How to set up Google Search Console, and the choice that comes first

Verification is the part everyone worries about and it is not where the mistakes happen. The consequential decision is made one screen earlier, when Google asks whether you are adding a domain property or a URL prefix property, and the two collect different data. Choose wrong and you spend a year looking at a partial picture without any indication that something is missing. There is no warning, no error and no obvious symptom, which is exactly what makes it worth two minutes of attention.

How do I set up Google Search Console?

By verifying the property and choosing between a domain property and a URL prefix, which is the decision that shapes everything after it. A new site should verify before it needs to, because data starts on the verification date; a mid-authority site usually needs both forms; a consolidated one also needs per-person permissions.

Quick answer

• The property type is the real decision. Domain covers everything; URL prefix covers one variant.
• Data starts accumulating from the moment you verify, never retroactively. So verify early.
• Submit the sitemap, and expect it to change nothing dramatic. It helps discovery, not ranking.
• Grant restricted access, not ownership, to anyone you hire.

Step 1: the property type, which decides what you see

A domain property covers every subdomain and both protocols at once: the www version, the bare domain, http, https, and any subdomain such as a blog or a shop. It is the complete picture and it can only be verified with a DNS record.

A URL prefix property covers exactly the address you typed and nothing else. https://www.example.com/ and https://example.com/ are two different properties, and if the site serves both, one of them will show data the other does not.

Pick the domain property whenever you can edit DNS. If you cannot, a URL prefix property is a perfectly usable second choice, and it is worth writing down which variant you verified so that a partial view is a known limitation rather than a mystery six months later.

Step 2: verification, which is the easy part

The methods available depend on the property type. For a domain property it is a DNS TXT record, added at whoever manages the domain, which is often not the same company as the hosting.

For a URL prefix, Google offers several: an HTML file uploaded to the root, a meta tag in the head, an existing Google Analytics tag, or Tag Manager. Any of them works, and each ties verification to something that can be removed later.

That last point causes a specific and slow failure. Google revalidates, and if the tag or file disappears, verification is withdrawn and access to the data goes with it. The notice arrives by email, late and without context. Whatever method you choose, the artefact has to survive the next redesign, which is why this site keeps its verification tag as a constant in the generator and audits its presence on every build.

Step 3: submit the sitemap

Under Sitemaps, submit the URL, usually /sitemap.xml. It helps Google discover pages, particularly on a site with weak internal linking, and it does not make anything rank.

Two things worth knowing. A URL in the sitemap is a suggestion rather than an instruction, so pages can be excluded after being read. And a sitemap containing pages that are blocked or marked noindex sends contradictory signals, which is a common state on sites where the sitemap is generated by a plugin nobody configured.

Step 4: give access without giving away control

If you hire anyone, they need Search Console access, and they need it at restricted level. That reads every report.

Ownership is different: it can add and remove other owners, and removing yourself as owner is uncomfortable to reverse if the relationship ends badly. A provider who insists on ownership rather than restricted access should be asked why. It is one of the questions worth asking before signing anything, along with the rest in how to choose an SEO agency.

Search Console and Analytics are not the same measurement

They get set up together and then get read as if they described the same thing, which produces arguments about numbers that never had a reason to match.

Search Console measures what happens on Google: impressions, positions and clicks, before anyone arrives. It is the only place the query is visible, because Google does not pass it to your site.

Analytics measures what happens after arrival: sessions, pages, conversions, from every source. It cannot tell you which query brought someone from organic search.

So the click count in one and the session count in the other will differ, permanently and for good reasons: different counting rules, different handling of bots and consent, different time zones. Chasing the gap is a well-known way to lose a week. The two answer different questions and the useful pattern is reading them in sequence: Search Console to decide what to write, Analytics to see whether it did anything for the business.

Linking the two accounts is worth doing because it saves navigation, and it does not merge the measurements.

