Google NavBoost Explained: How Click Behaviour Affects Rankings and How to Track User Engagement

It has been a matter of hot debate for a number of years as to whether Google uses clicks and user engagement to influence organic rankings.

Google’s public advice has generally encouraged website owners to focus on creating useful content rather than chasing behavioural metrics such as bounce rate or dwell time. In 2019, Gary Illyes, a Google Search analyst, rejected claims that RankBrain used metrics such as dwell time and click-through rate, describing the idea as “made up crap” and commenting that search was “much more simple than people think.”

However, evidence disclosed through Google’s antitrust trial, followed by the 2024 leak of internal Google Search API documentation, provided a much clearer picture. The documents referenced systems and attributes associated with how users interact with search results, including a system called NavBoost.

What Is NavBoost?

NavBoost is a system that uses aggregated user interaction data to help evaluate and rerank search results.

At a simplified level, Google can examine how people respond when a page appears for a particular query. It records details such as:

• Was the result clicked?
• Was it repeatedly ignored?
• Did the user return quickly and choose another result?
• Was the click the last and longest interaction within a related search journey?
• Did behaviour differ between mobile and desktop users?
• Did users in different locations respond differently?

Within NavBoost itself, there are a number of specific click-behaviour signals, including:

• clicks — total recorded clicks
• goodClicks — clicks Google classified internally as positive or successful, although the exact criteria are not publicly confirmed
• badClicks — clicks Google classified internally as negative or unsuccessful, although the exact criteria are not publicly confirmed
• lastLongestClicks — clicks that were both the last and longest within related search journeys
• unsquashedClicks — click counts before Google applies “squashing” or normalisation to prevent unusually large volumes from dominating
• impressions — how often the result appeared; not technically a click, but needed to interpret clicks and calculate click-through behaviour
• unicornClicks — a click-related field whose precise meaning and use are not publicly confirmed
• lastGoodClick — a field associated with the recency or date of the most recent good click in some document-level fields

As you can see, this disclosure appears to contradict some of what Google has said publicly on the matter.

However, this does not mean that every click or user behaviour directly moves a page up or down the rankings. Rather, aggregated search behaviour may be used to assess whether Google’s results are satisfying users over time.

How Does the Existence of NavBoost Impact SEO?

For SEO teams, the implication is significant: earning a ranking is only part of the job.

A page must also attract the right click and fulfil the expectation and intent created by the search result.

This expectation varies between different types of search results. For example, pages that answer straightforward questions such as “What time is it in Paris?” or “What is the temperature in New York now?” should not necessarily be viewed as unsuccessful because they generate short sessions or limited scrolling. The user may have found the answer immediately.

However, a fashion e-commerce brand with multiple collection and product pages that is experiencing short user sessions, low scroll depth and a high proportion of users returning to the SERP may be showing signs that its pages are not fully satisfying search intent.

This does not mean those on-site metrics directly cause a ranking demotion. However, they can help identify problems that may also lead to weaker search-result engagement over time.

It is also worth noting that there is no publicly available information about the weighting of user satisfaction signals within Google’s algorithm.

Moreover, there is nothing to suggest that UX or user satisfaction is more important than traditional ranking factors such as site authority, content quality, technical SEO, site speed and E-E-A-T.

That does not mean you should ignore UX. On the contrary, it should form part of your ongoing optimisation efforts.

What Metrics Will Help Measure User Engagement and User Experience?

There are a number of ways to measure user engagement. Below are some of the metrics you may want to track, depending on the goal of each page on your site.

• Scroll depth
• Bounce rate
• Average engagement time per session
• Outbound link clicks
• Site-search usage
• Events per session
• Views per session
• Rage clicks, where custom tracking or a behavioural analytics platform is installed
• Form interactions
• Add-to-cart and remove-from-cart activity
• Checkout progression
• Element visibility and impressions, where custom event tracking is configured
• Page exits or a calculated exit rate

Some of these metrics are available in GA4 by default or through Enhanced Measurement, while others require e-commerce implementation, Google Tag Manager, custom events or a separate behavioural analytics platform.

For example, GA4’s standard scroll event is triggered when a user reaches approximately 90% of a page. Tracking additional thresholds such as 25%, 50% and 75% requires custom configuration.

These metrics are not confirmed Google ranking signals. Instead, they are internal diagnostic measures that can help marketers understand whether a page is satisfying users and where the experience may need improvement.

SEO teams should define what successful engagement looks like for each page type rather than applying one site-wide benchmark.

For example, an e-commerce collection page with a low add-to-cart rate, poor scroll depth or a high exit rate may indicate that users are not satisfied with the page.

This could be due to the product range, sizing, availability, pricing or something else altogether.

By recording the metrics outlined above, marketers should be able to identify potential reasons why a particular page may not be performing well in terms of user interaction.

However, as already stated, you cannot apply one hard-and-fast rule to every page type.

For example, if a page says, “Get the best deals by calling our sales team directly,” that may reduce on-page engagement while still generating strong sales results.

How Do I Extract the Data From GA4?

One way to extract and analyse the data is to connect GA4 to Claude using Windsor.ai. The setup does not require advanced programming knowledge.

First, create a Windsor.ai account and connect the Google account that has access to the relevant GA4 property. New accounts currently receive a 30-day free trial, after which a limited free plan is available.

Once GA4 is connected, Windsor.ai generates a live data-query API endpoint for your account — a URL containing your unique API key that returns your GA4 data on request. Rather than installing it as a formal “connector” inside Claude, you simply give Claude that endpoint (specifying the date range and fields you want, such as sessions, bounce rate or landing page), and Claude queries it directly like any other web data source to pull back the underlying GA4 data.

You can then ask Claude to retrieve specific engagement metrics, such as:

• engagement rate and bounce rate
• average engagement time
• views and sessions by landing page
• events and key events
• internal site-search activity
• scroll events
• form interactions
• e-commerce actions, where tracking is configured

It is important to remember that Claude can only retrieve data that has already been collected in GA4.

Metrics such as detailed scroll thresholds, rage clicks, element visibility and checkout stages will only be available if the relevant events or custom tracking have been configured.