Ranking signals are about pages, not people

How Oernoe Search works publishes a section titled Privacy-Preserving Ranking Factors. It is the clearest public list of what the ranking system is supposed to look at, and what it is not supposed to look at. This Company essay stays on that list. Ranking signals are about pages: content relevance, authority and expertise, link analysis, freshness, aggregated anonymized satisfaction signals, mobile-friendliness, and page speed. They are not about building a personal dossier from search history, location, device type, age, or interests. The default is equal treatment for the same query. Explicit saved preferences are a separate, opt-in layer under an account—not a hidden profile inferred from every click.

## What the guide lists as primary signals

Content relevance is first for a reason. The algorithm is supposed to prefer pages that directly address the query over pages that only brush the topic. That is a page-to-query match problem. It does not require knowing who you are. Two people typing the same words should trigger the same relevance evaluation against the same indexed pages unless they have deliberately saved different preferences.

Authority and expertise ask whether a source looks established and trustworthy relative to other sources. Link analysis asks how other reputable sites cite the page. Freshness asks how recently the page was updated, which matters more for news and shifting topics than for stable reference material. None of those signals need a personal advertising profile. They need an index of public pages and relationships among those pages.

User satisfaction signals appear on the list with an important qualifier: aggregated and anonymized. The guide mentions bounce rate and click-through rate as examples processed at scale without tying data to individuals. That sentence is easy to misuse in either direction. Misuse one way: pretend Oernoe never looks at any aggregate click behavior. Misuse the other way: pretend aggregate satisfaction is a personal interest graph. The published middle is aggregate, anonymized, not a person-level advertising profile.

Mobile-friendliness and page speed are page qualities. A page that works on phones and loads quickly is easier to use. Ranking those qualities higher for the contexts where they matter is still about the page, not about tagging a visitor with a demographic interest label. The guide’s phrasing for mobile-friendliness notes that mobile-optimized sites rank higher for mobile searches—as a property of the page’s design relative to the request context for rendering, not as a claim that Search builds a device-based advertising persona of you.

## What is not on the list

The guide is explicit about the negative space. Notice what is not on the list: your search history, your location, your device type, your age, your interests, or any other personal data used as a personal ranking profile. The algorithm section earlier in the same guide repeats the equal-treatment idea: unlike systems that incorporate search history, location, device, or personal profile, Oernoe’s algorithm is described as treating every search equally by default.

That negative list is the company claim worth defending in public. It is also the claim that requires care. Technical logs may still receive IP address, browser and device type, and related diagnostics for security and reliability, as Privacy describes for the services generally. Ranking’s published story is that those operational facts are not turned into ranking personalization or advertising profiles. Keeping operational receipt and ranking inputs separate is part of how the guides ask to be read.

Interests deserve a special note because the word appears in both forbidden and allowed contexts. Interests as a personal profile inferred from behavior are what the ranking section says it does not use. Interests you deliberately encode by saving preferences—language, region, filters, collections—are explicit account choices. The guide’s personalization section calls that difference out: personalization is explicit and under your control; Search does not automatically learn about your interests or build a detailed profile from behavior.

## Equal treatment versus saved preferences

The pizza example in How Ranking Works is useful because it is concrete. If you and a friend search for the same phrase, you should get the same results unless an account has custom preferences you saved. That “unless” is not a loophole for silent profiling. It is the documented exception for settings you chose. Language, region, result filtering, blocked domains, saved searches, and bookmarked sources are the kinds of controls the guide lists. They are visible, editable, and deletable through account settings according to the product writing.

Equal treatment is therefore the default for ordinary queries, including queries without an account. Search does not require login for ordinary use. An account is what lets you save preferences you chose, not a hidden profile inferred from every click—language that also appears in Using an Oernoe account. The two guides agree: ranking starts from pages and queries; personalization, when present, is explicit.

This essay is not a rehash of “saved preferences are not an ad profile.” That Privacy piece centers the personalization-without-tracking section and the difference between a setting and an advertising graph. This Company piece centers the ranking-factor list itself: what signals exist, what signals are denied, and why the company presents ranking as page-centric. Preferences appear here only as the documented equal-treatment exception.

