Same algorithm for every person

How Oernoe Search works states the rule without decoration: the ranking algorithm works the same way for every person. Two people who type the same query get the same results unless an account holds custom preferences that person saved. That sentence is the whole subject of this journal note. It is not a soft slogan about fairness. It is an engineering claim about what ranking is allowed to read when it sorts pages.

Personalized search, as the guide contrasts it, changes results based on past behavior, location, and profile. Oernoe Search refuses that path. The algorithm analyzes the query, searches the index for candidate pages, evaluates those pages with ranking factors, orders them from most to least relevant, and returns the list. None of those steps attaches a personal dossier to the query. The scored objects are pages. The querier is not a ranking feature.

This essay stays on equal treatment of queriers. Neighboring pieces already cover related territory: equal results unless you chose otherwise, ranking signals about pages rather than people, ordinary search without an account, saved preferences that are not an advertising profile, and the refusal to claim Search is a clone of another engine. Those URLs exist. This one is narrower. It asks what “same algorithm for every person” means in practice, how to test it, and how not to overread it into a promise we do not make.

## What the guide puts on the page

Under How Ranking Works, the guide lists five steps. Analyze the query. Search the index. Evaluate each page. Rank results. Return them instantly with no personal data attached. The paragraph that follows is the line we are defending here: the algorithm works the same way for every person. The pizza example is concrete. You and a friend search for the same phrase. You get the same results unless saved custom preferences sit on an account.

Earlier in the same section, the guide says the algorithm does not incorporate search history, location, device, or personal profile the way personalized engines do. Every search is treated equally. That is the default. Preferences you chose and saved are the exception, and they are labeled as your exception, not as a hidden learning loop.

The ranking factors themselves reinforce the same boundary. Content relevance asks how well page text matches the query. Authority asks how trustworthy the site looks. Link analysis asks how the public web cites the page. Freshness, mobile-friendliness, and page speed are properties of the document. Aggregated satisfaction signals, when used, are processed at scale without tying data to individuals. The guide then lists what is not on the board: your search history, your location, your device type, your age, your interests, or other personal data used as ranking fuel.

If equal treatment failed, those “not on the list” items would quietly reappear as ranking inputs. The published story refuses that failure mode. Journal writing here is supposed to keep the refusal honest, not decorate it.

## Equal treatment is not identical life experience

Two people can still leave a results page with different feelings. One person may know a domain and click it first. Another may open three tabs and bounce. Equal ranking does not equalize curiosity, literacy, or taste. It equalizes the ordered list for the same query under the same preference state.

That distinction matters when someone complains that Search “feels personalized” because a friend recommended a different click. The friend personalized advice. The result list did not. The algorithm did not rearrange itself because one of you lives in a different city or typed a related phrase last week. If preferences were saved on one account and not the other, the lists can diverge for that reason alone. Divergence without a saved preference is a bug report, not a feature brochure.

Equal treatment also does not mean every index is equally thick for every topic. A rare phrase can return a thin set for everyone. Thinness is an index and discovery problem. It is not proof that one querier was profiled and another was not. Continuous crawl can thicken a topic over time. It still will not invent a private ranking branch for a person.

## How this differs from preference-driven change

Saved preferences are real. When you are signed in, How Search works says you can save searches, bookmark sources, configure language or region options, and create filters that block domains or content types. Those controls are explicit. You decide what to keep. You can delete them through account settings. They are not inferred interest graphs built from clicks.

The same-algorithm rule and the preference exception must stay in the same paragraph in a reader’s head. Without preferences, results match across people for the same query. With preferences, results can change because you asked them to. That change is still not past-behavior personalization. It is a filter or setting you can see and remove. Essays about saved preferences as advertising profiles, and about filters as your choice, cover the preference side in more depth. This essay only needs the boundary: preferences are the documented exception to sameness, not a license to rebuild a hidden profile under another name.

Search without an account remains ordinary. You do not need a login to run a normal query. An account is what lets you keep the preferences you chose. Absence of an account does not mean a weaker algorithm. It means the preference exception is empty, so the equal-treatment default is the whole story for that session.

## What ranking is allowed to notice about a page

Equal treatment of people does not flatten judgment about documents. The algorithm still prefers a page that answers the query over a page that only mentions the words in passing. It still notices authority, citations, freshness where the topic needs it, and ordinary quality signals such as mobile layout and load speed. Those judgments are about the page in the index.

