Search Console is the only place where Google tells you what it sees of your site: the queries that surface your pages, the ones that earn clicks, the URLs it refuses to index, the files it cannot read. The tool is free. Its value is not in the setup but in reading the reports regularly and knowing what each number is worth.
What follows is a reading guide, not a setup tutorial. It assumes your property is verified and collecting data. We go through the reports that support a decision — Performance, generative AI, Page indexing, Sitemaps, Core Web Vitals, robots.txt — and what their figures mean.
Before you read: what Search Console actually measures
The data is not real time. Google states that it is generally available within two to three days, with roughly a week of delay for a recently added site (About Search Console data). History stops at 16 months. Beyond that, only your own exports, or the BigQuery bulk export, let you compare one year with another.
Two invisible filters explain most reported discrepancies. The first is anonymised queries: Google removes any query that was not typed by more than a few dozen users over two to three months (Google Search Central, October 2022). The second is truncation: the interface displays and exports at most 1,000 rows per table, so the visible rows rarely add up to the total shown at the top. Past that ceiling you need the Search Analytics API or the Looker Studio connector: 25,000 rows per call, up to 50,000 per day, per site and per search type.
The Performance report: four metrics, one method
The report rests on clicks, impressions, average CTR and average position. The last is an impression-weighted average: it degrades mechanically as soon as a page appears, even badly ranked, on new queries. Never read it alone.
The most productive reading is always the same. Sort queries by impressions, keep those between positions 5 and 15 with a weak CTR, then look at the page attached. Google already knows that page; what fails is the title, the meta description or the angle. The example that opened the first version of this article, in 2014, still holds: on a site I followed then, Search Console listed dozens of queries collecting impressions and no clicks, because no page really dealt with the subject. Rewriting a title is cheap. Writing the missing page is the real work, and it belongs to your SEO content plan, not a technical backlog.
What changed in the report since late 2025
- Weekly and monthly granularity (10 December 2025): a selector replaces the default daily view, so weekend patterns stop distorting comparisons. A 24-hour view also exists, with preliminary data: keep it for watching an incident, not for reporting.
- Branded query filter (20 November 2025, extended to all eligible sites on 11 March 2026): it separates queries containing your brand from the rest. Classification relies on an internal, AI-assisted system, not a regular expression, so some queries are misclassified; it exists only for top-level properties with enough volume.
- Custom annotations (17 November 2025): right-clicking the chart adds a dated note of up to 120 characters. A redesign, a migration, a plugin change — the simplest way to stop wondering what happened in March. Notes are visible to every user of the property.
- AI-assisted setup (4 December 2025): you describe the analysis in plain language and the tool applies filters and comparisons. Experimental, limited rollout, and Google asks you to check the filters proposed.
The generative AI reports: what they show, and no more
Announced on 3 June 2026 on the Google Search Central blog, the generative AI performance reports were rolled out to all sites worldwide on 31 August 2026, in two versions: Search and Discover.
They show impressions, and nothing else: no clicks, no CTR, no position. Those impressions are broken down by page, country, device (search results only) and date, with hourly, daily, weekly or monthly granularity. The scope covers AI Overviews and AI Mode in Search, excluding Search Labs experiments. They are already counted in the overall Performance report: the dedicated view is a separate lens, not an extra counter.
You therefore know which pages AI answers draw on, not what that is worth. Google has said more metrics will follow. Until then, cross these impressions with organic clicks on the same pages, and treat the causal link as unproven.
Page indexing: separating anomalies from choices
This report sets indexed URLs against non-indexed ones, each exclusion carrying a reason: blocked by robots.txt, marked "noindex", duplicate page, soft 404, server error, "Crawled — currently not indexed", "Discovered — currently not indexed". Google is explicit that not every URL is meant to be indexed: a basket page in noindex, or a duplicate URL correctly canonicalised, is a normal exclusion.
Three habits are enough. Filter by sitemap first: the pages you declared and Google does not index are the real subject, the rest is often technical noise. Then open each reason and read the URL examples, capped at 1,000 per issue. Once the cause is fixed, start the validation and allow two weeks for the recrawl. If a page stays out after that, look upstream — internal linking, content quality, a server response — and our guidance on Google indexing takes the statuses one by one.
