Google Search Console is the free service Google uses to show site owners how their site appears in Search: how many times it was shown, for which queries, which pages are indexed and which are not. It is the only source of organic data that comes straight from Google, with no third-party estimates in between.
It is also a partial source, and further down we measure that on two real sites. Over the same quarter, the clicks Search Console attributes to a specific query are 34.5% and 38.6% of the total the same tool reports in its chart. The other clicks exist and are counted, but they don't get a row in the table.
How to set up a property
You start at search.google.com/search-console with a Google account and add a property. The first choice is between two types, and it can't be reversed without starting again.
- Domain property: collects everything under the domain, including subdomains and protocols. It can only be verified with a DNS record, so you need access to the registrar's control panel.
- URL prefix: covers only the exact combination of protocol and host you enter. It can be verified with an HTML file, a tag in the head, Google Analytics, Google Tag Manager or DNS.
For a site that lives on a single host the difference is marginal. For an ecommerce site with a support subdomain, a separate blog and an international version, a domain property saves you reading four disconnected dashboards. If the site has several languages, keeping separate URL prefix properties as well is still useful: the Pages dimension isn't enough to isolate one market quickly.
Nothing appears after verification. Data arrives with a delay that Google's documentation puts at two or three days, and on a new property the first weeks are inevitably empty.
The reports you actually use
Search Console has around twenty screens. In day-to-day work you open five.
Performance
Four metrics (clicks, impressions, CTR, average position) crossed with five dimensions (queries, pages, countries, devices, search appearance). The default view shows the last three months. This is where you read the queries that bring traffic, the ones that bring only impressions, and the pages where the CTR is out of line with the position.
Hourly granularity, with the 24-hour window, is the newest and least used part: it shows the effect of a new page or a fix within a few hours instead of three days later. The data in that view is preliminary and changes, so read it as a sign of direction, not a final number.
Page indexing
It splits the URLs Google knows into indexed and not indexed, with the reason for each group: alternate page with proper canonical tag, crawled but currently not indexed, discovered but not yet crawled, excluded by noindex. Read the count of non-indexed pages carefully, because it also includes URLs you excluded on purpose, such as a blog's tag pages. The Sitemaps report gives a cleaner reading, because it only compares the URLs you declared in the sitemap.
When a new site doesn't show up for its own name, this is the first place to look: nine times out of ten the answer is here and not in a penalty.
URL Inspection
It returns the status of a single address: whether it is indexed, which canonical version Google chose, which structured data it read and, with the live test, the HTML Googlebot actually receives. On a site built in JavaScript, that comparison between static and rendered HTML is often the only way to understand why content visible in the browser doesn't make it into the index.
Experience and Core Web Vitals
The data comes from the Chrome UX Report, meaning real Chrome users, not a lab test. That is why it moves slowly and a change made today shows up in the report weeks later. There are three thresholds: LCP, INP and CLS. What they really change and how much they weigh against other ranking factors is a longer discussion than the dashboard suggests, and page speed remains a condition, not a lever.
Links and manual actions
The Links report lists the domains linking to you, the most linked pages and the anchor text. It is the only free source of backlink data that isn't an estimate from a third-party crawler, although it updates slowly and doesn't replace a link building tool. Manual actions and Security issues are the two screens that are almost always empty: when they aren't, they become the day's top priority.
How many clicks the Performance report attributes to no query
This is the part most guides leave out. Using the Search Analytics API, we pulled data from two Italian properties we manage, over the same 92-day window (1 June to 31 August 2026, Web search only) and on the same metric, changing only the grouping dimension. The names stay out because the numbers have never been published.
| How the data is grouped | Property A (B2B industrial site) | Property B (B2B SME) |
|---|---|---|
| Total property clicks | 2,840 | 624 |
| Clicks summing the Query dimension rows | 981 | 241 |
| Share of clicks attributable to a query | 34.5% | 38.6% |
| Total property impressions | 244,383 | 12,922 |
| Impressions summing the Query rows | 172,965 (70.8%) | 3,977 (30.8%) |
| Clicks summing the Page dimension rows | 2,889 (101.7%) | 638 (102.2%) |
| Impressions summing the Page rows | 307,809 (126.0%) | 15,940 (123.4%) |
Two readings, both counterintuitive. On the Query dimension, two clicks out of three are lost. On the Page dimension the opposite happens: impressions rise 26% above the property total.
