URL: /rules/crawl/rendered-signal-mismatch --- title: "Rendered signal mismatch: JavaScript changes the canonical, noindex, title or H1" description: "A script that rewrites the canonical, adds or removes a robots noindex, or replaces the title or first H1 makes crawlers that read raw HTML see a different page from a browser." --- ## What it is Crawlers and agents that do not run JavaScript read the raw HTML the server sends. A browser shows the page after scripts have run. When a script rewrites the canonical URL, adds or removes a robots `noindex`, or replaces the title or first `

`, the two readers get different pages. The signals that matter most are the canonical and `noindex`. A canonical or `noindex` that differs between the raw and the rendered page is a common cause of pages dropping out of the index after a framework migration: the server sends one thing, the client script sends another, and which one a search engine acts on depends on whether it renders the page. [`ax/content-without-js`](/rules/ax/content-without-js) already compares the body content of the raw and the rendered page. This rule compares these four head signals. ## What squirrel checks It compares the raw HTML the crawl fetched with the DOM from the cloud render of the same page, and emits four checks: | Check | Fires when | Severity | |---|---|---| | `canonical-mismatch` | the canonical URL differs, or the script removes it or adds one that points elsewhere | error | | `noindex-mismatch` | `noindex` is present on only one side, in any robots or googlebot meta tag | error | | `title-modified` | the title text differs after whitespace and case are normalised | warning | | `h1-modified` | the first H1 differs, or exists on only one side | warning | - Canonical URLs are resolved against the page before they are compared, so a relative canonical equals the same absolute one. A trailing slash, the case of the host, a default port (`:443`) and a `#fragment` do not count as a difference. A different path, query, scheme or host does. - A raw page with no canonical where the script adds one that is the page's own URL is a **warning**, not an error. That is the usual shape of a single-page app: crawlers that do not run JavaScript pick a canonical themselves, and the one the script adds is the right one. A canonical the script adds that points anywhere else, one it changes, and one it removes still fail. - `noindex` is read from every `` and `` tag on the page, on both sides, with the name matched without regard to case; the page is `noindex` when any of them says `noindex` or `none`, as a token in the comma-separated list. A script that appends a second tag, so that the first still reads `index, follow`, is a mismatch. The finding says which side carries it: `raw` or `rendered`. - Only the first `

` is compared. - Titles and H1s compare equal when only whitespace or letter case differs. A finding shows both texts, cut at 200 characters. A page whose signals are the same on both sides passes all four checks. ## When it does not run The rule reports one skipped check, named `rendered-signal-mismatch`, with the reason, and never a pass or a fail, when: - the crawl browser-rendered the page, because the raw and the rendered HTML are then the same document. Run an audit that does not render the page to compare them; - there is no render result for the page: a CLI audit without the cloud, `--offline`, a render that was disabled, or one that returned no HTML; - the page is a soft 404, because both sides are then the error shell. ## Cost The rule reads the same cloud render that `ax/content-without-js` reads, which runs inside the audit and is billed as the page's own audit charge. It adds no credits. ## How to fix it Put the canonical and the robots meta in the HTML the server sends, and make client code leave them alone. ```html Pricing ``` In a framework that sets head tags on the client (a head manager, a router hook), move those tags into the server render, or into a static page, so they are in the first response. Keep the title and the first `

` in the server HTML for the same reason: what a crawler reads should be what a visitor sees, not a placeholder such as `Loading...` that a script replaces. | | | |---|---| | **Rule ID** | `crawl/rendered-signal-mismatch` | | **Category** | [Crawlability](/rules/crawl) | | **Scope** | Per-page | | **Severity** | error | | **Weight** | 6/10 | ## Enable / disable ### Disable this rule ```toml squirrel.toml [rules] disable = ["crawl/rendered-signal-mismatch"] ``` ### Disable all Crawlability rules ```toml squirrel.toml [rules] disable = ["crawl/*"] ``` ### Enable only this rule ```toml squirrel.toml [rules] enable = ["crawl/rendered-signal-mismatch"] ``` `enable` replaces the default `["*"]`, so this is the only rule selected. Leave `disable` out: a rule that matches any `disable` pattern is skipped even when `enable` names it. Only an `enabled` setting under a rule's own `[rule_options]` table overrides both lists. ## Related rules - [ax/content-without-js](/rules/ax/content-without-js): body content that only exists after JavaScript runs. - [core/canonical](/rules/core/canonical): the canonical on the raw page. - [core/robots-meta](/rules/core/robots-meta): the robots meta on the raw page. - [crawl/indexability](/rules/crawl/indexability): whether the page can be indexed at all.