Google algorithm leak: links, small sites and E-E-A-T myths
Short answer The 2024 leak was real: Google confirmed the documents but warned they were out of context. They list thousands of data fields, not their weights. They show Google tracks links by quality, clicks, a site-level authority score and how focused a site is. For a small business, the advice stays the same: earned links, a clear focus, pages people use.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callIs the Google search leak real?
Yes, but what leaked was a list of ingredients, not the recipe. In March 2024, internal documentation for Google's search systems was posted publicly on GitHub, apparently by accident, and taken down on May 7, as reported by SparkToro. SEO analysts Rand Fishkin (SparkToro) and Mike King (iPullRank) wrote it up in late May 2024: about 2,500 pages describing 14,014 attributes (named data fields) across 2,596 modules.
Google confirmed the documents were authentic, telling The Verge:
"We would caution against making inaccurate assumptions about Search based on out-of-context, outdated, or incomplete information."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)
What the documents can and can't tell you:
| The leak shows | The leak doesn't show |
|---|---|
| Which data Google stores about pages, sites, links and clicks | How much any field counts toward ranking |
| Names of systems and demotions (penalties built into the ranking) | Whether a field is live, retired or only used in tests |
| That some things Google long played down are tracked | What changed after March 2024 |
SparkToro's own write-up says the documents don't show how much any element weighs in ranking.
Do backlinks still matter after the leak?
Yes. The leak shows Google sorts links by quality, and Google says so in public too. Mike King's analysis describes link quality tiers, where a link from a page in a higher tier is worth more, and a demotion when the link's text doesn't match the page it points to. SparkToro's source said links in the lowest tier are simply ignored. Google's public explanation of ranking already says:
"One of the factors used to determine quality is understanding if other prominent websites link or refer to the content." (Google Search: How Search works, ranking results)
The trap is the advice that followed the leak: "build lots of different link types and vary your anchor text" (the clickable words of a link). Google's spam policies name several of those tactics as link spam:
| Common post-leak advice | What Google's spam policies say |
|---|---|
| Add directory and bookmarking links for variety | "Low-quality directory or bookmark site links" are link spam |
| Guest posts with keyword links back to you | "links with optimized anchor text in articles, guest posts, or press releases" are link spam |
| Buy a few strong links | Paying or trading goods or services for links, or excessive link swaps, is link spam unless the links are marked sponsored or nofollow |
| Rotate keyword-rich anchor text | Google asks for anchor text that is "descriptive, reasonably concise, and relevant" (Link best practices) |
For a local business, the links worth having come from things that happen anyway: a supplier's dealer page, the chamber of commerce, a local news story, a sponsorship that lists you. How to get them safely: backlinks for a local business.
Does Google have a site authority score?
The documents include one, called siteAuthority, and Mike King's analysis shows it is used in a ranking system. They don't say how it is calculated or weighted. Don't confuse it with Domain Authority, a 1–100 score made by the SEO tool Moz, which Moz itself says is not a direct ranking factor.
What it means for you: a site's overall reputation can lift or hold back each page. One great page on a site full of thin or copied pages starts at a disadvantage. Cleaning up weak pages can matter as much as adding new ones.
Does Google use clicks and Chrome data?
Yes for clicks; Chrome data is in the documents, but its use is unclear. The leak describes a system called NavBoost that counts good and bad clicks, and Google's search antitrust trial put it on the record. The court wrote:
"NavBoost, one of Google's core ranking models, runs on 13 months of Google click-and-query data." (U.S. v. Google, remedies opinion, Sept. 2, 2025, quoting the court's 2024 findings)
Google's own page says it uses "aggregated and anonymized interaction data to assess whether search results are relevant to queries." The documents also list site-level counts of Chrome views, per iPullRank. They don't say how Chrome data feeds ranking, so claims that time on page is a direct ranking factor go past the evidence.
The useful lesson: if searchers click your result and then go straight back to look for another, that is the behavior the data captures. A page that answers the question, shows the price range and puts the phone number where a thumb can reach it is what you control.
