What GA4 actually is
Google Analytics 4 is the current generation of Google's web analytics product. It replaced Universal Analytics in July 2023 and is the default analytics platform for nearly every Australian SMB website. It is free for standard use, event-based rather than session-based at its core, and the on-site half of the SEO measurement stack.
The shift from Universal Analytics to GA4 caught a lot of teams out. The two products look superficially similar but the data model is fundamentally different. GA4 treats everything as an event (page views, scrolls, clicks, conversions). Sessions and users are derived from events rather than being primary. The result is more flexibility once you understand it, and more confusion when you do not. Most of the "GA4 is broken" complaints we see are actually "the property was set up assuming the old data model still applied".
What GA4 covers. On-site behaviour (sessions, engagement, events, conversions). Traffic sources (the channel that brought the user to the site). User journeys (path analysis across pages and events). Revenue and e-commerce (for sites with purchase tracking). Audiences (segments of users for downstream analysis).
What GA4 does not cover. SERP-side data; that is what GSC does. Keyword-level position; that is what a rank tracker does. Technical health; that is GSC plus a crawler. The four tools are complementary. See the SEO Measurement pillar for the full stack and the GSC chapter for the SERP-side companion.
Setting it up properly
The default GA4 install is not configured for SEO measurement out of the box. Six setup moves change that.
Step 1: Install via Google Tag Manager
Use Google Tag Manager (GTM) to load GA4 rather than hard-coding the GA4 tag in the site template. GTM gives you control over event firing, lets you add custom events without code deploys, and is the standard approach for any team that runs both SEO and CRO. The GTM container loads once; everything else (GA4, conversion pixels, Microsoft Clarity, A/B test scripts) loads through it.
Step 2: Change data retention to 14 months
In the GA4 Admin, under Data Settings then Data Retention, change the event data retention from the 2-month default to 14 months. The 2-month default makes year-on-year analysis impossible. The 14-month setting is the maximum for standard properties. Do this on day one of every new property before any data builds up.
Step 3: Enable enhanced measurement
In the Data Streams section, enable Enhanced Measurement on the web stream. This adds automatic event tracking for scrolls, outbound clicks, site search, video engagement and file downloads. Useful, free, and most clients have it disabled because the previous agency turned it off and forgot why.
Step 4: Link to Search Console and Google Ads
Under Product Links in the GA4 Admin, link both your Search Console property and your Google Ads account. The GSC link enables the Search Console reports section inside GA4, which combines GSC query data with GA4 session and conversion data. The Google Ads link enables conversion-modelling features and lets you use GA4 audiences in Ads.
Step 5: Configure the BigQuery export
Under BigQuery Links, configure the free daily export. GA4 will copy your event data to a BigQuery dataset every day. The first 10 GB of storage per month is free in BigQuery; nearly every SMB stays within that. The export protects your data history (you keep the raw events even if the GA4 interface limits you) and lets you run SQL queries when the GA4 interface runs out of analytical depth. Set it up the same day you change data retention.
Step 6: Add Google signals carefully
Google Signals lets GA4 cross-reference signed-in Google users for demographic data and cross-device attribution. It adds value but it also triggers data thresholding on small audiences (GA4 will hide some data to protect privacy when the audience is small). For most Australian SMBs the thresholding hides more than the demographics reveal. We turn Google Signals off on small properties and on for sites with substantial traffic where the cross-device attribution is more useful than the thresholding is painful.
Defining conversions that matter
The single most common GA4 problem we see on takeover. Either no conversions are defined, or conversions are defined badly. Both produce a measurement stack that cannot answer the question that matters.
The honest definition of an SEO conversion is an event that represents a real business outcome. For most Australian service businesses, that means a form submission, a phone click, or a booking request. For e-commerce, it means a purchase event. For content publishers monetised by ads, it means a meaningful engagement event (newsletter signup, account creation). The conversion event must be something that, if the count doubled, the business would feel it in revenue.
Events to mark as Key Events (conversions)
- Form submission. The lead form completion event from your form plugin (Gravity Forms, Contact Form 7, HubSpot, Webflow form). Fire once per submission, not on field interactions. Tag with the form name in the event parameters.
- Phone click. A click on a
tel:link. Cheap to set up via GTM trigger. Pairs naturally with the form submission event for service businesses where 30 to 50 percent of leads come by phone. - Email click. A click on a
mailto:link. Smaller volume than phone or form, still worth tracking. - Purchase. For e-commerce, the standard GA4 purchase event with transaction value. Drives the revenue reporting.
- Booking or appointment scheduled. For services with online booking (Calendly, Acuity, custom). The completed-booking event, not the started-booking event.
Events to never mark as Key Events
- page_view. A page view is a step, not an outcome. Never a conversion.
- session_start. Same logic. The user arriving on the site is not a business outcome.
- scroll. Scrolling 90 percent down a page is not a lead. It is a diagnostic engagement signal.
- file_download. Sometimes worth flagging as a soft conversion for content sites with gated downloads. For most sites it is engagement data, not a conversion.
The rule of thumb. A conversion is something a CEO would feel if it stopped happening. If the CEO would not care, it is engagement data, not a conversion. Pick one or two real conversions and report against them relentlessly. See SEO KPIs for how the conversion events flow into the wider KPI structure and conversion rate for the underlying definition.
Custom channel groupings
GA4's default channel grouping has 17 channels. Organic Search, Direct, Referral, Paid Search, Display, Email and so on. For most SEO measurement work the defaults are fine. Two changes are worth making.