These are two different mechanisms, and it is worth keeping them apart because they lead to different mistakes.
First mechanism: the privacy filter
Google calls the queries left out of the tables to protect the people who typed them anonymised queries. The threshold is public. In the post A deep dive into Search Console performance data filtering and limits (Search Central, 19 October 2022), Daniel Waisberg writes that queries are anonymised when they are "not issued by more than a few dozen users over a two-to-three month period".
The behaviour is precise: those queries never get a row in the table or in the API, but they are included in the chart totals, unless you apply a filter on the query dimension. This produces two discrepancies that look like bugs and aren't: the sum of the rows doesn't match the total, and "contains X" plus "doesn't contain X" stays below the total.
The practical consequence is significant. On property A, 1,859 clicks out of 2,840 come from searches too rare to be shown individually. That is pure long tail, and it is why keyword research built only on the queries visible in Search Console describes a third of real demand. On a technical B2B site, where every buyer searches with their own variant of a product name, that share is structural and can't be recovered.
Second mechanism: property versus page
The other discrepancy is about aggregation, and it is documented in Google's help page on the Performance report. Data grouped by Queries, Countries, Devices or Dates is aggregated by property. Data grouped by Pages or Search appearance is aggregated by page.
The help page explains it in one line: if two results from the same site appear for one query, they count as a single impression in the chart total. In the Pages table, that same SERP counts as two impressions, one per URL. Hence the 126%.
Anyone building a monthly report by adding up impressions per page and comparing it with last month's total taken from the chart is comparing two different units. When the number then moves, the diagnosis of the traffic drop starts off on the wrong foot.
Average position is not the position you see
The third misunderstanding, and the most common. In the chart, average position is the position of your site's highest result, averaged across all impressions in the period. In the table it is the average for the row. A page that is second for a query with 40 searches a month and eighth for a query with 4,000 will have an average of around 7.9, and that average matches no position anyone has ever seen.
Google adds a warning worth reading in full: even if a query appears in your list, you might not see your site when you run the same query in Google Search, because results depend on the time, place, device and recent history of the person searching.
In other words, searching for your own keyword by hand to check a Search Console figure is not a check. For that you need a rank tracker that always queries from the same location and device, and comparing the two series is one of the most useful things you can put in an SEO KPI dashboard.
Over the last two years average position has lost even more meaning, because AI Overviews push the first blue link below a generated block, and Search Console doesn't separate those impressions from the rest. A falling CTR at the same position is now most often the sign of a SERP that has filled up above you, not of a badly written title. The zero-click search trend shows up here before anywhere else. Which question you take to Search Console and which to GA4 is the starting point of SEO analytics.
How to get the data the interface doesn't show
On top of the privacy limits there are volume limits, and these also have public numbers, from the same Search Central post:
- Export from the interface stops at 1,000 rows.
- The Search Analytics API goes up to 50,000 rows a day per site and per search type. It returns 1,000 rows by default, you raise that to 25,000 with the
rowLimitparameter and get the next 25,000 by paging withstartRow. - For requests that don't involve queries or URLs, such as countries, devices and search appearance, Google exports everything.
The next step is bulk data export to BigQuery, announced on 21 February 2023 by Daniel Waisberg, Gaal Yahas and Haim Daniel. You set it up once from the property settings and from then on Google writes three tables every day: searchdata_site_impression, searchdata_url_impression and ExportLog.
