Google search data leak: is it real and what it means for SEO
Short answer Yes. Google confirmed in May 2024 that the leaked internal documents are authentic, but warned they may be out of context, outdated or incomplete. They list about 14,000 data fields, including click counts and a site-wide authority score, with no weights. Use them as hints, not a recipe: earn clicks honestly and keep every page useful.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callIs the Google search data leak real?
Yes. Google confirmed it. A Google spokesperson told The Verge on May 29, 2024:
"We would caution against making inaccurate assumptions about Search based on out-of-context, outdated, or incomplete information." (Google spokesperson Davis Thompson, to The Verge)
So the documents are Google's own. What they prove is narrower than most headlines claimed.
| Date | What happened |
|---|---|
| March 27, 2024 | Internal reference documents for Google's "Content Warehouse" API appear in a public code repository on GitHub |
| May 7, 2024 | They are removed, but copies stay online on Hexdocs, a code documentation site |
| May 27–28, 2024 | SEO researchers Rand Fishkin (SparkToro) and Mike King (iPullRank) publish them; the source is later named as SEO practitioner Erfan Azimi |
| May 29, 2024 | Google confirms they are authentic |
The leak did not come from a court case. It is easy to mix up because Google's antitrust trial with the U.S. Department of Justice ran at the same time and made some of the same systems public. That trial's court record is a separate source, and it backs up the biggest finding below.
What the leaked documents are, and aren't
They are an API reference: a dictionary of data fields Google stores about pages and sites. SparkToro counted more than 2,500 pages and 14,014 fields.
They are not the ranking formula. There is no code, and no field says how much it counts. Mike King, who reviewed them first, wrote:
"Unless explicitly indicated, we don't know how things are being used." (iPullRank, May 27, 2024)
Think of it as the column list of a spreadsheet, without the formulas.
What the leak shows about Google ranking factors
Six findings matter for a business website. Each row says what the documents show, what Google says in public, and what it means for you.
| Field in the documents | What it says | What Google says publicly | What it means for you |
|---|---|---|---|
| NavBoost: goodClicks, badClicks, lastLongestClicks | Counts of good and bad clicks, and of clicks that were the last and longest in a search | Google uses "aggregated and anonymized interaction data to assess whether search results are relevant to queries" (How Search works) | Clicks from real searchers count. Be the result people pick and don't come back from |
| siteAuthority | A site-wide score "applied in Qstar", Google's page quality measure | Ranking works on the page level, but site-wide signals exist (Search Central: ranking systems guide) | Weak pages may weigh on strong ones, though Google says one weak signal doesn't sink a whole site |
| hostAge | The first date Google saw any page on the domain, "used in twiddler to sandbox fresh spam in serving time" | Google had denied a "sandbox" for new sites, as SparkToro and iPullRank note | A holding pen aimed at spam, not a penalty on every new site |
| chromeInTotal | "Site-level Chrome views" | Google representatives had said Chrome data doesn't feed rankings (The Verge) | Google stores Chrome visit counts per site; how it uses them is unknown |
| exactMatchDomainDemotion | A demotion score for domains built to match a search | Google's ranking guide openly lists an exact match domain system | A keyword domain won't carry a weak site. See exact match domains |
| smallPersonalSite | "Score of small personal site promotion" | Not addressed | A boost for small personal sites may exist; its use is unknown |
Does Google use clicks to rank?
Yes, in some form. This is the one finding confirmed from three directions. Beyond the leak, a Justice Department record of a January 2025 call with Google engineer Pandu Nayak describes NavBoost as a measure of how often users click a particular page for a particular search, using the most recent 13 months of data. Google's own How Search works page now says it uses interaction data.
The documents contain no field called "dwell time." The closest is lastLongestClicks: the click a searcher made last and stayed on longest, which suggests they found what they needed.
What to ignore
Most of what was sold off the back of the leak is noise.
- "14,000 ranking factors." A field is stored data, not a factor. Some are marked legacy or "DO NOT USE" in the documents. None has a weight.
- Treating it as current. The documents date from March 2024 or earlier. Google says it "regularly" improves its ranking systems, so some fields may have changed or gone since (Search Central: ranking systems guide).
- Click bots and paid clicks. The same Justice Department notes say Google avoids simply predicting clicks because clicks "are easily manipulated." SparkToro's source said Google filters click spam using cookie history, logged-in Chrome data and pattern detection. Fake clicks are a cost with a downside.
- Panic disavowing links. Disavowing means telling Google to ignore certain links to your site. Google says you rarely need it:
"In most cases, Google can assess which links to trust without additional guidance, so most sites will not need to use this tool." (Search Console Help: Disavow links to your site)
Google limits it to sites with many spammy links that caused, or likely will cause, a manual action (a penalty applied by a Google reviewer).
What a small business should do now
The leak mostly confirms advice that was already sound. In order:
- Earn the click. Write each page's title and description to promise exactly what the page delivers: the service, the town, the next step. On-page SEO for service pages covers the details.
- End the search. Put what the searcher came for near the top: price range, area served, hours, phone. A visitor who finds it doesn't go back to Google to try the next result.
- Fix or remove weak pages. A site-wide score exists, and thin or outdated pages may pull it down. See update, merge or delete old pages.
- Give a new site time, without shortcuts. The sandbox field targets fresh spam. Spammy links or mass pages on a new domain look like exactly that. See how long SEO takes.
- Don't buy clicks, reviews or link packages. Earn backlinks from real local ties instead.
Example
A plumber who ranks but gets few clicks
A plumber's "water heater repair" page sits at position 4 and gets few clicks. Its title is just the company name. Rewritten to "Water heater repair in Mesa: same-day, from $150", it gives searchers a reason to choose it. That is the kind of honest click signal the documents describe.
A new accounting firm worried about the sandbox
A firm launches a site in January and sees little Google traffic by March. A seller offers 500 links for $99 to "escape the sandbox." That is the pattern the hostAge field is built to hold back. The firm adds real pages, a Business Profile and local links, and waits.
A salon with 200 thin blog posts
A salon has 200 short posts copied from supplier newsletters. It merges the useful ones into 10 strong service pages and deletes the rest, so the site's overall quality reflects its best work.
FAQ
Where can I read the leaked Google documents?
Copies remain on Hexdocs under google_api_content_warehouse. They are technical; the SparkToro and iPullRank write-ups are easier to follow.
Did Google lie about using clicks?
Google representatives had said clicks don't feed rankings. The leak and antitrust record show they feed NavBoost, and Google now says it uses interaction data.
Does the leak mean domain authority is real?
A siteAuthority field exists, but it's Google's internal score, not the "Domain Authority" number SEO tools sell. Nobody outside Google knows its formula.
Should I change my SEO because of the leak?
Only if you use shortcuts. Clear pages, real reviews and useful content fit the documents. Click schemes and bought links get filtered.
Different for you?
- Your site never ranked at all: check indexing first in website not showing on Google.
- Traffic fell after an update: see traffic dropped after a Google core update.
- You got a manual action: follow Google manual action: find it, fix it.
- You want someone to do the page and link work: see SEO.