They are not judgments about the person who typed the query. A page does not rank higher for you because you clicked similar pages yesterday. A page does not sink for you because your device looks like a phone in a certain region. The guide’s “what is not on this list” paragraph exists so that sentence can be checked. If product behavior ever contradicts it, the guide needs a correction the same day, and this journal note becomes wrong until rewritten.

Aggregated satisfaction signals deserve a careful reading. The guide allows anonymized, large-scale signals that are not tied to individuals. That is not the same as building a per-person advertising profile from dwell time. Scale without identity is the published constraint. Equal treatment of queriers depends on that constraint holding.

## How to verify the claim without a private briefing

Open How Oernoe Search works. Read The Search Algorithm and How Ranking Works. Confirm the same-way-for-every-person sentence and the pizza example. Confirm the list of ranking factors and the list of personal inputs that are excluded. Open the Personalization Without Tracking section and confirm that saved controls are described as explicit and deletable.

Then open Search queries and Google ads. Confirm that Search at search.oernoe.com does not use query history to build an advertising profile and does not sell account or search data to advertisers for their own marketing. Confirm that the publisher site on www.oernoe.com is a different surface where selected finished pages may load Google ads. Equal ranking of queriers is a Search claim. Publisher advertising is a www claim. Mixing them produces the slogans we already had to take down.

Open About and the homepage directory if you need the operator story. Anoepal operates Oernoe. Angel Mejia Rodriguez founded the company. Live services are listed there. This essay does not need to sell those products. It only needs the ranking rule to stay checkable against public pages.

The Privacy Policy adds the retention language: search queries are not used to build an advertising profile, and technical or security logs needed to run Search are not a license to sell query history. Equal treatment of queriers sits comfortably next to that retention line. A system that ranked by personal advertising profile would need a different privacy page.

## Failure modes this claim does not cover

Same algorithm for every person does not mean every other engine ranks the same way. It does not mean our index matches another index in size. The FAQ in How Search works already refuses the clone-and-size-parity story. We stand behind a narrower difference: no personal advertising profiles for ranking, and disclosed ads on selected publisher pages.

It does not mean results never change over time. Pages enter and leave the index. Freshness moves. Authority signals shift when the public web cites differently. Two people searching a month apart can see different lists because the index changed, not because one person was profiled. Sameness is across people at a moment, for a query and preference state, not across calendar months.

It does not mean spam never appears. Equal treatment can still surface a thin page if the index holds it and ranking has not yet demoted it. Fixing spam is a quality problem. It is not solved by secretly personalizing one querier away from the mess while leaving another in it.

It does not mean preferences cannot be buggy. If a filter you saved cannot be found or deleted, that is a product defect, not proof that equal treatment was a lie. Using an Oernoe account says the control should be visible in account settings. Invisible controls are bugs.

## Why the sentence keeps needing its own URL

Reviewers and readers mash “private search” into “magically different results for me.” Some mash it the other way into “the company must be lying because my friend clicked something else.” Both mashes skip the published mechanism. The mechanism is boring on purpose. Score pages. Do not score people. Allow explicit preferences. Delete them when asked.

Journal space on www is for finished original writing that adds a fact, a method, or a correction. The fact here is the guide’s own line: ranking works the same way for every person, with equal results for the same query unless saved custom preferences apply, and without personalization by past behavior, location, or profile. Method is how to read that line next to ranking factors and the ads split. Correction is refusing to inflate the line into size parity, identical click paths, or an advertising-free network claim we do not make.

Selected finished publisher pages may show Google ads. Search still does not sell queries. Those two sentences can both be true. Equal treatment of queriers does not require pretending the publisher hostname has no advertising systems. It requires that ranking not read an advertising profile of the person who typed.

## Closing

If you remember one claim from this essay, remember this: the same algorithm for every person means the ordered results for a query do not rearrange themselves because of who you are, where you sit, what device you hold, or what you searched last week. Preferences you saved can change what you see because you asked. Pages can change because the index changed. People do not get a private ranking branch. How Oernoe Search works already said that. This journal note exists so the sentence has a durable home that does not collapse into neighboring essays about signals, accounts, or clone claims. Check the guide. Check the ads split. If product behavior drifts, correct the pages. Until then, sameness across people remains the default we publish.