For a decade Google’s public line was that it does not use clicks to rank pages, and for a decade SEOs argued in circles about it with no evidence either way. That debate is over. Navboost — a real, named Google system that re-ranks results using aggregated click behavior — was described under oath in the 2023–2024 antitrust trial and corroborated by the Content Warehouse API leak. The question is no longer whether clicks feed rankings. It is how, on what horizon, and whether you can do anything about it without getting your patterns thrown out as manipulation.
What Navboost Actually Is
Navboost is not the ranking algorithm. It is a re-ranking layer that sits on top of Google’s core retrieval and adjusts the order of results based on how searchers historically interacted with them for a given query. Google’s own VP of Search, Pandu Nayak, confirmed its existence and function in testimony. Think of the base system as producing a candidate ordering from relevance and links, and this layer as a second pass that promotes results people demonstrably chose and engaged with, and demotes the ones they bounced off.
The distinction matters because it tells you where clicks fit in the stack. Clicks do not conjure a page onto page one from nowhere. Navboost re-orders results that already qualified. If your page is not in the consideration set — not indexed well, not relevant, not linked — no amount of click behavior rescues it. Once you are in contention, though, engagement can be the tiebreaker that moves you three positions.
The Signals: Good Clicks, Bad Clicks, and the Last Longest Click
The API leak exposed the vocabulary Google uses internally, and it is more granular than “did someone click.” The named metrics include goodClicks, badClicks, and lastLongestClicks — a taxonomy that maps neatly onto searcher satisfaction rather than raw click-through rate.
- Good clicks are the ones that appear to satisfy the query — the user clicks, stays, and does not immediately return to the results to try something else.
- Bad clicks are the opposite: a click followed by a fast bounce back to the search page, the classic “pogo-sticking” pattern that says the page did not deliver.
- Last longest click is the most telling. In a session where someone clicks several results, the one they land on last and dwell on longest is treated as the result that finally answered them — a strong satisfaction signal.
This is why chasing raw click-through rate as a ranking lever misreads the system. A clickbait title that wins the click and then fails the visit generates bad clicks and short dwell, which is worse than not being clicked at all. The system is measuring resolution, not attention.
The 13-Month Window: Why Click Signals Have Memory
The system aggregates click data over a rolling window — roughly 13 months, according to the trial, narrowed from an earlier 18-month span around 2017. That single fact reshapes how you should think about the signal. It is not real-time. A day of good engagement does not move you, and a day of bad engagement does not sink you. The system is looking at a long, smoothed history of how searchers treated your result for a query.
Two practical consequences follow. First, incumbency is sticky — a page that has accumulated 13 months of good clicks carries momentum that a fresh competitor cannot instantly overturn. Second, recovery is slow: if a page earned bad-click history, cleaning it up takes months to re-average, not days. This rolling memory is also why click-manipulation schemes decay — a burst of fake clicks is a blip against a year of real behavior, and Google’s systems are built to discount exactly that shape.
Glue: The Same Idea Applied to the Whole Page
The system has a sibling. The trial and leak referenced a companion nicknamed Glue, which extends the same behavioral logic beyond the ten blue links to every feature on the results page — images, the knowledge panel, “people also ask,” featured snippets, local packs. Glue aggregates interactions across all these elements to decide what earns real estate and in what order. If searchers consistently interact with a video block over the web results for a query, that behavior informs how the page gets assembled.
The takeaway for practitioners: your competition for a query is not only the other web results. It is every SERP feature that can satisfy the intent, and behavioral data decides which features win space. Optimizing for a blue-link position while a video carousel eats the clicks is fighting the wrong battle.
How Google Stops You From Faking It
The obvious question is whether you can just generate the clicks. Google clearly anticipated this, and the leak surfaced the defenses. There is a squashing function — a mathematical dampener that caps how much any single signal can contribute, so no metric can be inflated to infinity. There are separate treatments for signals coming from different sources and geographies. And the 13-month window itself is a manipulation filter: sustained fake behavior at the scale needed to move a competitive query is expensive, detectable, and still gets averaged against real history.
Practitioners have tested click-injection schemes for years, and the honest verdict is that they are unreliable, short-lived, and risky. The systems that would need to be fooled — device diversity, session coherence, dwell realism, geographic distribution — are precisely what Google models. Treat click manipulation the way you would treat a private blog network: it can produce a twitch on a low-competition term and nothing durable on one that matters.