## Ranking is not the crawl promise

Privacy-first crawling and a finite index are nearby topics that can swallow a ranking essay if you let them. The crawl section says crawlers respect robots.txt and rate limits, avoid harvesting personal data during crawling, and index publicly visible content and links. That is how pages get into the index. Ranking is how pages already in the index are ordered for a query. A privacy-first crawl is not an infinite index, and a finite crawl is not a promise of coverage for every page on the web. Those are separate journal theses. This one stays on ranking signals.

Likewise, the How Search works guide is not the results page. The guide is publisher documentation on www. The results live on Search. Confusing the two produces empty claims about “checking ranking live on the guide.” You check the guide for the published factor list. You check Search by searching. Company writing should keep that split boring and clear.

## Aggregated satisfaction without person-level ads

The satisfaction-signals line is where skeptics rightly press. Any click-through or bounce metric can be abused into tracking if a company wants to. Oernoe’s published constraint is aggregation and anonymization, plus the broader claim that Search does not use click or dwell signals to build an advertising profile of you. Those sentences sit next to each other in the guide for a reason. Aggregate quality feedback can exist. Person-level ad profiling from Search clicks is what the product writing refuses.

That refusal aligns with the commercial line repeated across About and Privacy: Oernoe does not sell account or search data, and does not use search history to create advertising profiles. Ranking quality and advertising profile are different jobs. Ranking can care whether pages satisfy people in aggregate. Advertising profiles care who you are across sessions for targeting. The company claim is that Search ranking is doing the first job without the second.

Selected publisher pages on www may still show Google ads, measured under Google’s policies. That is a publisher-site fact, not a Search ranking input. Ranking signals about pages on the open web are not the same system as AdSense on a guide. Mixing them produces the false story that either ads prove ranking tracks you personally, or ranking privacy proves the publisher never shows ads. Neither follows.

## Why page-centric ranking is a company stance

Calling this a Company essay matters. Privacy essays often explain what happens to a person’s data. Company essays explain what the organization says it is building. Oernoe’s published ranking story is an organizational choice: compete on page quality signals, not on a personal dossier. Angel’s standing test on About—whether a feature gives people more control over their data or less—fits here. Silent ranking personalization from history and demographics would be less control. Explicit preferences you can delete are more control. Page-centric defaults are the middle that does not require inventing a profile to ship results.

The algorithm description also says it analyzes hundreds of signals. The public list is the primary set, not a claim that the only seven bullets exist in code. What the company will stand behind in journal form is the category split: page and query signals yes; personal advertising-profile signals no. If that split ever changes, the guide has to change in public. Until then, editorial should not soften the negative list to sound more “smart,” and should not harden aggregate metrics into a promise of zero measurement of any kind.

## Reading the ranking block without mythology

A careful reader should take three sentences home. One: ranking looks at how pages relate to queries and to each other. Two: ranking does not use your search history, location, device type, age, or interests as a personal profile. Three: saved preferences are explicit exceptions you control, not inferred identity. Everything else in the guide—crawl limits, HTTPS, referrer stripping, ads on www—supports adjacent claims and belongs in other essays. Keep those adjacent claims in their own shelves so this ranking list remains readable as a company stance rather than a dump of every privacy sentence on the site.

Mythology creeps in when writers turn “treats every search equally” into “knows nothing about the request.” A search system still sees the query string you just typed. It still retrieves candidate pages. It still scores them. Equality means the scoring does not rearrange itself around a secret biography of you. It does not mean the engine is blind to the words in the box.

Mythology also creeps in when “aggregated anonymized satisfaction” is either deleted from the story or inflated into personalization. Keep the words the guide uses. Process at scale. Do not tie to individuals. Do not build an advertising profile from clicks and dwell. Those constraints are the company line.

## Editorial close

Ranking signals are about pages, not people. That is the shortest accurate summary of the Privacy-Preserving Ranking Factors section as published on 2026-09-08. Content relevance, authority and expertise, link analysis, freshness, aggregated anonymized satisfaction, mobile-friendliness, and page speed describe pages and how people in aggregate experience them. Search history, location, device type, age, and interests-as-profile are named as absent. Equal treatment is the default; saved preferences are the explicit override. The crawl is not infinite, the guide is not the results page, and preferences are not an ad profile—those neighboring claims stay neighboring. This Company shelf entry exists so the ranking list itself has a meaty, grounded restatement instead of a slogan.