Google Analytics 4 for people who do not want to learn Analytics

GA4 is genuinely harder to use than the version it replaced. That is not your failing. Google rebuilt the data model around events instead of pageviews, which is better for the people who need it and worse for everyone who just wants to know whether the website is producing calls.

The good news is that the hard parts are the parts you do not need. A business owner who spends two hours once will get everything useful out of GA4 and never open the exploration builder.

What are the only three things I need from GA4?

Traffic volume, traffic source, and conversions. In that order of ease and reverse order of importance.

Traffic volume is sessions or users over time. It answers whether more people are finding you. It is the least interesting of the three and the one most reports lead with.

Traffic source tells you which channel sent them. Organic search, paid search, direct, referral, social, email. This is where the decisions live, because it is the only view that connects money spent to people arriving.

Conversions are the events you defined as mattering. A form submit, a phone tap, a booking confirmation, a menu download. If you have not marked any events as conversions, GA4 is a traffic counter and nothing more. That configuration step is the entire difference between a tool that informs decisions and a tool that produces charts.

How do I set conversions up without learning event syntax?

Most of it is already happening. GA4 collects a set of events automatically once the tag is installed, and the enhanced measurement setting captures outbound clicks, file downloads, and form interactions without any code.

Go to Admin, then Events. Look at the list of events already firing. Find the ones that represent a real outcome. Toggle them to count as key events. That is the whole job for most small sites.

Two events are worth adding manually because they carry most of the value for local businesses. A click on a phone number link, and a click on a directions or map link. Both are single clicks on the element, tagged through Google Tag Manager, and any developer will do them in under an hour.

Phone calls deserve their own attention beyond the tap, because a tap on a number is not a conversation and counting calls properly is a different exercise from counting clicks.

If your revenue actually happens in a point of sale or booking system rather than on the website, GA4 will never see the outcome. That gap is fixable, and joining transaction data to marketing data is where the real picture appears.

What does the gap between a key event and a real outcome look like?

Here is the reconciliation. Run it with one month of your own data.

What GA4 reports. 41 key events. Twelve form submissions and 29 taps on the phone number link.

What the phone system reports. 18 inbound calls that connected and ran longer than thirty seconds. Eleven of those came from the website, based on the tracking number placed there.

What the booking log reports. Nine jobs traceable to a form submission, out of twelve submissions. Three were spam or duplicates.

Do the arithmetic. Twenty-nine taps produced eleven connected calls from the site, which is 38%derived. The other 18 taps are people who tapped and hung up, tapped twice, or were already holding the phone. Add the nine real form outcomes and the month produced 20 outcomes against 41 reported key events, so GA4 is overstating by roughly a factor of two.

Why this matters more than it looks. If you are bidding on ads against the GA4 number, you are paying for taps. Your reported cost per conversion is roughly half your true cost per customer, and every channel comparison built on it is distorted in the same direction. LocaliQ's 2026 search advertising benchmarks put the median conversion rate across all industries at 8.18% and the median cost per lead at $66.69, using thousands of customer campaigns, and those are lead-level figures. Comparing your tap-inflated rate to them is comparing two different things.

The fix is not a better report. It is one tracking number on the website and a monthly count of connected calls, which turns the phone from an estimate into a measurement.

Which reports should I bookmark and which should I ignore?

Bookmark two. Reports, then Acquisition, then Traffic acquisition, with the date range set to last twenty-eight days compared to the previous period. And Reports, then Engagement, then Events, filtered to your key events.

That is it. Those two answer where people came from and what they did.

Ignore the Realtime report entirely. It is the most watched screen in GA4 and the least useful. Seeing four people on your site right now produces no decision.

Ignore bounce rate and engagement rate until you have a specific question. Both move for reasons that have nothing to do with quality, and both are routinely used in agency reports to fill a page. When a report you receive leads with either one, read it with more suspicion than its design invites.

Ignore the exploration builder unless someone technical is driving. It is powerful and it is where hours disappear.

Why do the numbers look wrong compared to what I expected?

