Broken Google
Broken Google is the story of a search engine whose internal systems—indexing, canonicalization, and crawl control—have drifted so far from predictable behavior that even sophisticated implementers can no longer reliably influence what gets indexed, what ranks, or what disappears.
1. When URLs won’t die
One of the clearest symptoms of Broken Google is the persistence of URLs that should be long gone. Site owners regularly report removed pages—returning clean 404 or 410, removed from sitemaps and internal links—still being crawled and hanging around in Google’s index or Page Indexing reports for months or even years. Even the official position concedes that 404s can remain visible for “a while,” but in practice the tail can be so long that it breaks any intuitive model of an index that reflects the live web.
The Removals tool in Search Console doesn’t really solve this; it only hides a URL for around six months, after which Google may quietly allow the listing to reappear if its systems still consider the URL “valid” or keep discovering signals to it. Meanwhile, Googlebot often continues to recrawl these supposedly dead URLs, consuming crawl budget and server resources for assets the host has explicitly killed.
2. Redirects and canonicals that don’t consolidate
In a sane system, clean 301 redirects and consistent rel=canonical signals should consolidate signals to a single URL and cause the old version to fall out of the index. In Broken Google, that often doesn’t happen. Many SEOs see old URLs remain indexed alongside their replacements even after long-standing 301 redirects and proper canonical tags, with Search Console insisting those old URLs are “Indexed” and sometimes even treating them as 200 instead of 301 when inspected.
More disturbingly, there are documented cases where a redirected page is still chosen as canonical by Google’s own logic, even though Google fully recognizes the redirect. The result can be canonical ghosts: Page A redirects to Page B, yet Google declares Page A the canonical, reports Page B as indexed, but neither URL actually surfaces in search. Official documentation frames rel=canonical as a “hint” that might be ignored, but that framing masks the reality that canonical and redirect signals can now conflict in ways no human can reliably predict.
3. Canonicals that can be poisoned or ignored
Canonicalization is one of the most visibly broken layers. Google’s own docs admit that even explicit canonicals may be ignored and that its systems can select a different canonical if they think it is “better.” In practice, this often manifests as “Duplicate, Google chose different canonical than user” in GSC, even on unique, self-canonical pages with no obvious duplicates, leaving site owners chasing phantom conflicts.
Because canonical is only a hint, it is also surprisingly easy to poison. During hacks, attackers frequently insert cross-domain rel=canonical tags or malicious redirects pointing to spam domains; Google can then decide that the spam URL is canonical and effectively demote or drop the original site’s legitimate content. Multiple case reports show Google selecting unrelated or scraper sites as canonicals, including proxies and spammy copies, even when the original page uses correct self-referencing canonicals and appears to be the clear source.
4. Stuck hacked URLs and negative history
Hacked URLs expose another broken edge. In large hacks—Japanese keyword hacks, cloaked pharma pages, and auto-generated spam trees—thousands of junk URLs can get indexed rapidly. Even after the site is fully cleaned and those URLs now return 404/410, they often linger in Google for an extended period, sometimes continuing to show in site: queries and index reports. The recommended “don’t 301 hacked URLs because of their spam backlinks” advice leaves webmasters stuck between preserving their index and inheriting toxic link equity.
Google’s recrawl cadence is the bottleneck here: each spam URL must eventually be revisited and verified as gone, but when a hack produces tens or hundreds of thousands of pages, that cleanup can take far longer than any reasonable owner or user would expect. The outcome is a search engine that remains haunted by ghosts of a compromised past.
5. Robots.txt that controls crawl, not index
The robots.txt layer is another place where expectations and reality diverge. Google is explicit that robots.txt is a crawl control mechanism, not an indexation control mechanism, which is why “Indexed, though blocked by robots.txt” appears as a standard status in Search Console. In practice this means URLs can be blocked from crawling yet still be indexed and shown in results based on external links and other signals—often without a snippet or with minimal information—directly contradicting what many site owners think “Disallow” should do.
On top of that, real-world cases reveal Google’s robots parser and caching behavior to be fragile. Using fully qualified URLs instead of paths, duplicating user-agent blocks, or relying on non-standard directives leads to situations where Googlebot appears to ignore clear Disallow rules because the file doesn’t match its expectations. Google also disregards directives like Crawl-delay, and caches robots.txt in ways that can make fixes feel delayed or randomly applied, fueling the perception that it is not respecting instructions, even when the explanation is a mix of subtle syntax and caching rules.
