Core Reference

What is NavBoost? Google's Click-Based Re-Ranking System Explained

NavBoost is Google's primary system for re-ranking search results based on how users click, dwell, and interact with them. Confirmed through leaked internal API documentation and sworn testimony during the DOJ antitrust trial, NavBoost represents the most significant public evidence that click-through rate directly influences Google search rankings.

Last updated: March 2026 ~16 min read

What is NavBoost?

NavBoost is an internal Google system that adjusts the ranking of search results based on aggregated user click behavior. When millions of users search for a query and consistently click on certain results — and stay on those pages — NavBoost interprets that pattern as a signal of relevance and quality. Conversely, when users click a result and quickly return to the search results page (a behavior known as pogo-sticking), NavBoost treats that as a negative signal.

The system does not create rankings from scratch. Instead, it operates as a re-ranking layer that sits on top of Google's other ranking systems. After Google's core algorithms produce an initial set of results based on factors like content relevance, backlinks, and page quality, NavBoost adjusts the ordering of those results based on historical click patterns. A page that consistently earns clicks and long dwell times may be promoted, while a page that users frequently abandon may be demoted.

NavBoost has been operational inside Google for over a decade, but its existence was not publicly confirmed until two major events in 2023 and 2024: the DOJ antitrust trial, during which a senior Google executive described the system under oath, and the 2024 Google API leak, which exposed thousands of pages of internal documentation detailing NavBoost's data structures and click classification categories.

In plain terms: NavBoost watches what happens after users click on search results. If people consistently find what they need on a particular page, that page tends to rank higher. If they consistently bounce back, it tends to rank lower. This feedback loop operates across billions of searches and is one of Google's most important ranking mechanisms.

How NavBoost Was Discovered

For years, Google publicly denied that click-through rate was a ranking factor. In 2016, Google's Gary Illyes stated at a conference that Google had "looked into using clicks for ranking" but found the data "too noisy." As recently as 2023, Google representatives maintained that user click behavior was used only for evaluation and experimentation, not as a direct input to the live ranking system.

Two events shattered that narrative.

The DOJ Antitrust Trial (2023–2024)

In the landmark United States v. Google LLC antitrust case, the Department of Justice argued that Google maintained an illegal monopoly in general search. As part of the proceedings, Google's VP of Search, Pandu Nayak, was called to testify about Google's ranking systems.

Under oath, Nayak confirmed that NavBoost is a system that uses click signals to re-rank search results. He described it as one of Google's most important ranking systems and acknowledged that it relies on user interaction data collected from Google Search and Chrome browser activity. Trial documents revealed that NavBoost had been in operation since at least 2005, making it one of Google's longest-running ranking innovations.

"NavBoost is a system that uses click data to adjust search rankings. It considers various types of clicks and user interactions to determine the relevance and quality of search results."

— Testimony summary from Pandu Nayak, VP of Search, Google. United States v. Google LLC, 2023.

The trial also revealed that Google had been collecting click data from Chrome users — including those who were not logged into a Google account — to feed into NavBoost. This was significant because it showed the scale of data collection powering the system.

The 2024 Google API Leak

In May 2024, a massive trove of internal Google API documentation was made public. The documents were first reported by Rand Fishkin (SparkToro co-founder) and Erfan Azimi (an SEO professional), who received the leaked materials and published analyses of their contents.

The leaked documentation comprised thousands of pages describing the internal data structures, API endpoints, and attribute definitions used by Google's ranking systems. Within this documentation, NavBoost appeared repeatedly as a core ranking module with specific, well-defined fields for tracking user click behavior.

The API leak revealed the specific click categories that NavBoost tracks:

  • goodClicks — Clicks where the user appeared satisfied with the result (clicked and stayed)
  • badClicks — Clicks where the user quickly returned to the search results page
  • lastLongestClicks — The final click in a search session where the user dwelled the longest
  • unsquashedClicks — Raw click counts before normalization
  • squashedClicks — Click counts after being processed through Google's squashing function

Each of these click types is covered in detail on the NavBoost click types page. The presence of both squashed and unsquashed variants confirmed that Google actively normalizes click data to prevent manipulation — a mechanism explored on the squashing function page.

The combination of sworn testimony and leaked internal documentation made NavBoost the most thoroughly documented ranking system in Google's history. For the first time, the SEO industry had both an official confirmation and technical details about how clicks influence rankings.

