Technical SEO Audit Checklist

robots.txt is at the root of each host and returns 200 or 404, never a 5xx.

Evidence: "curl -I" on the file for each protocol and host. A 5xx halts crawling for 12 hours; Google then reuses the last good copy for up to 30 days.

No robots.txt rule blocks a template you want indexed, or the CSS and JavaScript pages need to render.

Evidence: The "Crawl allowed?" field in URL Inspection for each sampled URL, and page indexing rows under "Blocked by robots.txt".

The XML sitemap is submitted or named in robots.txt, returns 200, and lists only absolute, canonical, indexable URLs.

Evidence: The Sitemaps report status, plus a crawl of the sitemap URLs exported with status, canonical and robots columns. Google crawls sitemap URLs exactly as written.

Every intended page has a plain HTML link (an "a" element with an "href") from an indexed page; no orphans.

Evidence: A crawl export filtered to zero internal inlinks, checked against the sitemap. Google only discovers "a href" links; the internal linking guide covers repairs.

No intended page carries noindex in its meta tag or an "X-Robots-Tag" header, and no page is both blocked in robots.txt and noindexed.

Evidence: The raw HTML head and "curl -I" headers per sampled URL, the "Indexing allowed?" field in URL Inspection, and page indexing rows under "Excluded by noindex tag".

Every indexable page declares one canonical: an absolute URL that returns 200 and is itself indexable.

Evidence: Crawl export columns for the canonical URL, its status and its robots state; count pages with none, two, or one pointing at a redirect or a noindexed page.

Google's selected canonical matches the declared one on every template.

Evidence: "User-declared canonical" and "Google-selected canonical" in URL Inspection (the live test does not check this), and page indexing rows under "Duplicate, Google chose different canonical than user".

Protocol and host variants (http, www, trailing slash, letter case, tracking parameters) reach the canonical URL in one permanent redirect, which internal links and the sitemap also use.

Evidence: "curl -I" on each variant of three sampled URLs recording every hop, and the crawl count of internal links to non-canonical URLs.

Sort, filter and pagination variants follow a written rule, and Bahasa Malaysia and English versions each declare their own URL as canonical.

Evidence: One rule per variant type with two URLs proving it, and the canonical tag read on both language versions of one page.

Each template's primary content (headings, copy, product or service details, prices) is in the raw HTML, or confirmed in the rendered HTML where JavaScript injects it.

Evidence: Two columns per sampled URL, "in raw HTML" (from "curl" or view-source) and "in rendered HTML" (from URL Inspection or the Rich Results Test), marked yes or no.

Title, meta description, canonical, robots meta and "a href" links are identical in raw and rendered HTML, and single-page routes use real paths rather than "#" fragments.

Evidence: The same comparison for the head plus a link count per template. Google may skip rendering when raw HTML contains noindex, so a script removing it later does not help.

Everything needed to render loads for Googlebot, and a missing product or bad route returns 404 or 410 (or noindex), never a 200 saying "not found".

Evidence: The live test's page resources and console output (its screenshot exists only for live tests), "curl -I" on a deliberately wrong URL per template, and page indexing rows under "Soft 404".

Every intended page returns 200 on a plain fetch and in URL Inspection; removed pages return 404 or 410.

Evidence: The crawl export status column for all sitemap URLs, "curl -I" on known removed URLs, and page indexing rows under "Not found (404)" and "Soft 404".

No sampled URL, and never robots.txt, returns 5xx or 429 during the audit window.

Evidence: Timestamped "curl -I" results, page indexing rows under "Server error (5xx)", and Googlebot entries in server logs if available.

Permanent moves use a server-side 301 or 308 reaching the target in one hop: no temporary codes, delayed meta refresh, JavaScript redirects, long chains or loops.

Evidence: A "curl -IL" trace listing the status and "Location" of every hop, and page indexing rows under "Redirect error".

A URL map lists every old URL and its one-to-one destination, each row tested; retired URLs return 404 or 410, not a catch-all redirect.

Evidence: The mapping spreadsheet with status and "Location" filled from a fetch of every old URL after launch. Google warns against pointing many old URLs at one unrelated page such as the home page.

Redirects stay for at least a year, old and new sitemaps were both submitted, and the Change of Address tool was used only for a domain change (not http-to-https, www or path moves).

Evidence: The Sitemaps report showing both files, the Change of Address status, a written retention date, and the page indexing report filtered by the new sitemap over successive weeks, new URLs rising as old ones fall.

Every finding is labelled measured, partial or hypothesis, carries at least three affected URLs and a site-wide count, and only measured findings become fixes.

Evidence: The findings table below with every column filled, and the "source" shown against each page indexing reason: the website or Google's systems. The Search Console guide explains each report.

Each fix has a named test and a verified-on date recorded in three states: raw response changed, live test passed, indexed version updated.

Evidence: Three dated entries per fix. Google says it may take months to revisit a page, and "Validate fix" in Search Console typically runs up to about two weeks.