Most e-commerce teams spend hours reading their traffic, their acquisition sources and their bounce rate. And all the while, a far more direct piece of data sits untouched in their GA4 account: the exact words their visitors type into the search bar.

These queries are not traffic to interpret. They are explicit requests. Someone who types "sleeping bag -10 degrees" or "wedding guest dress" is telling you precisely what they want to buy, in their own words. It is the voice of the customer, with no survey or panel. GA4 knows how to capture it, stores it for free, and makes it usable the moment tracking is on.

~30%
Of e-commerce visitors use the search bar during their session
Multiple studies
$0
The cost of site search tracking in GA4: native, built in, no third-party tool
Built-in feature
2-3x
Higher conversion rate for visitors who use site search
Multiple studies

The reasoning fits in one sentence: a significant share of your best buyers tell you what they are looking for, for free, and that data goes unused in most stores. This guide shows how to wire it up, read it, and turn it into decisions.

What GA4 already knows about your site search

Google Analytics 4 has a mechanism dedicated to site search. When it is active, each search fires a view_search_results event, along with a search_term parameter that holds the text the visitor entered. That parameter is what you want: it stores "sleeping bag -10 degrees" exactly as the person wrote it.

This tracking is part of enhanced measurement, the automatic collection block GA4 offers with no code. By default it invents nothing: it reads the URL of your results page and grabs the contents of the query parameter. That parameter still has to exist and be declared. This is where most setups fall short, and tracking stays empty without anyone noticing.

The first thing to check: run a search on your own site and look at the URL. If it reads like /search?q=shoes, GA4 can capture the query through the q parameter. If the URL does not change at all when you search (common with a widget or instant search), automatic tracking will see nothing: you will need to send the event manually.

Set up tracking in three steps

Count about fifteen minutes. None of these steps needs a developer if your search already changes the URL. The third one handles the special case of searches with no page reload.

1

Turn on enhanced measurement and declare your query parameters

In GA4, open Admin, then your web data stream. Make sure enhanced measurement is on, and that the "Site search" option is on too. GA4 watches five URL parameters by default: q, s, search, query and keyword.

The non-standard parameter trap
If your results page uses a different parameter (for example /results?term=shoes), GA4 does not recognize it out of the box. Add term to the list of search query parameters in the enhanced measurement settings.
The result: as soon as it is on, GA4 starts recording the view_search_results event with the right term. Collection is not retroactive: it begins when you enable the option, so give it a few days before drawing conclusions.
2

Find the data in an exploration

Unlike the old Universal Analytics, GA4 shows no ready-made "site search" report. The most effective route is to build an exploration: the Explore menu, a free-form table, then add the Search term dimension (search_term) as rows and a metric like users or sessions as values.

What you get
A sorted list of every term typed on your site, by volume. In two clicks, you see what your customers search most, in their real vocabulary. Save the exploration so you can reopen it each week without rebuilding it.
To go further: add a secondary dimension like source or device, or cross with a conversion metric to spot which terms actually lead to a purchase. That is where the analysis becomes actionable.
3

Handle a search with no page reload

Many modern engines, including instant search widgets, display results in a layer over the page without changing the URL. GA4's automatic tracking, which relies on the URL, then sees nothing. You have to send the event manually at the moment of the search.

The event to fire
When the visitor runs their search: gtag('event', 'view_search_results', { search_term: 'shoes' }). The same standard event, sent by hand, feeds exactly the same reports as automatic tracking.
The good news: if you use the Vectail dashboard, these queries are already listed and sorted on the engine side, with no GA4 setup. Sending the event manually is only useful if you want to centralize everything in Analytics.

The four readings that change your decisions

Once the data flows in, the point is not to admire a ranking of terms. It is to draw actions from it. Four readings stand out, from the most obvious to the most profitable.

1. The most searched terms. This is real demand, ranked by you without realizing it. If a product or category dominates your searches, it deserves a prime spot in navigation, on the homepage and in your campaigns. Conversely, a heavily searched term pointing to a thin catalog is a signal to buy stock or expand the range.

2. Your customers' real vocabulary. Your visitors almost never use the words on your product pages. They type "fridge", "trainers", "windbreaker" when your catalog says "refrigerator", "sports shoes", "waterproof jacket". This list of terms is a goldmine for rewriting product titles, feeding your synonyms and fuelling your SEO with the expressions people actually use.

3. High-volume terms with low engagement. By crossing the search term with a downstream metric (engaged sessions, product page views, conversions), you spot the frequent queries that lead nowhere. Technically the engine answered; commercially it missed. These are often relevance, title or categorization problems.

4. Emerging terms. Compare two periods. A term climbing from one week to the next signals a trend, a season starting or the effect of a campaign. Caught early, it gives you time to adjust your merchandising before your competitors.

See every query, without building GA4 tracking

Vectail connects to your Google Merchant Center and shows your search terms and zero-result queries directly, engine included. One line of code, powered by Google Vertex AI.

Start free - 14 days, no credit card

GA4's limit: it does not know whether the engine answered

GA4 tells you what people search. It does not tell you, out of the box, whether your engine found anything. A search that returns zero results and a search that returns thirty relevant products fire the same view_search_results event: nothing tells them apart in your default reports.

This is a significant blind spot, because zero-result searches are exactly the ones that drive away your most decided customers. To track them in GA4, you need to instrument a dedicated event (for example, send view_search_results with a parameter for a result count of zero), which takes a technical intervention.

The useful complementarity: GA4 excels at tying search to the rest of the journey (source, device, conversion). An engine like Vectail exposes the list of zero-result queries on its side, sorted by frequency, with no setup. The two answer each other: GA4 for the marketing context, the engine dashboard for the product diagnosis.

From analysis to action

A report you never turn into a decision is worthless. Here is how to convert your GA4 exports into concrete work, from the quickest to the most structural.

  • Export the 20 most searched terms and check, for each one, that the first page of results is relevant on your site. A single poorly served search in the top 20 carries real weight in volume.
  • Spot the vocabulary gaps (the customer says "fridge", your catalog says "refrigerator") and handle them with synonyms rather than rewriting the whole catalog.
  • Fix the product data for terms that find little or poorly: readable titles, unique descriptions, clean attributes and categories.
  • Track the trend over time rather than running a one-off audit. A saved exploration, reviewed each week, beats a thorough analysis forgotten the next month.

Turning on site search tracking in GA4 takes a quarter of an hour and costs nothing. What matters comes after: listening to what your customers ask for in their own words, and fixing what stops them from finding it. It is one of the rare data sources where the signal is already clean, already qualified, and already there.

Back to blog