How NavBoost Works: A Conceptual Overview

NavBoost operates as a feedback loop between user behavior and search result ordering. While the complete technical architecture is not publicly known, the leaked API documentation and trial testimony provide enough detail to construct a reliable model of how the system functions. A more detailed technical breakdown is available on the How NavBoost Works page.

Step 1: Data Collection

Google collects click and interaction data from multiple sources:

  • Google Search interactions — Every click on a search result, the time between clicking and returning to the SERP, and the sequence of clicks within a search session
  • Chrome browser data — Browsing behavior from Chrome users, including time on page, scrolling depth, and navigation patterns after clicking a search result
  • Android device data — Interaction signals from mobile devices running Google's Android operating system

This data collection operates at enormous scale. Google processes billions of searches per day, and each search session generates multiple interaction signals. The sheer volume of data is what gives NavBoost its statistical power — patterns that emerge across millions of queries for the same topic are unlikely to be coincidental.

Step 2: Click Classification

Raw interaction data is classified into the click categories revealed in the API leak. The classification distinguishes between positive signals (user found what they were looking for) and negative signals (user was dissatisfied).

The most important distinction is between goodClicks and badClicks. A click where the user dwells on the page for a significant period and does not return to the search results is classified as a good click. A click where the user quickly bounces back to try another result is classified as a bad click.

The lastLongestClick signal carries particular weight. When a user performs a search, clicks through several results, and finally settles on one page where they spend the most time, that final destination is tagged as the lastLongestClick. This signal is considered one of the strongest indicators of user satisfaction because it represents the result that ultimately answered the user's query.

Step 3: The Squashing Function

Before click data feeds into ranking calculations, it passes through Google's squashing function. This normalization mechanism compresses click volumes so that extremely high click counts do not disproportionately influence rankings.

The squashing function serves two critical purposes:

  1. Anti-manipulation — By compressing click volumes, the function ensures that artificially inflated click numbers (from bots, click farms, or other manipulation techniques) produce diminishing returns. A page receiving 10,000 artificial clicks does not get 10 times the ranking benefit of 1,000 artificial clicks.
  2. Signal normalization — Popular queries naturally generate more clicks than niche queries. Without normalization, pages ranking for high-volume keywords would accumulate click signals far faster than pages ranking for long-tail queries, creating an unfair advantage based purely on search volume rather than relative user satisfaction.

The API leak confirmed the existence of both squashedClicks and unsquashedClicks fields, indicating that Google retains both the raw and normalized versions of click data. The specific mathematical function used for squashing has not been disclosed, but it is likely a logarithmic or sigmoid-style compression that flattens large values while preserving distinctions at lower magnitudes.

Step 4: The 13-Month Rolling Window

NavBoost does not react to yesterday's clicks in isolation. Instead, it aggregates click data over a 13-month rolling window. This means the system considers approximately one year's worth of user interaction data when calculating click-based ranking adjustments.

The 13-month window has several important implications:

  • Resistance to short-term manipulation — A sudden spike in clicks over a few days or weeks is averaged against 13 months of historical data, diluting its impact
  • Seasonal sensitivity — The 13-month window (rather than 12) ensures that seasonal patterns from the same period last year are included, allowing NavBoost to account for annual trends
  • Gradual ranking shifts — Changes in user behavior take time to accumulate enough signal to move rankings, which is why NavBoost-driven ranking changes tend to be gradual rather than sudden
  • New page disadvantage — Pages that are new to the search results have limited click history, which means NavBoost has less data to work with. This can create a "cold start" challenge for new content competing against established pages with months of accumulated click signals

Step 5: Re-Ranking

After click data has been classified, squashed, and aggregated over the 13-month window, NavBoost applies its adjustments to the search results. This is a re-ranking process, not a ranking process. The distinction matters.

Google's core ranking algorithms (including systems for evaluating content quality, relevance, and authority) first produce an initial ordering of results. NavBoost then adjusts this ordering based on click signals. A page that earns a disproportionately high share of good clicks relative to its position may be promoted. A page that generates an unusually high rate of bad clicks may be demoted.

The magnitude of NavBoost's adjustments is not publicly known, but the fact that Pandu Nayak described it as one of Google's "most important" ranking systems suggests that its influence on final result ordering is substantial.

