Is Your GA4 Quietly Lying to You? 6 Tracking Mistakes We Find on Real Websites

GA4 almost never tells you when it is wrong. A broken tag shows no error. A goal that can no longer fire just reads zero, and zero looks like a slow week.

Every example below is from a real website, names left out. Most started with a normal change: a new theme, a plugin, a cookie banner, a moved page.

1. Every Page View Is Counted Twice

What it looks like. One online store sent GA4 every page view twice, about five seconds apart. Pages per session looked great and the bounce rate sat near 10%. Even we read those numbers as real at first.

Why it costs you. GA4 counts a visit with two or more page views as “engaged.” A duplicate makes almost every visit engaged, so bounce rate stops warning you about weak pages and ads.

Check it yourself:

1. Go to tagassistant.google.com, enter your site and load one page.
2. Count the page_view events sent to your GA4 ID (it starts with G-). One is right. Two is a duplicate, usually GA4 installed by a plugin and again in Tag Manager.
3. In GA4, open Reports, then Acquisition, then Traffic acquisition. A site-wide engagement rate above 85% is a warning sign, not a win.

Where Askalytics helps. Askalytics reads your data, not your page code, so it cannot see the extra tag. But ask Ask AI “What was our engagement rate each month this year?” and a rate stuck near 100% stands out.

2. A Goal Tied to a Page That No Longer Exists

What it looks like. A service business built its main lead goal from visits to /thank-you. After a rebuild, that page returns a 404 error, so the goal can never fire. A store we checked had a purchase key event with no events in 18 months.

Why it costs you. Leads keep coming, but GA4 shows none, so every channel looks worthless. If the key event feeds Google Ads, bidding loses its signal too. The missing days can never be filled in.

Check it yourself:

1. In GA4, open Admin, then Events, then Create event. Note any rule built on a page address, such as /thank-you.
2. Paste that address into your browser. It should load the real page.
3. Open Reports, then Engagement, then Key events. Over the last 90 days, look for any key event at or near zero.

Where Askalytics helps. Tracking Health, on the Data Health page, warns when GA4 stops recording key events and flags key events that count page visits instead of actions. AI Sentinel checks nightly for key events that drop to zero or fall sharply. Step 2 is still yours.

3. Ad Clicks That Never Become Visits

What it looks like. One Performance Max campaign recorded 733 clicks in Google Ads. Over the same days, GA4 recorded 75 visits from it. About 9 in 10 paid clicks never showed up.

Why it costs you. Some gap is normal: people decline cookies, block ads or leave early. Up to about 40% missing is normal. Past that, you cannot tell which campaigns work.

Check it yourself:

1. In Google Ads, note one campaign’s clicks for the last 30 days.
2. In GA4, open Traffic acquisition for the same dates, switch the first column to Session campaign, and find that campaign.
3. Under about 6 visits per 10 clicks? Check auto-tagging is on in Google Ads (Admin, then Account settings), and use Tag Assistant to confirm the GA4 tag loads on the landing page.

Where Askalytics helps. The Ads vs GA4 Match-Up on the Paid Search page puts both numbers side by side, says whether the gap is normal, and lists likely causes with a way to confirm each.

4. Ad Spend With Zero Recorded Conversions

What it looks like. Over 90 days, one store spent more than £1,300 on Google Shopping ads and Google Ads recorded 0 conversions. GA4 credited visits from those ads with 71 orders. The ads were selling. Google Ads could not see it.

Why it costs you. Automated bidding learns from conversions, so with none it bids blind, and cutting the “dead” campaign would cut ads that sell. The reverse hurts too: another account counted a conversion on 79% of clicks because a page view was set as one.

Check it yourself:

1. In Google Ads, open Goals, then Conversions, then Summary.
2. Make sure a purchase or lead action is Primary and “Recording conversions.” Page view actions should be Secondary.
3. Compare its count with GA4’s key events for google / cpc over the same dates. They never match exactly, but 0 against 71 is not rounding.

Where Askalytics helps. Tracking Health flags both cases. The Wasted Ad Spend panel on the Paid Search page lists search terms that cost money with no recorded conversion, plus a negative keyword file. If your conversion tracking looks broken, it says so and lists nothing, so you never block a term that sells.

5. A Cookie Banner That Lets Ad Tags Run First

What it looks like. On one site, the Google Ads tag ran before the cookie banner set its defaults. An Ads cookie landed before a UK visitor chose anything, and ad requests continued after they clicked Reject. Another banner had no Reject button at all.

Why it costs you. With UK or EU visitors, ad cookies before consent are a real privacy risk, and European regulators have fined companies over it. Expect some numbers to drop once it is fixed. That drop is the real picture. (This is not legal advice.)

Check it yourself:

1. Open your site in a private window. Do not touch the banner.
2. In Chrome, right-click, choose Inspect, then Application, then Cookies. Look for _gcl_au (Google Ads), _ga (GA4) or _fbp (Meta). If they are there, tags ran first.
3. Click Reject, reload and look again. New ad cookies mean the banner is not in charge.

Where Askalytics helps. Askalytics reads your data, not your site’s code, so this check is for you or your developer. When ad clicks go missing in GA4, though, the Match-Up names your cookie banner as a likely cause.

6. Bot Traffic Hiding in Direct

What it looks like. A small business had a “steady” week: 112 visits against a usual 148. But 63 were Direct visits from one country with zero engaged sessions, and the same traffic padded the usual weeks. Real visitors were about 49 against a usual of about 80, a drop of almost 40%.

Why it costs you. Traffic looks steady while real visitors fall away, and conversion rate looks worse than it is.

Check it yourself:

1. In GA4, open Reports, then Acquisition, then Traffic acquisition.
2. Click the plus sign by the first column header and add Country.
3. In the Direct rows, one country with an engagement rate near 0% is almost certainly not people. GA4 cannot remove it from past data, so add a comparison that leaves it out.

Where Askalytics helps. Every weekly AI Insight looks for a block of traffic from one country and channel with almost no engagement, and says how it changed the picture. AI Sentinel also flags sudden traffic spikes.

Where Askalytics Fits

GA4 stays where your data lives. Askalytics connects to it, plus Search Console and Google Ads, and tells you in plain English what matters. In every AI Insight, code traces the figures back to your data, and a second AI from a different provider reviews it. New this week: an “Also Worth a Look” note for lasting problems like ad clicks that never become visits. Read how we keep the AI honest or see data quality and tracking audits.

The Takeaway

Tracking rarely breaks with a bang. It breaks in a redesign or a plugin update, then feeds you wrong numbers for months. Spend 30 minutes on these six checks, fix what you find, then let something keep checking after you stop.