Is CTR a Google Ranking Factor? NavBoost & the API Leak Explained

What Google said publicly about CTR (2016 to 2019)

Google's public position for years was that click-through rate is not a direct ranking factor. At Pubcon Las Vegas in 2016, Gary Illyes said CTR is not a ranking factor, describing clicks as incredibly noisy because people click around a lot, and adding that click data is used for personalisation, for example to learn which kinds of results a user prefers. In a 2019 Reddit AMA he went further, calling dwell time, CTR and similar theories 'made up crap', and he also said that using clicks directly in ranking would be a mistake. Google also said it used click data in controlled experiments: Search Engine Journal reports that a Googler's SMX West 2016 presentation showed CTR used alongside other engagement metrics in search quality tests. Matt Cutts and John Mueller, for their part, said Chrome data was not used for organic ranking.

Read closely, many of these statements are narrow: 'not a direct factor' or 'not used directly'. That wording left room for click data to feed systems that are not called 'CTR'. Google's own How Search Works page today says it uses 'aggregated and anonymized interaction data to assess whether search results are relevant to queries'. That sentence is the bridge between the old denials and the court record: interaction data is used, but aggregated, anonymised and as a check on relevance rather than as a simple vote.

NavBoost in the DOJ v. Google trial

The US v. Google search antitrust case (No. 1:20-cv-03010, District of Columbia, Judge Amit Mehta) put Google engineers under oath. On October 18, 2023, day 24 of the bench trial, Pandu Nayak, then Google's vice president of search, was asked about a 2017 launch report. He testified that it concerned the NavBoost system, which used click and query data from session logs, and that Google reduced the amount of that data by about a third with 'no meaningful change in search quality'. SEO analysts, including Mike King, connect this to NavBoost moving from 18 months of data to 13. A later court exhibit, PXR0357, summarises a January 31, 2025 call with Nayak during the remedies phase. It defines NavBoost as 'a measure of how frequently users (subset by location and device type) click on a particular document for a particular query', calls it a traditional signal, and says it 'uses most recent 13 months of data'. It also describes NavBoost as a query-to-document lookup table of counts. Trial coverage dates NavBoost to around 2005.

What the same court documents say about clicks and manipulation

The DOJ material does not say 'CTR is a ranking factor' in the way SEO blogs often put it. It describes something more specific. PXR0357 notes that Google uses over 100 raw signals, that top-level signals combine them, and that how many signals are affected by user-side data is 'difficult to quantify'. It also says Google avoids simply 'predicting clicks' because clicks are easily manipulated and are a poor proxy for user experience. An anecdote in the same exhibit explains why raw CTR is a weak measure: an early signal that compared how often a link was shown with how often it was clicked turned out to be biased by position, because the first result gets clicked more whether or not it is better.

Trial reporting also quotes Nayak saying you get NavBoost only after documents are retrieved in the first place, meaning clicks help reorder and narrow a candidate set rather than decide which pages are eligible. The practical reading: clicks are a real input, aggregated over long windows, corrected for position and filtered for abuse. That combination explains why both camps feel vindicated. Those who said 'CTR is not a ranking factor' were right about the raw ratio; those who said 'clicks matter' were right about user behaviour feeding a ranking system.

The May 2024 Content Warehouse API leak

On May 27 and 28, 2024, Mike King (iPullRank) and Rand Fishkin (SparkToro) published analyses of internal Google documentation for its Content Warehouse API that had been accidentally exposed. Fishkin reported more than 2,500 pages and 14,014 attributes; King counted 2,596 modules. The source, SEO practitioner Erfan Azimi, later identified himself. At the end of May 2024 a Google spokesperson, speaking to The Verge, confirmed the documents were genuine while warning against drawing conclusions from out-of-context, outdated or incomplete information. The leak names NavBoost dozens of times and includes click fields such as goodClicks, badClicks, lastLongestClicks and impressions, as well as chromeInTotal, described as site-level Chrome views. King notes that click data appears to be segmented by country and device and that long clicks work much like dwell time even though the term is never used.

