Content analysis is public page text

When How Oernoe Search works describes content analysis, it is describing a narrow job. The crawler fetches a public page. The indexer reads what a person would see on that page: the headings, the paragraphs, the lists, the structure that ties those pieces together, and the links that point outward or inward. That work exists so ranking can answer a simple question later: does this page look relevant to the words someone typed? It is not a second job of vacuuming private identity data into an advertising file.

This essay stays on that line. It does not re-argue continuous crawl as complete coverage. It does not re-argue that privacy-first crawl is an infinite index. It does not re-argue that technical logs are an ad graph. It does not re-argue content relevance as the whole ranking story. Those pieces already exist. Here the subject is smaller and sharper: what “content analysis” is allowed to mean on a public page, and what it is forbidden to become.

## What the guide actually says

Under Web Indexing and Crawling, the How Search works guide lists four crawl behaviors. Continuous crawling keeps the index from going stale. Respecting robots keeps the crawler polite. Privacy-first crawling refuses to store identifying device or network fingerprints as part of the crawl itself. Content analysis sits in that list as the fourth bullet: analyze page content, structure, and links to understand relevance.

The next paragraph is the restraint. Unlike engines that treat a crawl as a harvest of personal material, Oernoe Search does not extract contact addresses for advertising lists, track user identities, or store behavioral signals during that crawl. The index is built from publicly visible page text and from how pages link to each other. That sentence is the whole thesis of this essay.

Publicly visible is doing real work. A page that requires a login wall is not the same object as a page that anyone can open without credentials. A footer that happens to contain a contact address is still public page text if the page is public; treating that string as a lead for advertising lists is a different act from noticing that the page exists and what it says. The guide draws that line on purpose. Analysis for relevance is not the same pipeline as list-building for ads.

## Content, structure, and links are ranking inputs

Content is the words and media a page chooses to show. Structure is how those words are arranged: titles, sections, lists, tables, the order that helps a reader (and an indexer) understand what the document is about. Links are citations and pathways: which other public URLs this page points to, and which pages point back. Together those three inputs help ranking decide whether a result should sit near the top for a given query.

That is a relevance machine, not a surveillance machine. When the algorithm later scores content relevance, it is asking how well the page text matches the query. When it looks at link analysis, it is asking how the public web cites the page. When it notices structure, it is noticing organization that humans already use. None of those steps require building a dossier of the person who typed the query, and none of them require turning a crawl into a marketing contact file.

The guide’s ranking section makes the same point from the other direction. The primary signals are content relevance, authority, link analysis, freshness, aggregated satisfaction signals processed without tying data to individuals, mobile-friendliness, and page speed. The list of what is not used is just as important: search history, location, device type, age, interests, and other personal profile material. Content analysis feeds the first set. It does not feed the second.

## Public page text is not a private store

Oernoe runs more than Search. Health, Chat, Docs, and Drive are workspaces and conversation surfaces. Those products are not the public web the crawler is supposed to vacuum. Search-and-ads says the Search index is the public web plus listings people chose to submit through an approval step. It is not a way to open Health notes, Chat threads, Drive files, or account settings. Content analysis inherits that boundary. If a page is not publicly fetchable as a web document, it is not “page content” for this crawl job.

That matters because people confuse operator proximity with data proximity. Anoepal operates Search and the publisher site on www. Sharing an operator does not authorize a shared advertising file of queries, and it does not authorize treating private product surfaces as crawl targets for relevance scoring. Content analysis stays on public documents. Private workspaces stay private.

The publisher site itself is also not a free pass to rewrite the crawl story. Selected finished pages on www may load Google ads. The Privacy Policy and the Search-and-ads guide disclose that split. Ads on the publisher hostname are not permission for Search crawl to harvest contact addresses into advertising lists. Two systems share an operator. They still do not share that harvest.

## What analysis does not do

Three refusals sit next to the analysis bullet in the guide, and they deserve plain language.

First, no harvesting of contact addresses for advertising lists. A public page may display how to reach an organization. Indexing that the page exists, and that it discusses a topic, is ordinary search work. Scraping every contact string into a sellable list is a different product. Oernoe Search’s published crawl story refuses that product.

Second, no tracking of user identities during crawling. The crawler is visiting pages, not assembling who you are from those visits. Privacy-first crawling already says identifying data is not stored as part of that process. Content analysis does not quietly reintroduce identity tracking under a nicer name.