The Five Click Types NavBoost Tracks

The 2024 API leak revealed five distinct click categories within NavBoost's data model. Each captures a different dimension of user interaction. These are covered comprehensively on the NavBoost click types page, but here is an introductory overview.

Click Type What It Measures Signal Interpretation
goodClicks Clicks where the user stayed on the result page Positive — user likely found what they needed
badClicks Clicks where the user quickly returned to the SERP Negative — user was dissatisfied with the result
lastLongestClicks The final click in a session with the longest dwell time Strong positive — the result that ultimately satisfied the query
unsquashedClicks Raw click counts before normalization Baseline volume data
squashedClicks Click counts after the squashing function is applied Normalized volume used in ranking calculations

The interplay between these signals creates a nuanced picture of user satisfaction. A page might receive many total clicks (high volume) but also generate a high rate of badClicks (low satisfaction). NavBoost would interpret this as a poor result despite its high click count. Conversely, a page that receives fewer total clicks but consistently earns lastLongestClick status would be interpreted as a highly satisfying result.

This multi-signal approach makes NavBoost considerably more sophisticated than a simple "more clicks equals higher ranking" system. It attempts to measure the quality of clicks, not just the quantity.

NavBoost and CTR as a Ranking Factor

The existence of NavBoost settles a long-running debate in the SEO industry: does click-through rate affect Google rankings? The evidence now overwhelmingly indicates that it does, though not in the simplistic way many assumed.

What NavBoost Confirms

NavBoost confirms that user click behavior is a direct input to Google's ranking system. This is not an evaluation metric used only in A/B tests. It is not a secondary signal used only in edge cases. According to testimony from one of Google's most senior search executives, NavBoost is one of the most important ranking systems at Google.

The specific click fields revealed in the API leak — goodClicks, badClicks, lastLongestClicks — demonstrate that Google tracks not just whether a user clicks, but what happens after the click. This post-click behavior data is what transforms raw CTR into a meaningful quality signal.

The Nuance: CTR Alone Is Not the Signal

It would be inaccurate to say that "CTR is a ranking factor" in the traditional sense. NavBoost does not simply measure whether a result was clicked. It measures a constellation of user behaviors:

  • Whether the user clicked (initial engagement)
  • How long they stayed (dwell time / satisfaction)
  • Whether they returned to try other results (pogo-sticking)
  • Whether the page was the last and longest destination in the session (terminal satisfaction)

A more accurate framing is that user satisfaction signals, as measured through click behavior, are a ranking factor. Raw CTR is one component, but it is enriched by dwell time, return-to-SERP behavior, and session-level analysis.

Supporting Evidence Beyond NavBoost

NavBoost is not the only evidence that clicks affect rankings. Multiple CTR studies have shown that organic click-through rates follow predictable patterns, and pages that deviate significantly from expected CTR for their position tend to change rank over time. Additionally, Google holds several patents related to click-based ranking adjustments, which are documented on the NavBoost patents page.

The complete timeline of public evidence shows that click signals have been part of Google's ranking system for nearly two decades, with NavBoost being the specific implementation that has operated since at least 2005.

Why NavBoost Matters for SEO

The confirmation of NavBoost has practical implications for anyone who depends on organic search traffic. Understanding that click behavior influences rankings changes how SEO strategy should be approached.

Title Tags and Meta Descriptions Carry More Weight Than Previously Assumed

If click behavior influences rankings, then the elements that drive clicks from the SERP — primarily the title tag and meta description — are more important than many SEO practitioners treated them. A compelling title that earns a higher-than-expected CTR may, over time, contribute to improved rankings through the NavBoost feedback loop.

This does not mean clickbait works. NavBoost tracks post-click behavior. A misleading title that earns clicks but generates high badClick rates would likely produce a net negative signal. The title must accurately represent the content while being compelling enough to earn the click.

User Experience Is a Ranking Factor (Through NavBoost)

When users click a search result and stay on the page, that generates a goodClick. When they bounce back immediately, that generates a badClick. This means every aspect of user experience that affects whether someone stays or leaves — page load speed, content quality, layout, readability, ad intrusiveness — indirectly feeds into NavBoost.

Sites that deliver a poor user experience are not just losing visitors. They are generating negative NavBoost signals that can erode their rankings over time.

Content That Satisfies Search Intent Gets Rewarded

The lastLongestClick signal is particularly important. It rewards the page that ultimately answers the user's question. This means content that thoroughly addresses the search intent behind a query — leaving the user no reason to return to the SERP — accumulates the strongest positive NavBoost signal.