What to do in the first week: nothing

Search Console is retrospective. There is no data on day one and very little in week one, and the reports look broken when they are simply empty.

The exception is the URL Inspection tool, which works immediately and is the most useful thing in the whole product. It fetches a page as Google sees it right now, and it is where a diagnosis starts when a page is missing, as described in why is my page not indexed.

Once a few weeks of data exist, the queries report becomes the most valuable free source of keyword information available, and how to work it is in keyword research without paid tools.

When this is not the priority

Never, honestly, because it costs fifteen minutes and everything else depends on it. The closer question is what to do afterwards, and the answer is usually to wait rather than to act on two weeks of noise.

What is worth avoiding is drawing conclusions early. Positions oscillate, and a rolling average built from a short window describes noise with great confidence.

What the reports do and do not separate

One thing is worth knowing before the first report is opened, because it prevents a whole category of wrong conclusion.

Google documents that traffic from its AI features, including AI Overviews and AI Mode, is included in the performance report, inside the web search type. A click from an AI summary to an external page counts as a click, and pages that appear count as impressions.

What does not exist is any filter to separate them. It arrives mixed with the rest of search traffic and there is no published way to isolate it. The two wrong conclusions are symmetrical: it is not missing, and it cannot be pulled apart.

A second detail changes how positions read: an AI Overview occupies a single position, and every link inside it is assigned that same position. There is no first or third place inside a summary, which is why average position is a poor signal for detecting one.

What to look at in the first week

Verifying the property does nothing if you then open the wrong report. This is the order that pays.

The pages report, which tells you how many URLs are indexed and why the rest are not. It is the only question that matters in month one, and the diagnosis path is in why is my page not indexed.

The performance report, the one everyone opens and almost nobody reads correctly, because average position moves with the mix rather than with your ranking.

Core Web Vitals, which brings field data rather than lab data, with the distinction explained in Core Web Vitals in Search Console.

And manual actions, a two-minute check that closes half the conversations about traffic drops before they start.

If the business serves customers at an address, most of what decides the result lives somewhere else entirely: local SEO for small business and how to add your business to Google Maps. What the whole job looks like from the top is in what is SEO.

Mistakes that repeat

• Choosing URL prefix without noting which variant. A partial view with no symptom.
• Verifying with a tag that a redesign will delete. Verification is withdrawn silently.
• Expecting the sitemap to improve rankings. It aids discovery and nothing else.
• Granting ownership to a provider. Restricted access reads the same data and reverses cleanly.
• Reading the reports in week one. There is nothing there yet, and it is not broken.
• Waiting until you need the data. It is never backfilled, so the wait costs history.

Data and transparency

That AI features traffic is included in the performance report within the web search type, that clicks from AI Overviews and AI Mode count as clicks, and that an AI Overview occupies a single position with all its links assigned that position, comes from Google's public documentation on AI features and on the performance report, checked on 12 August 2026. No filter to separate that traffic is documented.

The distinction between domain and URL prefix properties, the verification methods available for each, the fact that domain properties require DNS verification, the revalidation of ownership and the withdrawal of access when verification fails, and the role of sitemaps as a discovery aid rather than a ranking factor are all documented publicly by Google in the Search Console help, current as of August 2026. That data is not backfilled before verification is also Google's documented behaviour. The practice of keeping the verification tag as a build-time constant and auditing its presence on every page is this site's own, described because it is verifiable in the pages themselves. No figures appear here on how long data takes to become useful: it depends entirely on traffic volume, and any general number would be invented. Verified as of August 2026.

What this changes

The setup is treated as an administrative box to tick, and there is a reason it deserves more than that.

Everything downstream of it, every decision about what to write, what to fix and what to leave alone, gets made from this one data source or gets made by guessing. The property type decides how much of your site that source describes, and it is the only screen in the whole process where a wrong answer produces no error, no warning and no visible symptom. Two minutes there is worth more than the next six months of reading the reports.

Frequently asked questions