• It does not give weights: no field comes with a number saying how much it moves a ranking.
• It does not prove current use: some fields are marked deprecated or no longer populated.
• It does not describe thresholds: nothing says how many clicks, over what period, change a position.
• It does not separate experiments from production: an attribute can exist for testing only.
• It does describe what Google can store: clicks by query and document, quality of click, device and country segments.

How Rand Fishkin and Mike King interpret it

Fishkin's reading focuses on credibility. He argues the documents conflict with years of public statements, on clicks, Chrome data, subdomains, a sandbox for new sites and domain age, and that SEOs should weigh Google's public guidance with that in mind. King's analysis is more technical. He describes NavBoost as a click-driven re-ranking system, explains 'twiddlers' as re-ranking functions that run after the main scoring, and points to fields such as siteAuthority and hostAge that contradict earlier denials.

Long before the leak, Fishkin had run public click tests: Search Engine Journal recounts a 2014 experiment in which a blog post reached position one the same evening after 228 visits from people who searched and clicked, while Fishkin himself said it did not prove Google uses click volume directly. Eric Enge suggested the effect may come from freshness systems for trending queries. Taken together, the consensus among careful analysts is that clicks influence ranking through NavBoost-style systems, but slowly, in aggregate, and only for pages already in contention. That is very different from the claims made by many CTR manipulation vendors, which promise quick jumps from short bursts.

Why raw CTR misleads: position bias in the numbers

Google's own exhibit says an early click signal was biased by position, and public studies show how strong that bias is. Backlinko's analysis of about 4 million Google results (12.2 million queries from Search Console data provided by Semrush, last updated April 2025) found that the first organic result has an average CTR of 27.6%, that the top three results take 54.4% of all clicks, and that only 0.63% of searchers clicked anything on page two. Moving up one position raised CTR by 2.8 percentage points on average, and moving from second to first increased clicks by 74.5%. On top of that, SparkToro's 2024 zero-click study with Datos found that 58.5% of US Google searches and 59.7% of EU searches ended without a click, leaving 360 and 374 clicks to the open web per 1,000 searches. A page at position six with a 'low' CTR may therefore be performing normally for its slot. Any system that rewarded raw CTR would simply reward pages already at the top, which is why Google compares clicks against what is expected for a position.

So, is CTR a ranking factor? A precise answer

If 'CTR is a ranking factor' means Google divides clicks by impressions for your page and moves you up when the ratio rises, the evidence says no: Google itself found raw CTR biased by position and avoids optimising for clicks alone. If it means user clicks on search results influence rankings, the evidence says yes: sworn testimony and a court exhibit describe NavBoost counting how often users click a document for a query, by location and device, over 13 months, and the leak shows good, bad and last-longest clicks being recorded. Three nuances matter. First, it is the quality of the click, whether it ends the search, that seems to count, not just its existence. Second, the data is aggregated over long windows, so short bursts carry little weight. Third, NavBoost works on results that are already relevant, so clicks will not rescue a page Google does not consider a candidate. None of the sources lets anyone predict a ranking change from a given number of clicks.

What this means in practice for your SEO

Start with the clicks you can earn honestly. The levers below follow directly from what the record describes: satisfied clicks, aggregated over months, segmented by place and device, on pages that are already relevant. Use Search Console to find queries with high impressions and low CTR for your position, and test changes there. Remember that click data is segmented by location and device, so a page can perform very differently in different countries or on mobile. If you test real-user search campaigns, keep them proportionate, spread over time and limited to queries where you already rank on the first pages, and measure them against a control. Google's spam policies prohibit machine-generated traffic, meaning automated queries, so bots are both against the rules and the easiest pattern to filter; our comparison of CTR bots and real human clicks goes into the detection side.