Thin content, partial answers, and pages that force users to continue searching are penalized by this mechanism. Comprehensive, authoritative content that fully satisfies the query is rewarded.

The 13-Month Window Creates Momentum

Because NavBoost aggregates data over 13 months, established pages with long histories of positive click signals have a built-in advantage. This creates a form of ranking momentum: pages that have been satisfying users for months are difficult to displace, even if a newer page might be marginally better.

For new content competing against established pages, this means patience is required. Building a positive NavBoost signal takes time, and the effects of improved user engagement accumulate gradually over the 13-month window.

CTR Optimization Becomes a Legitimate SEO Discipline

Before NavBoost was confirmed, CTR optimization was considered secondary to traditional SEO activities like link building and on-page optimization. With NavBoost publicly documented, CTR optimization — improving the rate at which users click on a listing in the SERP, and ensuring they stay once they arrive — is now a recognized component of comprehensive SEO strategy.

What NavBoost Does Not Do

The confirmation of NavBoost has generated considerable attention, and with that attention has come some overstatement of its role. It is important to understand what NavBoost does not do.

NavBoost Does Not Replace Other Ranking Factors

NavBoost is a re-ranking system. It adjusts the order of results that have already qualified for ranking through Google's other systems. A page with zero backlinks, thin content, and no topical relevance will not rank simply because it receives clicks. The page must first meet Google's baseline quality and relevance thresholds before NavBoost has any opportunity to influence its position.

Content quality, backlink authority, topical relevance, page experience signals, and dozens of other factors still determine whether a page appears in the search results at all. NavBoost operates on the margins — adjusting the ordering of results that have already qualified.

NavBoost Does Not React Instantly

The 13-month rolling window means NavBoost is inherently slow-moving. A sudden change in click patterns — whether from a viral event, a site redesign, or a manipulation attempt — is averaged against months of historical data. NavBoost-driven ranking changes unfold over weeks and months, not hours or days.

This is by design. The extended averaging window makes the system more robust and resistant to noise, but it also means NavBoost cannot respond to rapidly changing user preferences or newly published content.

NavBoost Does Not Work the Same for All Queries

High-volume queries generate large amounts of click data, giving NavBoost strong statistical signals to work with. Low-volume queries, long-tail keywords, and brand-new queries may not generate enough click data for NavBoost to produce meaningful adjustments.

For niche topics with limited search volume, other ranking factors likely carry proportionally more weight than NavBoost. The system's influence scales with data availability.

NavBoost Does Not Penalize Pages Directly

NavBoost adjusts rankings based on relative click performance. A page that generates more badClicks than expected may be demoted, but this is a ranking adjustment, not a penalty. There is no evidence that NavBoost triggers manual actions or algorithmic penalties. It simply shifts the position of results within the normal ranking order based on click signals.

NavBoost Is Not Easily Gamed

The squashing function, the 13-month averaging window, and Google's click manipulation detection systems work together to make NavBoost resistant to artificial manipulation. While no system is perfectly immune to gaming, Google has invested heavily in ensuring that NavBoost's click signals reflect genuine user behavior rather than manufactured interactions.

The History of Click Signals at Google

NavBoost did not emerge in a vacuum. Google has experimented with and deployed click-based signals for over two decades. The complete NavBoost timeline covers this history in detail, but a summary of key milestones is useful context.

Year Event Significance
2005 NavBoost system reportedly begins operation Earliest known deployment of click-based re-ranking at Google
2009 Google patents on "modifying search result ranking based on implicit user feedback" filed Establishes the intellectual property framework for click-based ranking
2015 Google confirms RankBrain First public acknowledgment that ML-based signals influence rankings, though this is query interpretation rather than click ranking
2016 Gary Illyes states clicks are "too noisy" for ranking Public denial of click-based ranking, contradicted by later evidence
2023 Pandu Nayak testifies about NavBoost during DOJ trial First official confirmation that click data feeds directly into ranking
2024 Google API documentation leak Technical details of NavBoost's click types, squashing function, and data structures made public

The gap between 2016 (public denial) and 2023 (sworn confirmation) is significant. For seven years, Google maintained publicly that clicks were not a ranking factor while internally operating NavBoost as "one of the most important" ranking systems. This discrepancy has understandably eroded trust in Google's public statements about its ranking methodology.