6. Indexing vs. diagnostic reality
Perhaps the most confusing breakage is the growing gap between what is actually rankable in Google and what Google’s own tools say. SEOs increasingly report URLs that clearly rank for keywords and send organic traffic but do not appear via site: queries, quoted searches, or even direct URL search. This means Google’s core diagnostic workflows—used for decades to verify indexation—no longer faithfully reflect the underlying index.
Search Console adds another layer of contradiction. A URL may appear as “Indexed” with a perfectly acceptable canonical and no errors in the Page Indexing report, yet cannot be surfaced by any realistic search, making “indexed” almost meaningless in practical terms. Conversely, URLs flagged as “Crawled – currently not indexed” or “Alternate page with proper canonical” might quietly start ranking, despite supposedly not being part of the index. The upshot: you can no longer trust Google’s own interfaces to tell you what is really happening.
7. Conflicting and self-defeating signals
Broken behavior also emerges where Google’s systems encounter conflicting signals that webmasters would reasonably expect to be resolved deterministically. Common examples include chains where a noindex page canonicalizes to an indexable page (or vice versa), or where meta robots, x-robots headers, canonicals, and redirects disagree. Instead of a clear hierarchy, Google’s internal heuristics may produce outcomes where neither version is indexed as intended, or where the “wrong” signal wins.
The “Duplicate, Google chose different canonical than user” family of statuses is emblematic: Google will ignore a declared canonical if the content isn’t similar enough or if it sees stronger signals elsewhere, like sitemaps or internal linking, even when those are outdated or accidentally misconfigured. As a result, minor implementation mistakes become catastrophic, because Google’s override logic is aggressive and opaque.
8. Overloaded and misleading GSC error buckets
Search Console’s Page Indexing report tries to compress complex internal decisions into a small set of human-readable labels: “Crawled – currently not indexed,” “Duplicate, Google chose different canonical,” “Indexed, though blocked by robots.txt,” “Page indexed without content,” and so on. But those buckets are so generic that they often misdiagnose the actual cause of a problem, and two very different internal states can produce the same label.
For example, “Indexed, though blocked by robots.txt” can mean anything from “Google indexed a page from external links but can’t recrawl it to update” to “your robots rules accidentally hide your primary content.” Similarly, mass 404 or Soft 404 flags can hit perfectly valid, content-rich URLs after layout or template changes, suggesting that subtle shifts in how Google evaluates “quality” or “main content” can flip entire sections from normal pages to pseudo-errors overnight.
9. Fragment and alternate URL weirdness
Even at the URL level, Broken Google shows up in edge cases like fragments. Officially, URL fragments (the part after #) are not supposed to affect canonicalization or indexing, yet SEOs report GSC and ranking tools showing fragment URLs—page#section—as if they are distinct entries and sometimes even attributing better positions to them than to their fragment-less parent. This contradicts the mental model Google has taught the industry and makes it harder to reason about what Google considers the “real” URL.
Similar confusion surrounds alternate pages with canonicals and parameterized URLs. Reports of “Alternate page with proper canonical” or “Duplicate without user-selected canonical” often surface on pages that are meant to stand alone, forcing site owners to guess whether Google is treating them as second-class copies of something else or simply mislabelling them.
10. A search engine that no longer feels controllable
Individually, many of these issues can be rationalized: robots.txt is for crawl, not index; canonical is a hint; 404s take time to drop; reporting is approximate. But taken together, they describe a Broken Google: a system where:
- Legitimate, robust content often fails to index at all or is sidelined by phantom duplicates and mis-chosen canonicals.
- Removed URLs, hacked junk, and redirected pages keep showing up long after they should be gone.
- robots.txt, canonicals, and redirects—once reliable levers—have become probabilistic suggestions with undocumented edge cases.
- Search Console and
site:searches no longer give a trustworthy picture of what’s actually in the index or eligible to rank.
For site owners and SEOs, the practical consequence is loss of control. You can implement best practices, clean technical setups, and clear intent, and still have Google ignore, misclassify, or partially index your content for reasons that are neither observable nor fixable within the tools provided. That’s what people mean when they say Google is broken: not that it fails occasionally, but that its failures are now systemic, opaque, and increasingly indifferent to the webmasters trying to work with it.