How to Connect Analytics to an AI-Built Site in 2026

Connect Google Analytics from the integrations panel and paste your measurement ID. On Zugo that is the whole setup: the tag goes into the published site, data starts arriving within minutes, and you confirm it in the realtime report by opening your own page. No code, no tag manager, no plugin.

The setup takes about five minutes. Deciding what to look at afterwards takes longer, and it is the part that separates a useful measurement habit from a dashboard nobody opens twice.

What do you actually need before you start?

A Google Analytics account with a property for this site, and the measurement ID that property gives you. The ID looks like G-XXXXXXXXXX and lives in the property's data stream settings. That is the only value you have to move between the two systems.

Create the property before the site is live if you can. A property created afterwards starts counting from its creation date and cannot backfill, so a week of launch traffic is simply absent from your history.

One property per site is the correct default. Putting three different projects into one property saves an hour now and costs you every comparison later, because you cannot cleanly separate what you never separated at the start.

If you have a custom domain planned, still connect analytics on the zugo.run address first. Traffic to both is measured by the same tag, and moving the domain later does not require a new property.

How do you connect it, step by step?

Five steps, in order, and none of them involve editing code.

Step 5 is not optional. An unverified analytics setup is the most common silent failure in small sites: everything looks connected, the dashboard shows zero, and you assume nobody visited for three weeks.

Check it from a phone on mobile data rather than your own network, and avoid checking from a browser with tracking protection turned all the way up, since your own visit may be blocked and you will misread that as a broken tag.

If realtime stays empty after a few minutes, the usual causes are a mistyped ID, a project that was edited but not republished, or a browser extension blocking the request. Work through them in that order.

What should a small site actually measure?

Four numbers. Everything else in the interface is available to you and almost none of it will change a decision you make this month.

Conversions are the row people skip and the only one that matters. Set up an event for the action your site exists for: the contact form submitted, the booking requested, the checkout completed. A site with traffic and no conversion tracking is a site you cannot improve on purpose.

Bounce rate and time on page will tempt you. On a small site they are noisy enough to be misleading, and reacting to them produces changes that feel productive and move nothing. Ignore them until you have hundreds of sessions a week.

Compare like with like. A jump in sessions after you posted in a forum is not evidence your page improved. Hold the traffic source constant when you judge a change to the page, or you are measuring the audience rather than the work.

Does the custom domain change anything?

The tag follows the site, so a project measured on yourname.zugo.run keeps reporting after you connect your own domain. You do not need a new property or a new ID.

What does change is your referral picture, and it changes for the better. Traffic that arrived through the temporary address and traffic on your own domain end up in the same property, so a domain switch shows up as continuity rather than a cliff. The custom domain guide covers the DNS records.

One thing worth doing at the same time: add the site to Google Search Console. Analytics tells you what people did after they arrived; Search Console tells you what they searched to get there. Together they answer questions neither answers alone.

The one habit worth adopting on the day you connect a domain is an annotation. Write down the date of the switch somewhere you will find it, because three months later you will look at a traffic dip and spend an afternoon deciding whether it was the domain, a Google update, or a page you changed the same week.

Republish after any integration change. The tag lives in the published site rather than in the editor, so a project that was edited but not published carries the previous configuration, and that is the single most common reason a correctly entered ID reports nothing.

What about consent, cookies, and privacy?

Google Analytics sets cookies, which puts it inside privacy law in most of the world. In the EU and UK that means a consent mechanism before analytics loads, and in several other jurisdictions it means at minimum a privacy policy that names what you collect.

Zugo can build a consent banner and a privacy page into your site through normal edits, and the wording is yours or your lawyer's. What no builder can do is tell you which regime applies to your visitors, because that depends on where they are and what you do with the data.

The practical minimum for a small site: a privacy page that says you use Google Analytics, what it collects, and how to opt out, plus a consent banner if you have European visitors. That is not legal advice, it is the floor.

If the compliance overhead outweighs the value, the honest alternative is to measure less. A site that counts form submissions in its own Supabase table and nothing else is a legitimate choice, and for a small brochure site it is often the right one.

What can analytics not tell you?

Three limits, each of which has caused someone to draw a confident wrong conclusion.

It does not tell you why. You will see that most visitors leave the pricing page. Analytics cannot say whether the price is too high, the page is confusing, or those visitors were never buyers. That answer comes from asking five people.

It does not count everyone. Ad blockers, tracking protection, and consent refusals all remove visitors from the data. Your real traffic is higher than the number, by an amount that varies by audience. Technical audiences under-report the most.

It does not make a build correct. Every Zugo project is booted in a sandbox before it reaches you and a failure is reported as a failure, which lowers the risk of publishing a blank page, but it does not remove the need to check the result yourself. Analytics measures a site that works; it does not verify one.

Connect the tag, verify it in realtime, set one conversion event, and then leave it alone for a fortnight. The contact form guide covers the event most small sites should be counting, and the landing page guide covers what to change once the numbers tell you where people are leaving.