The Squashing Function and Manipulation Resistance

One of the most significant revelations from the API leak was the existence of Google's squashing function, a normalization mechanism designed to prevent click manipulation from distorting NavBoost signals.

How Squashing Works

The squashing function compresses click counts using a mathematical transformation (likely logarithmic) so that large click volumes produce diminishing marginal impact. In practical terms, doubling the number of clicks on a result does not double its NavBoost signal. The first hundred clicks may produce a substantial signal, but going from 10,000 to 20,000 clicks produces a much smaller incremental change.

This compression makes manipulation expensive and inefficient. An attacker would need to generate exponentially more fake clicks to achieve linear gains in ranking signal, while simultaneously avoiding Google's click manipulation detection systems.

Additional Anti-Manipulation Measures

Beyond the squashing function, Google employs multiple layers of protection against artificial click manipulation:

  • Behavioral analysis — Bot clicks tend to follow mechanical patterns (consistent timing, uniform dwell times, predictable navigation). Human clicks are inherently variable. Google's systems look for statistical signatures of artificial behavior.
  • Geographic consistency — Clicks from geographic regions that do not match the query's expected audience can be flagged as suspicious.
  • Device and browser fingerprinting — Clusters of clicks from identical or similar device configurations suggest automation.
  • Chrome integration — Data from authenticated Chrome users provides a high-trust baseline against which anonymous or suspicious click patterns can be compared.
  • The 13-month window — Short-duration manipulation campaigns are diluted by months of legitimate historical data.

No anti-manipulation system is perfect, and practitioners have reported varying degrees of success with click manipulation campaigns. However, the combination of squashing, behavioral analysis, and the 13-month averaging window creates significant barriers to sustained manipulation.

Frequently Asked Questions

Is NavBoost the same as RankBrain?

No. RankBrain is a machine learning system that helps Google interpret the meaning of search queries, particularly ambiguous or novel ones. NavBoost is a separate system that re-ranks search results based on aggregated user click behavior. They operate at different stages of the ranking pipeline: RankBrain helps Google understand what a user is looking for, while NavBoost adjusts result ordering based on how users have historically interacted with those results. A more detailed comparison is available on the NavBoost vs. RankBrain page.

Did Google admit that clicks affect rankings?

Yes. During the 2023–2024 DOJ antitrust trial, Google VP of Search Pandu Nayak confirmed under oath that NavBoost uses click data to re-rank search results. This was further corroborated by the 2024 Google API leak, which revealed specific click-related fields including goodClicks, badClicks, and lastLongestClicks within Google's internal ranking systems. The full evidence is documented on the CTR as a ranking factor page.

How long does NavBoost store click data?

NavBoost operates on a 13-month rolling window. Click behavior data is aggregated over approximately 13 months, meaning that user interaction patterns from the past year inform current rankings. This extended window helps smooth out short-term fluctuations and makes the system more resistant to manipulation attempts.

Can you manipulate NavBoost with fake clicks?

Google employs a squashing function specifically designed to normalize click data and reduce the impact of manipulation attempts. The system compresses click signals so that artificially inflated volumes carry diminishing returns. Combined with behavioral analysis, geographic consistency checks, device fingerprinting, and the 13-month averaging window, Google has multiple layers of defense against click manipulation.

Does NavBoost affect all search queries?

NavBoost primarily affects queries with sufficient click data to produce meaningful signals. High-volume queries generate large amounts of click data and are most influenced by NavBoost. Low-volume or brand-new queries may have insufficient click history for NavBoost to meaningfully adjust rankings, in which case other ranking systems carry more weight.

What is the difference between squashed and unsquashed clicks in NavBoost?

Unsquashed clicks represent the raw, unprocessed click data collected from user interactions. Squashed clicks have been passed through Google's squashing function, a normalization algorithm that compresses click counts to prevent any single large signal from disproportionately influencing rankings. The squashing function reduces the impact of click volume spikes — whether from genuine viral interest or artificial manipulation — ensuring that click signals remain proportional and resistant to gaming.

Further Reading

This page provides a foundational overview of NavBoost. For deeper exploration of specific topics, the following pages offer detailed analysis:

About this site: NavBoost.com is an independent resource on Google's click-based ranking systems. For businesses looking to improve their organic click-through rates, we recommend SerpClix — the only crowd-sourced CTR service using real human clickers.