Should you really hide navigation links on an e-commerce site?

Why does Google caution against hiding links?

Mueller's statement addresses a recurring issue in large catalog e-commerce sites. Many developers try to limit the exposure of certain pages by hiding links via CSS (display: none), loading them only with JavaScript, or using obfuscation techniques.

The problem? Googlebot interprets these practices as an attempt at manipulation. If a link is invisible to the user but present in the code, Google sees this as cloaking. If the link is only accessible through complex JavaScript, the crawler may never discover it—or may do so with significant delays.

What concrete alternatives does Google propose?

Mueller suggests two complementary approaches: rel="nofollow" and canonical tags. The nofollow tells Google not to pass PageRank through this link while still allowing for its discovery. The canonical tag consolidates SEO juice towards a preferred version when multiple similar URLs exist.

These two techniques allow for keeping links visible to the crawler while controlling the distribution of crawl budget and PageRank. This is particularly useful for navigation facets, sorting filters, or deep pagination pages that shouldn't be indexed but must remain crawlable.

In what contexts does this recommendation apply?

This guideline primarily pertains to sites with thousands of possible combinations of filters, sorts, and facets. Think of fashion sites with size × color × material × price, or marketplaces with hundreds of thousands of listings.

The risk is wasting crawl budget on low-value pages. Google could spend days exploring unnecessary variations while your strategic pages remain under-crawled. This is why it's essential to actively guide the bot with clear signals rather than hiding everything.

• Accessibility first: links must be discoverable by Googlebot, even if they don't pass PageRank
• Avoid exclusive JavaScript: don't rely solely on JS to load your deep navigation links
• Use rel="nofollow" strategically: on facets, filters, and sorts that create infinite combinations
• Canonicalize variations: when multiple URLs display the same content with different parameters
• CSS hiding is risky: display: none on links can be interpreted as cloaking if misused

Is this approach really the most effective?

Let's be honest: Mueller's recommendation is cautious but sometimes insufficient for large sites. The rel="nofollow" has lost much of its controlling power since Google turned it into a "hint" rather than a strict directive. [To be verified]: no public data confirms how much Google ignores this nofollow on internal links.

In practical terms, it’s observed that Google does crawl some nofollow links, especially if they are prominent in the site’s architecture. The canonical, however, works better — but only if the content of the pages is truly similar. In cases of very different facets, Google might reject the canonical.

What are the limits of this recommendation?

Mueller does not mention robots.txt, meta robots noindex, nor URL parameters in Search Console — three crucial levers for managing catalog crawling. The robots.txt completely blocks access, noindex allows crawling but prevents indexing, and URL parameters allow signaling to Google how to treat each type of parameter.

A second limitation: the statement does not distinguish between crawl budget and PageRank. A nofollow limits (in theory) the PageRank, but not necessarily the crawl. If you have 500,000 facets set to nofollow, Googlebot can still spend weeks crawling through everything. The real question is: how to structure the architecture so that Google doesn’t even discover these useless URLs?

What to do when JavaScript links are unavoidable?

Some frameworks (React, Vue, Angular in SPA mode) generate all content client-side. In these cases, Server-Side Rendering (SSR) or static generation becomes essential. Google does crawl JavaScript, sure — but with a delay and a significant crawl budget cost.

The pragmatic alternative: a hybrid navigation system with classic HTML links for critical paths and JavaScript for secondary interactions. Not glamorous, but effective. And if you truly cannot avoid pure JS, ensure that key URLs are at least present in a crawlable XML sitemap without executing JS.

How to audit your current link architecture?

First step: crawl your site like Googlebot with Screaming Frog, Oncrawl, or Sitebulb in "Googlebot smartphone" mode. Identify links hidden with CSS (display: none, visibility: hidden on parents) and links that are loaded only via JavaScript after user interaction.

Next, compare the number of pages crawled by your tool versus the number of pages indexed in Google (site: query). A significant discrepancy suggests that Google is ignoring part of your content — or conversely, indexing facets you thought were blocked. Also examine server logs to see which URLs Googlebot is actually visiting.

What critical mistakes must be avoided at all costs?

Critical mistake number one: placing important links solely behind event-driven JavaScript. If a user must click a button to dynamically load a dropdown menu, Googlebot will never see those links. Even in 2025, JS crawling remains costly and incomplete.

Critical mistake number two: using display: none on entire navigation sections to hide them on mobile. Google detects this and may consider that you're intentionally hiding content. Opt for CSS that visually reduces importance without hiding (size reduction, collapse with aria-expanded, accessible off-canvas).

What strategy to adopt for a catalog of thousands of products?

The most robust solution: a strict silo architecture with limited horizontal navigation. Each main category has its subcategories and products, but facets (color, size, price) generate canonicalized URLs pointing to the base category page.

Use robots.txt to completely block sorting parameters (ex: ?sort=price), meta robots noindex on deep pagination pages (page 10+), and rel="nofollow" on combined filter links. Finally, submit a clean XML sitemap containing only the URLs you really want indexed — not the 3 million possible variants.

• Crawl the site with a tool simulating Googlebot and compare it with actual indexing
• Identify all links hidden with CSS or loaded only with JavaScript
• Replace hiding techniques with rel="nofollow" and canonical where appropriate
• Block non-essential parameters via robots.txt or Search Console
• Noindex deep pagination pages and combined facets
• Create an XML sitemap containing only priority URLs