Is there a small publisher classifier?
The documents include a flag for a "small personal site", per iPullRank. Nothing in them says what it does: it could demote, protect, or just label. Claims that Google targets small sites with it are guesses.
What Google has said in public points the other way. In its August 2024 core update post, which it said reflected feedback from creators, Google wrote that it aims to show high-quality sites:
"including small or independent sites that are creating useful, original content, when relevant to users' searches." (Google Search Central Blog: August 2024 core update)
A local plumber or dentist is also not a "personal site" in the usual sense; it's a business site. If your traffic fell after an update, the checks are in traffic dropped after a Google core update.
Is E-E-A-T fake?
No, but it was never a single score. E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness, the qualities Google's human raters look for. Mike King's analysis names no E-E-A-T field, though it notes the documents record who authored a page. Google said the same long before:
"While E-E-A-T itself isn't a specific ranking factor, using a mix of factors that can identify content with good E-E-A-T is useful." (Google Search Central: Creating helpful content)
The same page says trust matters most and asks whether pages carry a byline where one might be expected. For a business site, that means a real owner or staff name, licenses, photos of real jobs and an address that matches your Google Business Profile. More: E-E-A-T for a small business.
Does word count or topic focus matter?
Length doesn't; focus does. Google's guidance asks whether you write to a set length because you heard Google prefers one, and answers "(No, we don't.)". The leak fits this: per iPullRank, the documents suggest Google only reads up to a set number of tokens (chunks of words) per page, so padding a page adds nothing, and short content gets scored for originality.
The documents also include siteFocusScore, which measures how closely a site sticks to one topic, and siteRadius, how far a page wanders from the site's main theme. Google's helpful content page asks a matching question: did you enter a niche "without any real expertise" because you thought it would bring traffic?
| Do | Skip |
|---|---|
| One page per service you actually sell, said fully | Stretching pages to 2,000 words |
| Posts that answer your customers' questions | Posts on trending topics outside your trade |
| Merge or delete thin pages that drift off-topic | Adding pages just to look bigger |
Page-by-page checklist: on-page SEO for service pages.
Example
Illustrative cases with invented numbers.
A roofer offered a $500 link package (US dollars)
The seller promises 50 "diverse" links: directories, bookmarking sites and guest posts with "roof repair Dallas" as the link text. Google's spam policies list each of those as link spam. The roofer spends the $500 sponsoring a youth team whose site links to him, and asks two suppliers to add him to their installer pages.
A cleaning company blog (Canadian dollars)
The owner pays C$150 a post for 1,500-word articles on home decor, travel and recipes. Traffic rises a little; calls don't. She stops, deletes six off-topic posts, and writes four short posts on move-out cleaning prices and what's included, the questions callers ask.
A dental clinic with no names on the site
The site says "our team" everywhere. The clinic adds each dentist's name, licence, photo and a short bio, and a byline on its treatment pages. That is the practical E-E-A-T fix.
FAQ
Is the Google leak still relevant in 2026?
As a snapshot, yes. The documents date from March 2024, and Google changes its systems constantly, but the court record from 2025 confirmed the click data parts.
Did the leak show a list of ranking factors?
No. It showed fields Google stores, not which ones rank pages or how much. Anyone selling a "leak-proven" checklist is going past the documents.
Should I build more backlinks after the leak?
Only earned ones. Links still count, and the leak suggests low-quality links are ignored rather than helpful. Bought or swapped links break Google's spam policies.
Does the leak change anything for local search on Google Maps?
The analyses focused on web search. Google names relevance, distance and prominence as its local ranking factors (Google Business Profile Help: Tips to improve your local ranking). See how to rank higher on Google Maps.
Different for you?
- You want links without risking a penalty: backlinks for a local business.
- You want to show trust on your site: E-E-A-T for a small business.
- Your traffic dropped after an update: traffic dropped after a Google core update.
- You're deciding whether to blog at all: do blogs help SEO?
- You'd rather have the pages written and checked: SEO & GEO Agent.