Three months after opening clearcontext.dev to search crawlers, Google Search Console recorded 453 domain impressions and a single click on the about page. Two technical posts accounted for 328 of those impressions, while the top visible queries (mostly literal deprecation and error notices) account for just 58 impressions domain-wide.
Counting entries in data.jsx dated on or before the 2026-09-28 snapshot window end, ClearContext had published 18 articles across four months. The Search Console snapshot pulled on 2026-10-01, archived at docs/news-research/sc-snapshot-2026-10-01.md, showed search traffic behaving under a completely different logic.
The numbers cover the 90-day window from 2026-06-29 through 2026-09-28. Here is what actually happened.
Two posts captured 328 impressions, but top queries explain only 58
Across the entire domain, Google reported 453 impressions, 1 click, a click-through rate of 0.2%, and an average search position of 33.5. Search Console's performance table listed 18 distinct URLs that received impressions: 13 articles and 5 site pages (/about/, /, /archive/, /privacy/, and /terms/). Five published articles received zero impressions in the window.
When summing per-page impressions across the export, the individual URLs total 465 impressions, compared to the 453 property-level total reported by Search Console (which deduplicates multiple appearances in a single query). Two articles accounted for 328 of those impressions (72.4% of domain property impressions, or 70.5% of the 465 per-page sum):
| Page | Clicks | Impressions | Average Position |
|---|---|---|---|
/posts/assistants-migration-prompt-gap/ |
0 | 166 | 16.3 |
/posts/mcp-install-trust/ |
0 | 162 | 60.1 |
/about/ |
1 | 5 | 31.8 |
| Remaining 15 pages in report | 0 | 132 | — |
The two articles, assistants-migration-prompt-gap (166 impressions) and mcp-install-trust (162 impressions), held 328 impressions combined. But the snapshot does not measure query-to-page attribution. Search Console reports aggregate page performance and top query performance separately, so no measurement links specific search queries to individual pages.
In the query export, Google listed 37 total queries for the domain. The top 10 queries together accounted for only 58 impressions (12.8% of the 453 domain total). Several visible terms in that top 10 are literal deprecation notices or configuration errors: prompt objects are being deprecated across three punctuation variants totaled 31 impressions (at positions 7.2, 7.6, and 7.5), and could not connect to mcp server mcp-registry accounted for 4 impressions (at position 44.8).
While literal error and deprecation strings appear in the top queries, they explain only a small share of domain visibility. The data does not support claiming that the two top posts were powered primarily by exact error messages. The vast majority of impressions came from queries outside the top 10 that Search Console did not break down.
The only click on the domain was on the about page
The performance table records a single click across the entire three-month window. That click was not on an article, a benchmark, or an API analysis. It landed on /about/, which had 5 impressions and an average position of 31.8.
Every article on the site received exactly zero clicks.
The explanation is in the rank distribution. Pages that achieved first-page average rankings recorded low impression counts: cross-model-reasoning-portability averaged position 6.1 across 22 impressions, qwen-3-8-two-licenses averaged position 7.8 across 5 impressions, and gemini-temperature-deprecated-but-live averaged position 8.3 across 29 impressions. Those counts were too small to produce clicks at current click-through rates.
Conversely, the two articles with higher impression counts sat deeper in search results. An average position of 16.3 lands in the middle of page two, while position 60.1 lands on page six. In search engine terms, page two rarely generates visits.
Google indexed 21 pages and left Recordy in crawl limbo
The Search Console page indexing report, updated 2026-09-20, listed 21 indexed pages and 9 non-indexed URLs:
| Indexing Status | Source | URLs | Details |
|---|---|---|---|
| Indexed | 21 | Known pages on clearcontext.dev | |
| Excluded by 'noindex' tag | Website | 4 | Excluded by 'noindex' tag |
| Page with redirect | Website | 3 | Page with redirect |
| Crawled – currently not indexed | Google systems | 1 | /rss.xml (last crawled 2026-07-16) |
| Discovered – currently not indexed | Google systems | 1 | /posts/recordy/ (first detected 2026-07-01, last crawl: N/A) |
Search Console divides the 9 excluded URLs into two source categories: Website (7 URLs) and Google systems (2 URLs). Under Website reasons, 4 URLs are excluded by a 'noindex' tag and 3 URLs are marked as page with redirect. Google systems excluded 1 URL as crawled but not indexed (/rss.xml, last crawled 2026-07-16) and 1 URL as discovered but not indexed (/posts/recordy/).
The case of recordy is notable. The article is 3,145 words, returns HTTP 200, links from site navigation, and appears in sitemap.xml. Yet Google detected it on 2026-07-01 and marked it "Discovered – currently not indexed" without fetching the page bytes. Its "Last crawl" timestamp remains N/A.
A priority crawl was requested via Search Console URL Inspection on 2026-09-18. We chose a two-week waiting window before evaluating whether Google crawled the page, leaving the outcome pending under issue #32. We made no attempt to rewrite /posts/recordy/ or edit its lastmod date. Google has never crawled the URL, so it holds no content evaluation that editing could alter.
Search Console shows where impressions landed, not which queries drove them
This dataset carries the Tested evidence label: all numbers come directly from the Search Console domain property export for clearcontext.dev on 2026-10-01, covering 2026-06-29 through 2026-09-28. The limits are substantial: this is a single domain with 18 published articles on or before the snapshot window end, 453 domain impressions, and 37 recorded queries. The export provides aggregate page totals and query totals without query-to-page attribution. Attributing the impressions of the top two posts to specific error queries is unsupported, especially when the top 10 queries account for only 58 impressions domain-wide.
What the data demonstrates is concentration: search demand focused on two specific technical posts rather than distributing across the catalog. The literal strings in the visible queries are a lead worth testing, not a finding. The next snapshot can check it by filtering the query report to each of the two pages.