JavaScript rendering and content delivery to crawlers
The site delivers crawlers a prerendered document and visitors a client-rendered app. 3.1 moved from Gap to Pass after a fix, on the same tool and the same deciding field. Two things are carried openly rather than folded into the Pass: spinners elsewhere are reduced, not gone, and a few URLs outside the priority set still fail, in a variant our own check could not detect at first.
Re-verified after the fix. In two sweeps of the 25 priority pages, none served a 'Loading...' placeholder in the crawl-stage HTML. Pages that had shown only a placeholder now carry their full text. Spinners elsewhere on the pages are reduced, not gone, and a few URLs outside the 25 still fail, as recorded under 1.9 and 2.8. The sweeps read one cached snapshot, so they do not sample two independent prerenders.
Corrected from Gap to Pass after the development team's fix shipped. Before the fix, 7 of 25 priority pages failed. The verdict moved on the same tool and the same deciding field that produced the Gap.
JavaScript execution produces most of what Google sees: 46 to 152 links per page in the crawl HTML against 100 to 483 in the rendered DOM. Google's own render corroborates this on a sampled page (111 raw links, 275 rendered).
All links in the crawl HTML carry an href, with no '#!', javascript: or empty values. In the rendered DOM there are no href-less anchors, no onclick attributes and no non-anchor role=link elements. Limit: handlers bound with addEventListener are invisible to DOM inspection, so non-anchor navigation can be under-reported.
No '#!' fragment-scheme hrefs in crawl HTML or rendered DOM. A navigation probe on 3 pages showed pushState called, no document reload, the path changed, and no fragment in the URL. All 25 priority URLs return 200 standalone.
The site runs dynamic rendering: all 25 priority pages serve two different documents at one URL. Googlebot gets a full document with title, main content and links, while a browser gets about 4 KB with none of them, and everything a visitor sees is built in the browser. Google positions this as a workaround rather than a long-term approach. The Gap records only that the site does not follow that recommendation. Nothing is broken and Google indexes the site through it. It matters because every other rendering check depends on knowing which document was read.
Genuine misses are handled correctly: 5 of 5 probes return 404 with a noindex robots meta tag. But any arbitrary subpath under a valid page returns 200 with the parent's full content, 5 of 5, which creates an unbounded URL space. A correct rel=canonical to the parent mitigates it, so this was recorded rather than raised.
Rendered DOM, page resources and console messages were captured from Google's Rich Results Test for 3 pages. On one page Google reported 3 of 122 resources not loaded, plus 2 console errors, and a spinner in a hidden panel. That did not reproduce in our own render, so it was recorded, not raised.
No page gates content on scrolling. The one material case was re-checked and the words were already present at load: late async content, not scroll gating. About 98% of images use loading="lazy", and none carry data-src without a src. Stated separately: 12 of 25 pages gate content behind a 'load more' click. Googlebot does not click either, but that is a different claim (see 5.8).
Want this run on your site?
We test your site against the same claims, record a verdict and the evidence for each one, and tell you what to fix first.
Start a conversation