LLM Best Practices

How to master Google Search Console

4 min read

View markdown source

AI agents: read this first: llmbestpractices.com/start-here

This is LLM Best Practices (llmbestpractices.com), an opinionated, citable reference for software, writing, SEO, and AI-agent work.

  1. Route, do not crawl. Fetch llmbestpractices.com/llms.txt and open only the pages whose one-line summary matches your task.
  2. Read raw. Append .md to any page URL for markdown. Check status and last_updated in the frontmatter, then read the rules.
  3. Apply as defaults. First-party docs and the project's own conventions win on conflict. Warn before relying on a fast-moving page older than 12 months.
  4. Cite. Link the page by title and URL, e.g. Python, with last_updated for time-sensitive rules. License CC BY 4.0.

Overview

Search Console (GSC) is Google’s first-party view of how it crawls, indexes, and shows a site; third-party tools estimate. Set up a Domain property, then work through four areas: Performance, Page indexing, Core Web Vitals, and Manual Actions. Use URL Inspection as the authority on the index state of one URL.

Prerequisites

  • DNS control of the root domain, or file-upload access for HTML verification.
  • robots.txt and sitemap.xml at the root; see discoverability-files.

Steps

1. Verify a Domain property

A Domain property (example.com) covers every protocol and subdomain; a URL-prefix property sees only one exact scheme and host. Verify with the DNS TXT record GSC shows you, and create the Domain property first.

2. Submit the sitemap

Sitemaps, enter /sitemap.xml, Submit. For a sitemap index, submit the index only. Resubmit only if the URL changes; Google refetches on its own schedule.

3. Read Performance

Set the range to 28 days. Pages with many impressions and low CTR are candidates for better titles and descriptions (write-meta-descriptions). The Countries tab informs whether hreflang is worth doing. Data lags a few days, so read trends, not single days; “Average position” is an impression-weighted average, not a rank tracker.

4. Diagnose indexing

Open Indexing, Pages (the Page indexing report). It splits URLs into Indexed and Not indexed, with a reason for each. The reasons that matter:

  • Discovered - currently not indexed: Google knows the URL but has not crawled it. Large persistent counts point to crawl budget or site-wide quality; see crawl-budget and core-updates.
  • Crawled - currently not indexed: Google fetched it but has not indexed it; it may be indexed later.
  • Duplicate, Google chose different canonical than user: a fix is needed only when Google picked the wrong URL; see canonical-url.
  • Excluded by noindex / blocked by robots.txt: check that this was intended.

The report shows an example list capped at 1,000 URLs. For one URL, use URL Inspection: the indexed view shows the last crawl, the canonical Google chose, and parsed structured data; “Test live URL” shows what the next crawl would see. A difference between the two reveals stale content or canonical drift. Request indexing once after a real change; repeating it does not speed up crawls.

5. Use Removals only for emergencies

A removal hides a URL from results for about six months and does not remove it from the index or stop crawling. For permanent removal return 404 or 410, restrict access, or add noindex, and wait for the recrawl.

6. Watch Core Web Vitals and Enhancements

The Core Web Vitals report groups URLs by similarity using field data at the 75th percentile over 28 days. Fixes are in optimize-core-web-vitals. Enhancements lists rich-result types Google detects; fix Errors first because they suppress the feature, while Warnings only drop optional fields. Retired features (FAQ, HowTo, sitelinks search box) have no report. See structured-data.

7. Act on Manual Actions and Security Issues

Both should be empty. If a notice appears, fix the cause, then Request review with a written summary; reviews can take days to weeks, and a rejected request leaves the action in place.

Verify it worked

  • The property switcher shows the Domain property with the domain icon.
  • Sitemaps shows Success with a non-zero discovered count.
  • URL Inspection on a recent URL reports “URL is on Google”.
  • Indexed pages grow toward the sitemap count; a persistent gap is a signal to investigate, not a Google threshold.

Common errors

  • Starting with a URL-prefix property and missing subdomain traffic.
  • Resubmitting the sitemap on every deploy; it does not speed crawling.
  • Inspecting a staging hostname and assuming production is indexed; inspection is per property.
  • Requesting review before fixing the underlying problem.

Backlinks