What Google Actually Ranks: Court Evidence and a Live Audit

">

SEO

What Google Actually Ranks: Court Evidence, a Leak, and a Live Audit

Google denied using click data, domain authority, and a sandbox for 15 years. Then the evidence came out. Here is what it says, plus a measured head-to-head of four real sites.

July 25, 2026

6 min read

Nobody outside Google has "the algorithm." Anyone who says otherwise is selling something they cannot deliver.

But for the first time in Google's history, we have hard evidence instead of guesses. Three kinds of it:

Sworn testimony. In the US v. Google antitrust trial, Google executives confirmed under oath how core ranking systems work, including systems Google had publicly denied for 15 years.

A leak of Google's own internal API. In May 2024, more than 2,500 internal documents describing 14,014 ranking-related attributes from Google's Content Warehouse were published. These are Google's own variable names and descriptions.

Live data. We audited four competing sites in one commercial niche, three of them less than a year old, with full Search Console access to one. You can see exactly what separated page 1 from page 5.

They Denied It, Then It Showed Up in the Evidence

This table is why "Google says" is a weak source category:

Google's public position<br>What the evidence showed

"We don't use clicks for ranking" (said repeatedly, 2009 to 2022)<br>Navboost, confirmed under oath as one of Google's strongest signals, built on 13 months of click data sliced by position, device, and location

"There is no domain authority score"<br>A siteAuthority attribute in the quality signals module of the leaked API

"There is no sandbox for new sites"<br>A hostAge attribute, documented as used to sandbox fresh spam

"We don't use Chrome data in search"<br>Chrome data confirmed at trial and present in the leak (chromeInTotal)

The lesson is not "Google lies about everything." It is: when official guidance conflicts with observed behavior, trust observed behavior.

The Live Audit: Four Sites, One Niche, July 2026

The niche: one-time-payment QR code generators. Commercial intent, real money, and small enough that new sites can compete. Domain ages verified by WHOIS on July 25, 2026.

Site<br>Age<br>Money page<br>Off-page footprint<br>Google outcome

qrcd.com<br>3.4 years<br>~1,300 words, no schema, exact-match H1<br>Social profiles, sparse links<br>Page 1

lifetimeqrcodes.com<br>7 months<br>~4,500 words, 100+ internal links, FAQ<br>Near zero third-party mentions<br>Page 1

ownqrcode.com<br>7.5 months<br>Blog post targeting the exact query<br>Near zero third-party mentions<br>Page 1

The test site<br>3 months<br>42 pages, full schema, tuned titles, 38/42 indexed<br>Zero third-party mentions<br>Position 47 to 92, 0 clicks

The test site is the one we have full Search Console access to. Its weekly impression series tells the story on its own (every week had 0 clicks):

Period<br>Weekly impressions<br>What is happening

Week of launch53Google discovers the domain<br>Week 2139New-domain honeymoon: Google samples you<br>Week 34Honeymoon ends, trust window begins<br>Weeks 4 to 911 to 20Flat, position 70 to 100<br>Weeks 10 to 1322 to 45Slow climb as trust accrues

What This Experiment Proves

On-page perfection does not beat the trust window. The 3-month-old site has objectively heavier on-page SEO (schema on 40+ pages, tuned titles, clean snippets) than any of the winners. It ranks 40+ positions behind a 1,300 word page with no schema.

You do not need years or a big backlink profile in a micro-niche. Two of the page 1 sites are seven months old with almost no links. Past the trust window, exact intent match plus minimal entity legitimacy (social profiles, a few mentions, consistent branding) was enough.

The honeymoon is real, and it is not a ranking. The week 2 impression spike then crash is Google sampling a new domain, not a lost position. Do not "fix" things in a panic when it crashes.

Migrating domains resets the clock. The 3-month-old domain is a working product with sales history; it moved from a subdomain it had ranked on, and the rankings did not transfer with the 301s in any meaningful timeframe. If you rebrand, you are choosing to re-enter the trust window. Price that in.

What To Do With This

Five things the evidence supports for small and new sites:

Pick fights you can win. If page 1 for your query is all household brands, find the adjacent query where sites like yours already rank.

Match the query in your title, exactly. The leak has an attribute for it (titlematchScore). The page 1 winner's H1 is the query, word for word.

Keep your domain. Age is an input, and migration re-enters the trust window. Expect roughly 3 to 9 months before a new domain ranks for commercial queries.

Get your first ten legitimate mentions. Directories, social profiles, real communities. The winners had thin footprints, but thin beats zero, and mention #1 has the highest marginal value you will ever get.

Win the click, then end the search. Once you reach pages 1 and 2, Navboost...