Usually because GA4 is measuring something narrower than you assumed, and because search itself is sending fewer clicks than it used to.

Consent banners, ad blockers, and browser privacy settings all remove sessions from the count. Users on multiple devices get counted more than once unless you have signed in users. And GA4's default attribution assigns credit differently from every ad platform you use, which is why the totals will never reconcile.

Then there is the wider shift. SparkToro, working with Similarweb panel data across US Google searches from January to April 2026, reports 68.01% of searches ending without a click, up from 60.45% in 2024, and calculates that per 1,000 US Google searches only 276 clicks now reach the open web against 374 in 2024, a 26% reduction in two years. Grade that honestly: it is panel-based, Google has publicly disputed SparkToro's methodology, and a separate 2026 measurement using a different definition of zero-click put the figure at 27.6%. Use the direction, not the decimal. The direction is that impressions and clicks have come apart, which is exactly the pattern an owner experiences as "we are more visible and the site is quieter."

None of that makes the tool useless. It makes it a consistent instrument rather than a census. Trends inside one tool are trustworthy. Absolute numbers across two tools are not. Use GA4 to see whether organic traffic is climbing, not to prove exactly how many humans visited.

Also check your data retention setting. The default is short, and it silently caps how far back some reports can look. Set it to the longest option available on the day you open the account.

Can GA4 show me traffic from AI answers?

No, and any report or plugin claiming an "AI traffic" channel in GA4 is constructing it rather than reading it.

The reason is first party and specific. Google states that sites appearing in its AI features are included in overall search traffic in Search Console within the Web search type. Clicks from AI Overviews and AI Mode arrive in your analytics as ordinary Google organic sessions. There is no referral parameter, no UTM, and no separate channel to group by. The largest AI surface in local search is fundamentally unattributable inside GA4.

The official reporting confirms the boundary rather than solving it. Google Search Central announced Search Generative AI performance reports in Search Console in June 2026, giving impressions by page, country, device and date. Clicks, click-through rate, position and the user's prompt are not available, and AI Mode is not separated from AI Overviews.

Two smaller leaks make it worse. Native app clients such as the ChatGPT and Perplexity mobile apps often strip the referrer, so those sessions land in Direct. And referrer-policy stripping on secure connections and inside in-app browsers removes more.

There is one proxy that has been validated, and it lives in Search Console rather than GA4. Ahrefs isolated AI Overview-shaped queries, meaning informational, question-starting and non-branded, and compared click-through rate before and after a known rollout at constant average position. Against confirmed AI Overview keywords the guessed set showed a 42% relative click-through drop and the confirmed set 44%, a difference of only 2% relative. That is vendor research, and it is the most honest instrument currently available. It measures decay on a query shape, not clicks from an AI answer, and it should be described that way.

What to do this week

Confirm the tag is actually installed on every page, including the thank you page and any booking subdomain. Missing pages are the most common cause of numbers that look wrong.

Mark your key events. Open Admin, then Events, and toggle the two or three that represent an outcome. If none of the existing events qualify, that tells you the site has no measurable outcome, which is a website problem rather than an analytics one. Frequently it comes down to what the page is asking people to do, and a page with a weak or missing call to action gives GA4 nothing to record.

Run the reconciliation above once. GA4 key events against connected calls and real bookings, for one month. Write the ratio on the front of the file.

Set the comparison date range to previous period and leave it there.

Then stop. Do not build custom dimensions, audiences, or funnels this month. Come back in ninety days when you have enough data to say something, and by then you should know which handful of things are worth measuring in the first place. If you are hiring help around this, treat the analytics access question the same way you would treat any supplier term, because what you agree to at the start decides what you keep at the end.

Questions about your GA4 setup?

Email me at eric@seod.com with a screenshot of your GA4 Events screen, the one under Admin. I will tell you which events should be marked as key events, which ones are noise, and what is missing. No account access needed, just the screenshot.

I read these personally. If the answer is that your setup is already fine, that is what you will get back.

More on measurement sits in the analytics library.