's job search. Nothing on the page looks broken.
You only see it when you paste the address into Google's Rich Results Test, its free page that shows whether Google can read the label, or read the page source.
To get the label onto every job, follow how to get jobs on Google for Jobs, and hand the JobPosting tutorial to whoever builds the job template.
Is the JobPosting data that exists complete?
Often not: where the data was present, salary appeared in 50% of pages and an expiry date in 72%. Both fields are recommended rather than required, so the job can still be listed. It is listed without pay or a closing date.
63 pages (5%) carried JobPosting code that is not valid JSON, the format the label is written in, and Google ignores code it cannot read. For Google, those pages have no data either.
And 17 live jobs had an expiry date already in the past, so the label told Google that a job still live on the board had closed.
Does the ATS decide, or the setup?
The setup decides more than the ATS: the share of job pages with JobPosting data ran from 80% on one type of board to 23% on another. Table 5 of the report splits the same checks by type of job board, for the four types with 100+ tested boards.
Search visibility by job board type · share of tested boards of each type
| Job board type | JobPosting data | Readable without JavaScript | Hash URLs | Canonical tag | Sitemap |
|---|---|---|---|---|---|
| Staffing website vendor281 boards | 80% | 98% | 0% | 97% | 98% |
| WordPress job plugin220 boards | 70% | 98% | 4% | 96% | 95% |
| Custom build339 boards | 44% | 89% | 1% | 72% | 80% |
| ATS-hosted job board267 boards | 23% | 47% | 44% | 21% | 53% |
Higher is better in every column except hash URLs, where higher is worse.
| Job board type | Boards | JobPosting data | Readable without JavaScript | Hash URLs | Canonical | Sitemap |
|---|---|---|---|---|---|---|
| Staffing website vendor | 281 | 80% | 98% | 0% | 97% | 98% |
| WordPress job plugin | 220 | 70% | 98% | 4% | 96% | 95% |
| Custom build | 339 | 44% | 89% | 1% | 72% | 80% |
| ATS-hosted job board | 267 | 23% | 47% | 44% | 21% | 53% |
A staffing website vendor is a firm that builds agency websites, job board included in most cases. A WordPress job plugin adds the ATS jobs to a WordPress site. Boards of both types almost always had readable job pages with structured data.
An ATS-hosted board is a job board run inside the agency's applicant tracking system, shown on its own address or in a frame. These boards more often render jobs in the browser, and 44% of their pages used hash URLs.
The report is clear on where that comes from. "These are setup choices, not limits of the ATS: the same ATS can feed a board that passes every check." For an agency on an ATS-hosted board, the first place to look is how the board is set up.
Custom builds, boards a web agency or developer built for one agency alone, sit in the middle. Their pages were mostly readable (89%), but only 44% carried JobPosting data, so the job reads well to a person and a crawler but often lacks the label Google for Jobs needs.
UK boards did better than US boards on readable job pages (96% vs 80%) and on hash URLs (2% vs 14%). A US agency has more reason to run the view-source check below first.

Which blocks do agencies not know about?
robots.txt kept Googlebot out of 3% of the 1,585 job boards we checked, and where crawlers were blocked, the rule often came from the platform, not the agency. 3% of job pages were set to noindex, a tag that tells Google not to show the page.
Crawlers kept out by robots.txt · share of 1,585 job boards
The AI search crawlers, which the assistants use to find pages they can quote (OpenAI's OAI-SearchBot, PerplexityBot and Claude-SearchBot), were blocked on 4%. OpenAI's training crawler, GPTBot, was blocked on 6%. Those are two different decisions: one is about model training, the other about whether ChatGPT search can find your jobs.
Three patterns in the robots.txt files we archived come from a platform, not from rules written for the agency's site. One ATS's shared job domain disallows all crawlers, Google included. A CRM-based site platform lets Googlebot in on its shared domains and keeps every other crawler out.
And several agency sites block AI crawlers through a managed robots.txt setting in Cloudflare, a security and delivery service, rather than through rules written for the site.
In each case the agency may not know the block is there. The robots.txt tutorial for Google and AI crawlers shows how to read the file and what to change.
The hardest block to see is bot protection, a security layer that turns away visitors that do not behave like a normal browser. 8% of job pages refused requests that did not come from a full browser. A candidate never notices, but a crawler can get nothing.
Are agencies hiding text aimed at AI?
Some are: 16 of the 356 job pages we checked by hand (4%) contained text hidden from human visitors that tells AI tools to add a specific phrase to the application. We do not reproduce the wording. It appears to be a way to spot applications written by AI agents.
If your board does this, make it a decision you took on purpose. Candidates cannot see the hidden text.
How we tested
I ran these tests from 26 to 28 October 2026. A script opened each job board in Chrome on a desktop with a US connection, found one live job and tested that page. It read the JobPosting data, the HTML the server sends without JavaScript, the URL format and the canonical tag. It also read the meta robots and X-Robots-Tag settings, which can tell Google not to show a page, the robots.txt rules per crawler, and the sitemap. It never clicked Apply, submitted a form or logged in. robots.txt files, sitemap summaries and the HTML of each tested page are archived.
We started from the 1,819 job boards identified in our review of more than 2,000 staffing firm websites. 624 boards had no page we could test: no job could be reached automatically, the page did not load, the board blocked the browser, or the jobs had no address of their own. That left 1,195 tested job pages. robots.txt was read on 1,585 boards.
To check the script, we also tested 307 boards by hand. The two agreed on 100% of JobPosting and hash URL results, 98% of robots.txt results, 99.7% of noindex results and 95% of raw HTML results.
How do I check my own job board?
Three checks cover the biggest gaps in the data, and none needs more than a browser.
The full set is in the 15-check job board checklist. To see which of your jobs Google has kept, follow which of your jobs Google has indexed, which uses Search Console set up for a subdomain. Short answers on hash URLs, noindex and canonical tags are in the findability FAQ.
If you would rather have someone else look, the free audit below runs the same tests we used for this report on your board and sends you a short report.
Our interest in these results is direct: we build a job board for staffing agencies, which agencies connect to Bullhorn and run on their own domain or subdomain.
Interested in this? Go there
Be found
7 reasons agency jobs miss Google and AI search, most common first
Why jobs go missing from Google and AI search, most common first
Be found
Audit your own job board in ten minutes: seven steps, with screenshots
Check your own board yourself in ten minutes, click by click
Be found
Googlebot vs the AI search crawlers on agency job pages
Why Googlebot and the AI crawlers see different pages