Two points that are often skipped. First, bulk export is not retroactive, so the day you switch it on decides from when you will have full history. Second, stated plainly in the announcement, the export contains all performance data except anonymised queries. BigQuery removes the row limit, not the privacy filter. The two thirds of clicks missing from the table above don't come back there either.
For a site under a few thousand URLs, the interface and the API are enough. The export is for when pages run into the tens of thousands, or when you want to join search data with Google Analytics, the CRM or the ERP on the same key.
Platform properties for Instagram, TikTok, X and YouTube
Since July 2026 Search Console no longer covers only websites. Platform properties let you verify an account on four platforms (Instagram, TikTok, X and YouTube) and see how that content performs in Google Search, Discover and Google News.
The metrics are the same as for sites: total clicks, impressions, average CTR and average position, with a default 28-day window. The official documentation draws a clear line, though: "Platform properties only show how your content performs on Google Search. They don't track when people see your content on the platform itself". This is not Instagram analytics; it is the slice of Instagram that passes through Google.
For a company that publishes product videos on YouTube and product pages on its site, this is the first time the two channels can be read on the same scale. They are worth switching on just to find out which queries lead to the profile instead of the site: it happens more often than you'd expect, and it changes the search intent you are serving.
What to do with it on a small business site
Context helps set expectations. According to the ONS article Artificial intelligence in UK businesses (20 July 2026, Business Insights and Conditions Survey), around 35% of UK businesses with ten or more employees reported using AI in June 2026, up from around 12% in late 2023; the figure is 28% for businesses with fewer than ten employees and 49% for those with 250 or more. Among smaller firms, most digital tools are still used at a basic level.
For most small and medium businesses, sophisticated analysis of Search Console data is rarely the first problem. The first problem is that the property doesn't exist, or it is verified in the name of an agency the relationship ended with three years ago. The three things that make the most difference, in order:
- Verify a domain property in the company's name, not the supplier's, and set up the BigQuery export straight away if the site has more than ten thousand pages.
- Look at the Indexing report before the Performance report. A page that isn't in the index has no data to analyse, and that is half the work of an SEO audit.
- Stop using the sum of the rows as if it were the total. That goes for client reports and for year-on-year comparisons.
An example of what happens when that data is read properly and used for two years in a row is our case study on SEO for manufacturing. If the point is to bring that work into a project with commercial goals, we cover it on our SEO services page.
Frequently asked questions
Yes, it is free with no plan limits. All you need is a Google account and a way to verify ownership of the site, with a DNS record or with a file, a meta tag, Google Analytics or Google Tag Manager.
Because queries searched by no more than a few dozen users over two or three months are anonymised to protect the privacy of the people who typed them. They don't appear in the table or in the API, but they are counted in the chart total. On two Italian properties we measured between June and August 2026, the clicks attributable to a query were 34.5% and 38.6% of the total.
Search Console measures what happens before the click, inside Google's results: impressions, queries, position, indexing status. Google Analytics measures what happens after the click, on the site: page views, paths, conversions. They are not alternatives and their numbers never match, because they count different things.
Google's documentation generally gives two or three days between collection and availability in the reports. The 24-hour view shows preliminary data, marked by a dotted line, which can change over the following hours.
A domain property if you have DNS access: it covers subdomains and protocols in one dashboard. A URL prefix is still handy alongside it when you want to isolate a section or a language without filtering the Pages dimension every time.
No. Bulk data export removes the row limit of the interface and the API, but Google's official announcement specifies that anonymised queries are excluded there too. It recovers volume, not rare searches.
No. Impressions and clicks generated inside Google's AI experiences are included in the Performance report totals without a dedicated dimension. That is why a CTR falling at the same average position should now be read by looking at how the SERP has changed, not only the title.
One last thing the industry rarely says. The share of clicks Search Console attributes to no query is not noise to discard: it measures how much of your traffic depends on searches that no keyword research tool will ever see, because below the volume threshold they don't exist for anyone. The higher that share, the more the site is capturing demand that competitors can't even plan for. It is worth calculating every quarter and watching whether it rises.
