A useful website can still struggle to reach buyers through Google. The experience of aitrainer.work, a career service for AI training and evaluation roles, shows why marketers should separate technical access, content differentiation, and channel performance before judging their organic strategy. Its founder’s account also raises a broader question: which investments build lasting demand when publishing more pages offers no guarantee of visibility?
Separate Audience Demand From Search Engine Access
The founder reports a striking acquisition split for October 1–23: Bing delivered 2,387 visitors, ChatGPT delivered 1,103, and Google delivered 253. Those figures describe one site over a limited period, not a benchmark for other businesses. They nevertheless show why a Google-only report would give an incomplete picture of its reach.
A separate reporting window reveals another distinction. During the 30-day period ending September 8, the founder recorded 63 Google clicks, with 48 coming from searches for the site’s name. Bing generated approximately 320 clicks from job-related searches.
Branded discovery and category discovery serve different roles. Someone searching for a business by name already has some awareness of it. A category query gives that business a chance to reach an unfamiliar audience. Marketing teams should measure both, rather than treating all organic visits as interchangeable.
| Reported signal | Strategic reading |
|---|---|
| Most Google clicks in the earlier window were branded | Separate existing awareness from new audience acquisition |
| Bing supplied substantially more job-query clicks | Assess channel value against the audience it reaches |
| ChatGPT referrals exceeded Google referrals | Keep assistant traffic visible in acquisition reporting |
The founder suggests Bing and ChatGPT traffic may be connected because ChatGPT’s web search has used Bing’s index. That remains an explanation offered for the pattern, rather than proof that either channel will rise or fall with the other.
For B2B marketers, the practical implication is to keep acquisition reporting broad enough to show where discovery actually happens. A weak result from one platform should trigger investigation, not an immediate verdict on the entire content program.
Diagnose Delivery Before Commissioning More Content
The first obstacle at aitrainer.work was technical. Its founder says category pages processed the entire collection of jobs in memory for each request. On Cloudflare Workers, uncached requests could exceed the available CPU allowance and return a 503 response.
Caching concealed much of the problem from ordinary visitors. Requests for less recently accessed pages exposed it.
In June, the founder tested ten uncached category pages using Googlebot’s user agent. Every test returned a 503. Database-level filtering and prerendering resolved the errors, but indexing did not recover.
That distinction should shape resource decisions. More guides, social promotion, or paid distribution cannot repair a crawler’s inability to retrieve a page. Equally, successful delivery does not establish that a search engine will include it in its index.
Checks before expanding the publishing budget
- Test uncached pages as well as popular, cached destinations.
- Inspect responses using the relevant crawler’s user agent.
- Review index exclusions separately from server errors.
- Compare generated descriptions and structured data for duplication.
These checks create a better starting point for content planning. They help avoid spending editorial resources on a problem rooted elsewhere in the website.
Manage Page Inventory Around Business Value
The next issue involved repetition. The founder reports that Search Console showed roughly 3,050 crawled but unindexed pages in August. Across 1,765 listings, FAQ answers embedded in structured data were identical.
The founder attributed the exclusion problem to that sameness and changed the templates to draw answers and descriptions from individual listings. Afterward, 98.6% of job pages contained distinct answers. That is a reported implementation result, not evidence that Google accepted those pages.
By late September, exclusions also affected reviews, guides, and geographic hubs. At that stage, the founder interpreted the distribution as a possible site-wide judgment, with a large volume of temporary listings affecting more substantial material. The account does not establish Google’s reasoning.
The response was selective: Google indexing was restricted to live listings within a group of approximately 230 pages that had previously earned clicks. Bing retained a separate sitemap containing all job pages.
This is a useful planning question for any inventory-heavy site: which pages deserve ongoing acquisition investment? A page can serve existing users without being a priority search landing page. Conversely, a detailed comparison or guide may deserve sustained maintenance because it supports evaluation as well as initial discovery.
For a hypothetical B2B directory, that could mean retaining temporary entries for navigation while concentrating editorial work on durable category resources. Paid campaigns and social distribution could then direct audiences toward the same carefully maintained destinations. That is a strategic option, not an outcome demonstrated by this case.
Use Service Evidence to Support Brand and Demand
The founder describes competitors producing large volumes of similar material. However, rapid publishing and overlapping page formats alone do not establish a spam-policy violation. Nor does this account prove why competing sites ranked above aitrainer.work.
A stronger direction emerges from the service’s own operations. The founder reports approximately 400 hires resulting from referrals and says the site publishes hiring outcomes alongside pay information. Its dataset is available under CC BY 4.0.
The account also describes The Next Web citing pay figures from an Ethos review when covering the company’s Series A. The review drew on 87 tracked listings. That example connects editorial depth with outside recognition more directly than page volume does.
Put Original Evidence to Work
A useful workflow would begin with verified service data, then build a maintained resource around it. Search-focused pages could answer specific evaluation questions, while social posts could explain selected findings and point readers toward the underlying methodology. Sales or demand-generation teams could use that same evidence where it helps prospects assess the service.
AI assistance can support production, but the editorial priority should remain the evidence and its limits.
The business conclusion is practical: fix retrieval problems first, make deliberate decisions about indexable inventory, and assess discovery across channels. Then invest in material grounded in what the business actually knows. For the next reporting cycle, track whether those resources attract relevant visitors and support meaningful outcomes, alongside their rankings.
