DOJ Leak; Google’s ranking stack exposed

NavBoost, RankEmbed & Twiddlers

[Whitey]

Another great MUST READ article from Brett_Tabke. from the DOJ’s unsealed files, shows rare insight into how Google actually ranks search results.

Forget the theories, this is based on internal documentation cited in the U.S. antitrust case against Google.

TL:DNR Here’s What We Learned:

•PageRank still lives; Google still weights traditional factors like backlinks, keywords, and relevance using a linear formula.

•NavBoost matters (a lot) ; Aggregates click data over 13 months to boost content users engage with. This is not machine learning, just a massive table of behavioral signals.

•RankEmbed & Q*Star; AI and language models are in play. RankEmbed handles semantic relevance. Q*Star is an internal “quality score” guiding rankings.

•Twiddlers; Post-ranking tweaks to elevate freshness, demote spam, and fine-tune the final results.

•Stack structure revealed. From crawling to query understanding to logging; the entire flow from GoogleBot to GWS is detailed.

Full article with breakdown:
[searchengineworld.com...]

What does this mean for SEO in the age of AI? Are traditional tactics more valuable than we thought? Let’s discuss.

[Fluff_Nutz]

Does any of this still matter when there is an AIO above everything, so people can read that and instantly leave? Also, Google just de-index your articles because they don't care about your or your tiny site. I no longer see any of my articles on the G SERP any more, none, zip.

Went from ranking top spot for weeks during my first year with a site. Felt surreal, it sure was. WHAM ''update'' and all lost. Don't care about Google. Thankfully I can get traffic elsewhere..

[tangor]

The list offered is kind of interesting in that all of it is HUMAN MANAGED (even the AI stuff).

So much for the mythical, marvelical, magical Black Box algo...* too complicated to explain to ordinary webmasters.

^{*(IT HAS ALWAYS been human driven and will CONTINUE to be human driven... after all who writes the AI code? OR who VETS any AI generated code which goes horribly wrong when put in practice? HUMANS)}

[Whitey]

@Fluff_Nutz - Fair point. AIO is a traffic siphon. But knowing how Google works still helps us adapt, even if it’s just to pivot smarter.

@tangor - Exactly. Spot on. The myth of full automation is overstated. These leaks prove there’s still a human hand on the dial.

[Whitey]

For those unfamiliar with a lineal scoring method used by Google, here’s a deeper but simplified explanation.

TL;DNR: Google Still Uses a Traditional Scoring Formula

Google ranks search results, yes, even in the AI era, with traditional SEO factors that still matter.

These internal documents from the DOJ antitrust case confirms that the linear weighted formula to calculate rankings, includes factors like:

•Backlinks (PageRank)
•Keyword relevance
•Anchor text
•Freshness
•On-page content signals

…which are each assigned a numerical weight, and then combined to generate a score for each page.

So think of it like this:

Score = (Weight1 × Backlinks) + (Weight2 × Keywords) + (Weight3 × Freshness) + …

It’s not all machine learning. This is an old-school model, updated with AI on top, but still rooted in a human-coded ranking stack. We don’t know the exact weights, but we now know the structure still exists.

Bottom line: SEO basics are far from dead. I hope that clarification helps those grappling with the AI related turmoil.

[dvduval]

Really fascinating that so much stayed the same, especially how PageRank continued to play a major role. I agree with others that rankings are becoming less important, but it is still really interesting to me, as I remember when Google was just getting started, and being so happy to feel like I had "won" by ranking for so many things because I knew how to manipulate the search results. As many more people figured this out, Google felt that manipulation was such a problem that we got regular updates from people like Matt Cutts explaining how Google was fighting back.

Then the internet kept growing until big players more and more "owned" the key positions in Google and "long tail keywords" were the rage. Now even ranking for long tail is not easy with so much content being produced, first by humans and now increasingly by AI.

Google first indexed all the sites, but now AI has tokenized all the content from all publicly available (and a lot of privately available) data. The search engine itself is becoming less relevant.
- Any informational question can be answered by AI.
- major players dominate their markets (social, shopping) and we don't need Google to use them
- Google itself acknowledges this with AIO
- Ads have depreciated relevant results long ago, and make up so much of the page

And finally, the idea of having your company or page or profile "found" is much less dependent on Google for all of the above. I believe if Google would have focused more on relevancy and helping people find information as was their original purpose, they would be the clear leader in AI now. But instead they were so focused on advertising and profits that they forgot their purpose. It may not be too late...