• Rewrite titles and meta descriptions so they match the query's intent; Backlinko found 40 to 60 character titles had the highest CTR.
• Put the answer near the top of the page so your result is the last, long click of the session.
• Compare your CTR with the benchmark for your position, not with an absolute target.
• Segment Search Console by country and device before drawing conclusions.
• Build brand demand: people who recognise your name choose your result over unknown ones.
• Change one thing at a time and wait several weeks; click systems aggregate over months.

Red flags when a provider sells 'CTR ranking'

Because the topic is murky, it attracts overconfident sellers. The court record and the leak give you a practical filter: anything that contradicts aggregation, position correction or manipulation filtering is a warning sign. Use the list below when comparing offers, whether you are reading a review of the best CTR manipulation tools or a vendor's sales page. A serious provider will tell you what it cannot control, show you per-visit proof, and let you stop at any time. It will also ask which queries you already rank for, because the court record says NavBoost re-ranks documents that were already retrieved, and it will refuse volumes that are out of proportion with real search demand for the keyword. If a seller cannot explain where its visitors physically are, what devices they use and how a visit is verified, assume the traffic is automated and that Google's spam systems are designed to discount it.

• Guaranteed ranking gains or 'page one in 7 days' promises: nothing in NavBoost evidence supports a guarantee.
• Traffic from software, emulators or residential proxies presented as 'human-like': it is machine-generated under Google's policies.
• No proof per visit (search, result page, click), only a counter in a dashboard.
• Hundreds of visits a day on a query that gets ten real searches a month.
• City-level or 'any country' targeting promises with no explanation of where the supply actually is.
• Advice to target keywords where you rank nowhere: NavBoost acts on already-retrieved documents.

What our own platform data shows about real search behaviour

Our Google SEO Boost product asks a real worker to search a keyword, find the advertiser's site and click it, and each worker reports where the site was found. Between October 26 and November 12, 2026, workers completed 179 visits across 4 campaigns. Among the 134 proofs with a readable answer, 56% found the site on the first results page, 24% on pages two to five, 16% did not find it in the first five pages and reached it through an address search, and 4% were unclear. Workers declared 56% mobile and 44% desktop devices. Two lessons follow, with the caveat that this is a small, early sample from a few advertisers. First, many advertisers test keywords where they are not yet on page one, which is exactly where NavBoost is least likely to help, since clicks reorder pages already in contention. Second, a real searcher who cannot find a site in five pages is a useful signal in itself: the keyword is probably too ambitious for click testing and needs content and links first. These figures describe behaviour, not ranking outcomes; we have no controlled evidence that the visits changed any position.

Testing click behaviour with real people on Hitworkers

If you want to observe how real search behaviour affects a page, Hitworkers' Google SEO Boost uses real people, not bots. A worker in the country you target searches your keyword on Google, finds your site in the results (up to page five, or through an address search), clicks it, visits up to the number of pages you set for about 30 seconds each, and sends proof. Visits cost $0.30 each, flat in any country, in packs of 30, 100 or 200 or a custom volume, with a drip feed over one to seven days. A worker never repeats the same URL for the same advertiser. Be realistic about geography: our 1,392 verified workers are in 106 countries, but the 651 active in the last 30 days are concentrated in Kenya, Indonesia, Nigeria, Pakistan and the Philippines, with only 16 active in the United States, 4 in the United Kingdom and 14 across the EU. Large US, UK or EU volumes are not available today.

• Pick 3 to 5 queries where you already rank on page one or two, and note baseline CTR and position.
• Choose a country where real supply exists, and a modest volume spread over seven days.
• Launch from the Google SEO Boost page and review each proof.
• Leave comparable queries untouched as a control.
• Compare Search Console and rank tracker data after four to six weeks, then stop if there is no value.

We cannot promise that any number of visits will change your rankings, and nothing in the DOJ record or the leak lets anyone promise that. Treat it as a measured experiment alongside better titles, content and links. For the wider picture of methods and risks, read our guide to Google CTR manipulation, and for local results, see how engagement works in Google Maps CTR.