Third, no storing of behavioral signals for ads as part of the crawl. Behavioral advertising graphs are a separate industry. Search may keep technical and security logs needed to run the service, as the Privacy Policy describes. Those logs are not a license to sell query history, and crawl-time content analysis is not a license to stockpile behavioral advertising signals from page visits.

If a sentence ever blurs those refusals—if “we analyze content” starts to mean “we mined your identity for ads”—the guide is wrong and should be corrected the same day. Journal writing here is supposed to keep the published line honest, not decorate it.

## Relevance is query-to-page, not person-to-profile

Content analysis exists so ranking can compare a query to a page. That is a document problem. Personalized search that changes results based on past behavior, location, and profile is a person problem. How Search works says the algorithm treats every search equally unless the account holder saved explicit preferences. Two people who type the same phrase get the same results under that rule. Content analysis supports equal treatment by keeping the scored object as the page, not the dossier.

Saved preferences are a different mechanism. When someone is signed in, they can save searches, bookmark sources, set language or region options, and create filters. Those choices are explicit and under the account holder’s control. They are not the crawler learning interests from public pages and smuggling that learning into an ad graph. Content analysis still reads public page text. Preferences still live in account settings the person can delete.

This is also why content analysis is distinct from the ranking factor named content relevance, even though the words overlap. Relevance is the later comparison between query and page. Analysis is the earlier step of understanding what the page contains and how it is built. Confusing the two turns a crawl description into a slogan about quality. Keeping them apart keeps the engineering claim checkable.

## How a reader can check the claim

You do not need a private briefing to test what we publish. Open How Oernoe Search works and read the crawling section. Confirm the content analysis bullet sits next to respecting robots and privacy-first crawling. Confirm the following paragraph limits the index to publicly visible content and link structure, and that it refuses advertising-list harvests, identity tracking, and behavioral signal storage for ads.

Then open Search queries and Google ads. Confirm that Search is at search.oernoe.com, that the index is described as the public web plus approved listings, and that private product surfaces are named as out of scope. Open the Privacy Policy for how technical logs and advertising disclosures are worded on the publisher site. If those pages disagree with this essay, this essay is wrong.

What we publish adds the editorial rule: journal pieces have to add a fact, a method, or a correction you cannot get by swapping a logo onto a template. The fact here is modest. Content analysis, in Oernoe’s own guide, means reading public page text, structure, and links for relevance. It does not mean building advertising lists from contact strings, tracking identities, or storing behavioral advertising signals during the crawl.

## Why the distinction keeps needing repetition

People hear “analysis” and imagine the worst version of the word. In consumer software, analysis often means profiling. In search infrastructure, analysis often means parsing a document. Those are different industries wearing the same noun. Oernoe’s guide chooses the document meaning and then lists the refusals so the noun cannot drift.

Repetition is not marketing filler. It is maintenance. Older homepage language mashed Search privacy and publisher advertising into one careless sentence. Search-and-ads exists because that mash was dishonest. Content analysis needs the same hygiene. If the crawl story ever starts sounding like a brochure about total insight into users, the fix is not a louder slogan. The fix is to return to the bullet list: content, structure, links, public visibility, no advertising-list harvest, no identity tracking, no behavioral ad signals from the crawl.

Authority and expertise still matter for ranking. Freshness still matters. Aggregated satisfaction signals still matter when they are processed without tying data to individuals. None of that rewrites content analysis into a private dossier. The page remains the object. The query remains the question. The person remains outside the advertising profile that Search refuses to build from search history.

## What this essay is not claiming

It is not claiming Oernoe Search indexes the entire web. Continuous crawl is a schedule of revisits, not a promise of complete coverage. It is not claiming robots.txt is a cover story; politeness rules are a separate essay. It is not claiming technical logs do not exist; they do, for abuse and uptime, and they are not an ad graph. It is not claiming the publisher site never shows ads; selected finished pages may, and that is disclosed.

It is also not claiming that every string on a public page is morally free to exploit. Public visibility makes a page eligible for ordinary indexing. It does not create a new right to turn contact addresses into advertising inventory. The guide’s refusal is the published behavior we will stand behind.

## Closing the loop

Content analysis is public page text, plus structure, plus links, scored later for relevance to a query. That is the whole honest sentence. Everything else—private workspaces, advertising lists, identity tracking, behavioral ad signals—belongs to other systems or to refusals. Keep the noun small. Keep the object public. Keep Search’s ranking about pages matching words, not about people becoming profiles.