Sitemaps: three pieces of information, no more
The Sitemaps report can be read in thirty seconds. The status says whether the file was fetched cleanly, with errors, or not at all. The last read date exposes a file that has been forgotten or has become unreachable. The number of discovered pages is compared with the pages you have published: a wide gap points to an incomplete or frozen sitemap, common where the plugin generating it was disabled during a redesign.
Then watch for the frequent misreading. A URL discovered in a sitemap is not crawled, let alone indexed, for that reason alone; the Page indexing report is what settles it. Official limits: 50 MB uncompressed and 50,000 URLs per file, 50,000 sitemaps per index file.
Experience: Core Web Vitals and HTTPS
The Core Web Vitals report is not a laboratory test: it draws on real measurements from Chrome users (the CrUX dataset) over a rolling 28-day window. For the "Good" category, the documented thresholds are LCP under 2.5 s, INP of 200 ms or less and CLS of 0.1 or less. The table is limited to 200 rows and URLs are grouped by similarity, so fixing one page moves a whole group.
An empty report is not good news: it mostly means your audience is too small to reach the minimum data threshold. Fall back on PageSpeed Insights and Lighthouse, and treat their scores as diagnostics rather than targets, as we argue in our notes on website speed. The neighbouring HTTPS report lists URLs still served over HTTP. The Mobile Usability report was retired on 1 December 2023.
robots.txt and crawl statistics
The robots.txt report, in Settings, shows the files found for the property's 20 main hosts, ordered by crawl rate, with up to two origins per host (HTTP and HTTPS). For each file: last crawl date, errors and warnings, and a history of the versions from the last 30 days. This is the first place to look when a site disappears from the results after a release — a staging file shipped to production remains a common cause. It exists only for a domain property or a URL prefix with no path.
Crawl statistics, in the same place, show the volume of Googlebot requests, the response codes and the host status. They exist only for root-level properties and are not attributed to canonical URLs. Read them as a server diagnostic, not a ranking indicator; when they look wrong, the follow-up belongs in a technical SEO audit.
Links, manual actions, security
The Links report lists your most linked pages, linking sites and anchor texts. Its tables are capped at 1,000 rows and Google is clear that it is a sample, not an inventory. Manual actions and Security issues should stay empty; if either is not, it takes priority over the rest.
A realistic reading routine
Checking Search Console every day is pointless for a small business site: the noise is larger than the signal. A fixed rhythm, tied to what you change on the site, produces better decisions.
| Report | Frequency | What you are looking for |
|---|---|---|
| Performance | Monthly | Comparison with the previous period, queries with unrealised potential |
| Page indexing | After each change to the site | Useful pages missing from the index, a sudden rise in one exclusion reason |
| Sitemaps and robots.txt | After a release | Last read date, errors, accidental blocking |
| Manual actions and security | On notification | No alert; otherwise, immediate handling |
| Core Web Vitals | Quarterly | URL groups that have moved to "Poor" |
One last point of method. Before concluding that traffic has fallen, check the period compared, the search type and the absence of a leftover filter. Many reported "traffic drops" are badly framed comparisons: 28 days against 30, or web results against all types. Rule that out first.
Common questions
How long does Search Console keep the data?
Google keeps Search Console data for the last 16 months; beyond that it is no longer available in the interface. Export your reports regularly, or set up the BigQuery bulk export, to compare several years.
Why does the total number of clicks not match the sum of the queries?
Two reasons combine. The interface tables are limited to 1,000 rows, and Google excludes anonymised queries — those typed by fewer than a few dozen users over two to three months. The total at the top of the report remains the reference figure.
What does the generative AI performance report show?
Impressions only, broken down by page, country, device and date, for AI Overviews and AI Mode in Google Search, and for Discover. It provides no clicks, no CTR and no position. Google rolled it out worldwide on 31 August 2026.
Why do the figures differ from Google Analytics?
Search Console counts clicks in Google results; Analytics counts sessions measured by a script on the site. Time zones, bot filtering and counting methods all differ. Expect the trends to agree, not the exact values.
My Core Web Vitals report is empty, is that a problem?
No. The report relies on real measurements from Chrome users over a rolling 28-day window, and a minimum volume of data is needed before a URL group appears. A low-traffic site does not reach it. Use PageSpeed Insights or Lighthouse instead.