What This Means for Real Optimization
The correct response to Navboost is not to game clicks. It is to earn the behavior the system rewards, which turns out to be ordinary good SEO with sharper priorities:
- Win the click honestly. A title and meta description that accurately promise what the page delivers earn clicks that turn into good clicks, not bad ones.
- Satisfy on arrival. The answer to the query should be visible above the fold. Dwell and the “last longest click” reward pages that resolve intent fast, not pages that bury it under an intro.
- Match intent exactly. Pogo-sticking is almost always an intent mismatch — the searcher wanted a comparison and got a sales page. Aligning format to intent is the single biggest bad-click reducer.
- Cover the whole question. If the visit ends with the user going back to search for the sub-question you skipped, that is a bad signal. Completeness keeps them on the page.
How to Diagnose a Navboost Problem
You cannot see Navboost directly, but you can read its fingerprints. A page that ranks well on relevance metrics yet underperforms its expected position, or one that slid without any obvious content or link change, is a candidate for a click-signal problem — searchers are choosing and preferring competitors. Cross-reference three things: your click-through rate from Search Console (is the click being won at all), your engagement and bounce behavior from analytics (is the visit satisfying), and the actual SERP (has a feature or a sharper competitor changed what “good” looks like for this query).
This diagnostic is exactly the kind of pattern-reading that SEO Rocket is built to support. Its rank tracking uses top-100 snapshots so you see the trend line rather than a single noisy day — essential when the underlying signal has a 13-month memory and moves slowly. Pairing that with its real-crawler site audit tells you whether a slide is technical or behavioral before you waste a month fixing the wrong thing.
Where Content and Competitor Analysis Fit
Because the system rewards the result that satisfies the query, the highest-leverage work is building pages that genuinely beat the current winner on that specific intent — not on word count, on resolution. That starts with understanding what the page-one results already do and where they leave the searcher wanting. SEO Rocket’s competitor gap analysis surfaces the topics and angles rivals rank for that you do not, and its validation-gated AI writer enforces completeness and structure so drafts actually cover the query rather than skimming it — the difference between a page that earns the last longest click and one that gets pogo-sticked. This is the same playbook that scaled a portfolio past 1,000,000+ ranking pages: build the result searchers prefer, then let the behavioral systems notice.
The Honest Limits of What We Know
Navboost is real and confirmed, but precision here is a trap. We know the metric names, the window, the existence of squashing, and Nayak’s confirmation of the system. We do not have the exact weightings, how it interacts with every other component, or how much of the leaked schema is live versus deprecated. One internal note surfaced in the trial suggested a senior engineer considered the system extraordinarily powerful relative to the rest of ranking — a striking claim, but one internal opinion, not a published weighting. Treat anyone selling a precise “Navboost formula” or a guaranteed click-injection service with deep suspicion. The reliable conclusion is directional: satisfied searchers help you, dissatisfied ones hurt you, over a long horizon, and you influence that by being the better answer.
Frequently Asked Questions
Does Navboost mean click-through rate is a direct ranking factor?
Not in the naive sense. The system uses click behavior, but it distinguishes good clicks from bad clicks and weighs dwell via the last longest click. Raw CTR won by a misleading title generates bad clicks and can hurt you. It is satisfaction, aggregated over ~13 months, not click count, that the system rewards.
Can I improve rankings by buying or automating clicks?
Reliably, no. Google’s squashing function caps signal contribution, the 13-month window dilutes bursts, and manipulation patterns are detectable. Injected clicks may cause a brief flicker on low-competition terms and nothing durable on valuable ones — with reputational and detection risk attached.
How is Navboost different from Glue?
Navboost re-ranks the standard web results using click signals. Glue applies the same behavioral logic to the whole search page — images, snippets, “people also ask,” local packs — deciding which features earn space. One orders the blue links; the other assembles everything around them.
How long does it take to recover from bad click signals?
Months, not days, because the data is averaged over a roughly 13-month rolling window. Fixing the underlying problem — intent mismatch, weak titles, unsatisfying pages — starts re-averaging your history, but the memory in the system means patience is part of the fix.
The Bottom Line
Navboost ended the “does Google use clicks” debate, but it did not hand SEOs a new lever to pull. It confirmed what the best practitioners already built toward: the durable win is being the result searchers choose and stay with, measured patiently over a year of behavior and defended hard against anyone trying to fake it. You do not optimize for it by chasing clicks. You optimize for it by deserving them.