[tangor]

But instead they were so focused on advertising and profits that they forgot their purpose. It may not be too late...

Forgot? Sorry, that was rule number One. It just took a bit of bait and switch to get it started, a few webillionaires here and there, then turning the screws for every last penny while killing off all the gamers and scammers chasing THEIR dollars with every animal update, now unnamed. These days AI just makes it easier:

Our content. Their AI Answer. Their Ads (not yours). What's not to like?

Remember: "What's mine is mine. What's yours is mine, too." Paraphrased from a human source (Aging Intelligence---a form of AI that can't recall at the moment the source, but if somebody asked ChatGPT I bet an answer can be found---AND NO SOURCE WILL BE REVEALED!).

^{If not obvious, much of the above is intended as humor. Up to you to decide what is humorous.}

[Whitey]

Roger Monti aka @martinibuster just dropped a solid analysis worth weaving into this discussion:

Google’s Trust Ranking Patent Shows How User Behavior is a Signal [searchenginejournal.com...]

TL;DNR Summary

Roger unpacks a Google patent focused on “trust ranking” which leverages user behavior signals (such as time-on-site, click paths, and user satisfaction indicators) to determine the trustworthiness of a document. Unlike simple engagement metrics, this patent focuses on whether users seem to complete their task after visiting a page (suggesting a deeper layer of behavioral evaluation tied to user intent and fulfillment). Pages that consistently meet user goals are promoted, while those that don’t are potentially downgraded.

The patent outlines multiple behavior-based scoring mechanisms including satisfaction scores, bounce analysis, and even backward click data (i.e. if users return to the SERP to refine their search, the initial page may lose trust points). While it’s not confirmation of live deployment, the signals align closely with what we’ve seen in NavBoost and the behavioral logic exposed in the DOJ leaks.

Tying It Back to the DOJ Docs

This trust-ranking patent dovetails with the DOJ revelations about NavBoost, a system based on 13 months of clickstream data shaping what gets surfaced in search. It reinforces the idea that user behavior isn’t just a side metric, it’s central to ranking decisions, both in pre-ranking (NavBoost) and post-ranking (Twiddlers). We’re seeing a clearer picture emerge: Google’s stack may still use PageRank and linear weighting, but user satisfaction is the long game signal they’re betting on to refine AI output and SERP integrity. It’s all part of the same machine, optimized to keep users clicking, trusting, and staying.

[Brett_Tabke]

Thanks Whitey, turned out to be top thread so far of the year on SearchEngineWorld.

[Kendo]

focused on “trust ranking” which leverages user behavior signals (such as time-on-site, click paths, and user satisfaction indicators)

And just where do they get these stats?

From Google Analytics? So web sites not running their logging script will not be rated?

From Chrome web browser? So only Chrome user activity is rated?

If such information is logged and analyzed, why can't anyone using Google's WMT see their own data?

[clairemiso299]

Interesting that everything ties back to behavioral signals rather than pure AI. NavBoost seems to be the missing piece explaining why some high-engagement pages rise without backlinks. Clearly, Google hasn’t abandoned classic factors, but now balances them with click and retention data. In practice, SEO is once again about useful content and good user experience, not just technical optimization.

[Whitey]

Interesting indeed. I wonder what the data engagement thresholds are for being part of Navboost.

Seems like the classic signals will be around for a while, but with AI Mode/Overviews muddying the playing field I guess SEO’s and site owners will have to wait for feedback.

[christianz]

Interesting that everything ties back to behavioral signals rather than pure AI. NavBoost seems to be the missing piece explaining why some high-engagement pages rise without backlinks. Clearly, Google hasn’t abandoned classic factors, but now balances them with click and retention data.

That was likely HCU, but Google has abandoned that concept and seem to rank useless zero engagement sites just fine in 2025.

[Whitey]

I reckon this is where things get messy…

NavBoost needs consistent positive engagement signals before it can do anything. Most “zero-engagement” sites probably don’t have negative signals either - especially if they’re thin but satisfy the click long enough that users bounce back to Google without pogo-sticking immediately.

In other words:

No love, no hate = stay invisible but not punished

Slight engagement = potential to rise

Searchers return fast = suppressed

So the mystery might not be why some bad sites rank well; it’s why so many better ones don’t get any engagement to activate the behavioral side.

Classic SEO still matters, but without clicks/retention, the algorithm’s blind.

Feels like Google is waiting for users to vote - and if your page never even gets a ballot, you’re stuck outside the election.