A page can look simple while carrying an unnecessarily complicated underlying document. When planning a redesign, consider how the essential information is delivered as well as how it looks. Useful content should not become an afterthought buried behind a large amount of repeated interface material.
Reconstructed archive: March 2026. Written in November 2026 using contemporary historical sources. The archive date organizes this series; this text is not a recovered original.
The March 2026 crawler explanation
Google explained its crawling and fetching infrastructure on March 31, 2026. The post described per-URL fetch limits and the risk of oversized inline resources pushing important HTML beyond the retrieved portion. Separate resource limits should not be confused with a single allowance for every asset on a page.
Check whether HTML size is the relevant problem
Current technical review: November 11, 2026. Fetch limits can change; verify the linked documentation before using this checklist.
This check concerns the response delivered for one URL. It is different from the total weight of a page’s images, scripts and other downloads. A slow page is not automatically an oversized HTML document, and a visually simple page can still contain a large amount of repeated or embedded code.
At this review, Google’s current Googlebot documentation specifies a 2 MB fetch limit for supported non-PDF files, applied to uncompressed data, and a separate 64 MB limit for PDFs. The March 2026 explanation, corrected in October, clarifies that HTTP response headers count toward the fetch limit. Referenced resources have their own limits; they do not share one combined allowance with the parent HTML. Other crawlers can have different limits.
Ask for the response, not a page-speed score
Have the developer examine an affected public URL and record:
- The actual response: its final URL, status, content type and the conditions under which it was retrieved.
- The relevant size: the uncompressed HTML response size, with response headers accounted for. A compressed transfer number or the sum of all page downloads is not the same measurement.
- What occupies the document: unusually large embedded image data, inline scripts or styles, repeated menus, or duplicated template output.
- Where essential information appears: service facts, important links, metadata and relevant structured data in the response.
A small response does not establish that every later rendering step works, but it can show that this particular fetch-size concern is not the next problem to pursue. Do not reduce useful content simply because a report contains a large number.
Separate truncation from content that loads later
Google’s explanation says it stops fetching at the applicable limit and passes the retrieved portion onward. If important bytes fall beyond that point, they are not available from that fetch. Ask the developer to distinguish this from content absent from the initial HTML because it is supplied by a later script or data request.
For a smaller response whose content still fails to appear, investigate the actual dependency or rendered-content behavior. That is a different diagnosis. A successful browser view does not prove that the crawler received the same information.
Remove excess code while preserving the answer
Hypothetical example: a template repeats a large navigation block and embeds image data before the service description. The developer finds that this pushes important information beyond the relevant response boundary. The proposed repair is to remove duplicated output or change how the heavy assets are supplied, while preserving the service explanation and customer route. This is not a measured Guardian client result.
Ask the implementer to explain the chosen change and its dependencies. Moving code to a separate file changes how it is requested; that file still needs to be accessible, usable and within its own applicable limit. Moving key information earlier can reduce exposure to a cutoff, but it does not replace checking the complete response.
Keep a before-and-after retrieval record
- Repeat the same response measurement after the approved change.
- Confirm that the essential text, links and metadata remain available and that the public page still works.
- Check another page using the affected template, including one with unusually long content.
- Record the URL, date, method, relevant byte measurement, change and observed result. Keep any remaining rendering or indexing questions separate.
Use lab and field performance evidence when the actual issue is speed or responsiveness. Passing this retrieval check does not guarantee indexing, rankings or improved customer outcomes.
Explore website design, digital marketing strategy, or contact Guardian Solutions to discuss your next step.
Historical source: Google announcement dated 2026-03-31.