Change 1: Add an AI Referrers channel
By default, AI search referrers (ChatGPT, Perplexity, Gemini, Copilot) land in the Referral channel mixed with everything else. The AI traffic is small but high-intent; lumping it in with random referral traffic hides it. Build a custom channel grouping that adds an AI Referrers channel matching the relevant hostnames.
The hostnames worth catching as of mid-2026: chat.openai.com, chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, bing.com/chat. The list will grow; revisit it quarterly. See the tracking AI referrals chapter for the exact match conditions.
Change 2: Split Organic Search by branded and non-branded
Harder to do natively in GA4 than in GSC. The cleanest approach is to import GSC query data and use the BigQuery export to run branded versus non-branded splits in SQL. The fallback is to use the GA4 landing page report and segment by which pages serve branded queries versus non-branded queries, which is approximate but usable. See the GSC chapter for the branded regex pattern that feeds this work.
The four reports that matter
GA4's interface has a lot of reports. Four of them do most of the SEO measurement work.
Report 1: Traffic acquisition
Found under Reports then Acquisition then Traffic acquisition. Shows sessions, engaged sessions, engagement rate, average engagement time, key events and conversions by Session default channel grouping. This is the headline channel report and the one most stakeholder reports key off.
For SEO measurement, filter to Organic Search (or use a custom channel that includes AI Referrers if you want to count them together). Compare 28-day rolling versus prior 28-day rolling. Watch sessions, engaged sessions and key events. If sessions grow but key events do not, the new traffic is wrong-intent or the conversion rate has dropped.
Report 2: Pages and screens
Under Reports then Engagement then Pages and screens. Shows views, users, key events and conversions by page. The landing-page-level performance view.
The most useful filter for SEO work is to add a secondary dimension of "Session default channel grouping" and filter to Organic Search. That surfaces the pages bringing in organic traffic, ordered by views. Cross-reference with the GSC Pages tab for the SERP-side picture of those same pages.
Report 3: Search Console
Appears in the Reports section once you link GSC to GA4. Combines GSC query and landing-page data with GA4 session and conversion data. The Landing pages report inside this section is especially useful: it shows GSC clicks, impressions and average position alongside GA4 sessions, engaged sessions and key events for the same landing page. Single view, both halves of the funnel.
Report 4: Monetization (e-commerce only)
Under Reports then Monetization. The revenue and purchase-event reporting. Skip this if you do not have e-commerce tracking; lean on it heavily if you do.
For e-commerce SEO, the key views are Revenue by Session default channel grouping (to see the revenue contribution of organic) and Purchase by landing page (to see which landing pages produce purchases). Combine with the Search Console section above for a full view of which keywords are driving revenue.
Attribution and how to read it
GA4 defaults to data-driven attribution, which uses machine learning to distribute conversion credit across the channels the user touched on the way to converting. Theoretically more accurate than last-click. Practically harder to explain to a board.
Three things worth knowing about GA4 attribution for SEO.
- Data-driven is the default and the right default for internal analysis. It reflects multi-touch reality better than last-click. SEO usually gets more credit under data-driven than under last-click because SEO is often the first or middle touch, not the last.
- Last-click is better for stakeholder reporting. Because it is easier to explain. "The session that ended in a conversion came from organic" is a sentence a CEO can understand. The data-driven equivalent is harder to defend in a meeting.
- You can switch the model and compare. In the Admin under Attribution Settings, you can change the reporting attribution model for the whole property, or use the Attribution section under Advertising to compare models side-by-side. We typically show stakeholders the last-click view and use the data-driven view for internal analysis.
The bigger attribution honesty conversation lives in the attribution chapter. The short version: SEO attribution is imprecise because of dark traffic, AI referrers, and the multi-channel reality of how people actually buy. Use the data, accept the imprecision, and frame the SEO contribution as a band rather than a point estimate.
Common mistakes
- Setting data retention to 14 months on day one of every new property.
- Defining one or two real conversion events and treating them as the scorecard.
- Building a custom channel grouping that surfaces AI referrers separately.
- Linking GA4 to GSC and Google Ads to enable cross-tool reporting.
- Setting up the BigQuery export to protect data history and enable SQL analysis.
- Marking page_view or session_start as a conversion.
- Leaving data retention on the 2-month default and losing year-on-year history.
- Letting AI referrers sit invisibly in the Referral channel.
- Reporting GA4 numbers next to GSC numbers and treating mismatches as broken data.
- Switching attribution models mid-quarter and breaking the trend continuity.
Perth and WA context
Two GA4 patterns specific to Perth and WA businesses.
Phone clicks are often the biggest conversion channel. Australian service businesses, especially trades and healthcare, take a large share of leads by phone. The phone-click event on the tel: link is the closest GA4 proxy. Set it up in GTM, mark it as a Key Event, and report it next to form submissions. Without it the GA4 conversion picture under-counts real lead volume by 30 to 60 percent for most service businesses. See trades SEO and healthcare SEO for the industry context.
Local search intent maps poorly to GA4's default location data. GA4 shows the user's IP-derived location, which is usually accurate to city level but not to suburb. For Perth businesses with multiple service-area pages (Fremantle, Joondalup, Mandurah), the landing-page-level analysis matters more than the geo report. Cross-reference Pages and screens with the suburb-specific landing pages. See Local SEO and Local SEO Perth.
For the wider context, the GSC chapter covers the SERP-side companion to GA4, the attribution chapter goes deeper on GA4 attribution models, the reporting chapter covers how to surface GA4 data in a CEO-readable format, the search intent chapter covers how query intent feeds the analysis, and the tracking AI referrals chapter covers the AI-referrer build inside GA4. Sites starting from zero can begin with a free SEO audit; clients ready for a measurement-stack rebuild engage the website audit service or the full SEO service.