The starting point: a jobs board AI crawlers couldn't read

GoTeachingJobs is a global marketplace for teaching jobs, and almost the whole site is job listings. Those listings were built in the browser. That worked for a person with a modern browser, but many AI crawlers read only the HTML the server sends, and on GoTeachingJobs that HTML held a loading message where the jobs should be.

What the AI Discoverability audit found

WriteWorks audited a sample of goteachingjobs.com drawn from every section of the site. The overall readiness score looked healthy, but that hid the problem, because the site is almost entirely job pages.

  • The listings needed JavaScript. Only 5% of the audited job pages kept 30% or more of their main content in the raw HTML. On the main jobs page, crawlers got "Loading jobs..." instead of the listings.
  • No H1 until JavaScript ran. Only 5% of job pages had an H1 in their raw HTML, which is why only 45% of audited pages passed the answer-first check.
  • Listings weren't described as structured data. None of the listing pages exposed its jobs as an ItemList, even after rendering. (Job detail pages already carried JobPosting.)
  • Sitemap dates didn't reflect change. Only 11% of audited pages had a sitemap date of their own; the rest shared the time the sitemap was generated.
  • Access was fine. robots.txt allowed every AI bot checked, and no AI user agent was refused.

What changed after the fixes

GoTeachingJobs moved the listings, counts and H1 into the server HTML, added ItemList markup to listing pages and gave the sitemap per-listing dates. The re-audit used the same method.

  • Resolved: job pages AI crawlers can read. 100% of audited job pages now keep 30% or more of their content in the raw HTML, up from 5%. The main jobs page now sends all of its rendered content in the HTML.
  • Resolved: headings. Every audited page has an H1 in its HTML, and pages passing the answer-first check rose from 45% to 99%.
  • Improved: the rest of each listing. The median job page now carries 73% of its main content in the raw HTML, up from 11%, and job pages with 80% or more rose from 5% to 37%. The rest still load part of the listing with JavaScript, and they're next.
  • Improved: listing markup. Listing pages carrying ItemList in their raw HTML rose from 0% to 37%.
  • Improved: sitemap dates. Audited pages with a sitemap date of their own, rather than the most common one, rose from 11% to 53%.

How these numbers were measured

Before = the first WriteWorks AI Discoverability audit, after = the re-audit once the fixes were deployed, on the same pages with the same rules. Percentages are shares of the audited pages, not of every URL on the site. A fix is marked Resolved when the re-audit shows it fixed, Improved when it is partly fixed, and Open when nothing changed. Speed comparisons come from one vantage point and are indicative. AI platform coverage is the share of the AI platforms measured that cited the site, at the baseline and six months on: a timeline, not a claim about cause. A job page counts as readable by AI crawlers when 30% or more of its main content is in the raw HTML.