# Semantic content networks: turning a map into linked pages

A topical map tells you which pages to make. It does not tell you what to put on them, where each link goes, or which pages go live together. That step has a name, Koray Tuğberk Gübür's semantic content network, and this lesson walks through it on one page you can see built.

By Kevin Jeppesen, AI growth engineer. Founder, FrameworkSEO & Ascendily. Updated 2026-11-13. Source: https://frameworkseo.com/learn/semantic-seo/semantic-content-network

**Semantic content network:** A semantic content network is a topical map built into real pages, where each page's angle, headings, order of ideas and links follow the map so the site reads as one connected body of knowledge.

## Key takeaways

- A topical map decides which pages exist; the semantic content network is those pages built, down to every heading, link and link position, which is how Koray Tuğberk Gübür defined it in January 2022.
- Every page needs one main subject that its headings carry down to each section: Google has said it ranks single passages, and a Google patent describes scoring an answer passage by the headings above it.
- Put each internal link inside the section whose small angle is the linked page's whole subject, and stop there: that overlap is both the bridge between the two pages and the border of the first one.
- A link's words, the sentence around it and its place on the page all carry meaning: a 2010 Google patent lists position, footer, sidebar and the words before and after a link as link features.
- Koray launches networks as groups of pages, from 15 at a time to nearly 400 in a day, so each new page arrives with its neighbors and links in place; his results are his own reported numbers.

## A topical map is the plan, the network is the pages

A semantic content network is a topical map built into real pages. The map decides which pages exist, what each one answers and what it links to. The network is those pages written so that each one's angle, heading order, order of ideas and links follow that plan.

If you are new to the plan itself, start with [what a topical map is](https://frameworkseo.com/learn/topical-maps/what-is-a-topical-map). This lesson starts where that one stops: one row of a map, and what it takes to turn it into a page that belongs to the rest.

### Where the term comes from

Koray Tuğberk Gübür named it. His case study [Creating Semantic Content Networks with Query and Document Templates](https://www.oncrawl.com/on-page-seo/creating-semantic-content-networks-with-query-document-templates-case-study/) went up on Oncrawl on 11 January 2022. It followed two Turkish sites, an education institute and a new visa application site, and on purpose left out backlinks, technical work and design changes, so the content network was the only thing being tested. I learned this method from Koray, and his full framework is in [Koray Tuğberk Gübür's topical map framework](https://frameworkseo.com/learn/semantic-seo/koray-topical-map-framework).

His short definition is "a content network that has been prepared based on the semantic network components and understanding". The useful one comes at the end of the same article. There he calls it "the sum of topical map, and individual content designs that include all of the headings, questions, heading levels, anchor texts, content hierarchy, and positions within the site-tree".

> **Wider than a template:** One newer explainer shrinks the idea to "one template per entity type", meaning one template for each kind of page. Templates are part of it. Koray's own definition is wider: every heading, every anchor and every link's position counts. Two pages built from the same template can still form no network if their links sit in the wrong places.

### What the map decides, and what the build decides

The clearest way to see the difference is one row. In the topical map lesson, a made-up coffee gear store has a row for a comparison page, "Burr vs blade grinders", in the cluster under its hub page, Grinders. The row holds four decisions. Building the page adds six more.

![A made-up store's map row for a grinder comparison beside what building the page adds: angle, headings, links, stops, batch.](https://frameworkseo.com/learn/semantic-content-network/map-row-to-page.webp)

*The map row decides that the page exists and what it links to. Building it decides everything a reader and a search engine actually see.*

| Decided in the map | Decided when the page is built |
| --- | --- |
| The search the page answers | The angle it takes on that search, and the order of its headings |
| Its page type (here, a comparison) | The shape of each section: verdict, table, list, short parts |
| The pages it links to | Which section each link sits in, and the sentence around it |
| Its place in the publishing order | Which pages go live with it as one batch |
| Its cluster and its neighbors | Where it stops, so it never covers a neighbor's subject in full |

A topic cluster groups pages by subject. A network goes further and plans how each page reads and joins its neighbors. The difference between the two plans is in [topical map vs topic cluster](https://frameworkseo.com/learn/topical-maps/topical-map-vs-topic-cluster).

## Six things every page in the network has to get right

Koray's articles name six things a page in the network has to get right. They are its angle, which angle sits inside which, the shape of each part, how many angles it covers, what it connects to, and the order of its ideas. He calls most of them "contextual" something. I give the plain version first, then his term.

Apart from Koray's own article, none of the guides I read from Google's top results for "semantic content network" on 13 November 2026 names any of the six. Koray spreads them over several articles from 2021 to 2023. In his January 2023 article on topical maps he lists four of them together as the parts of a brief for one piece of content: "contextual vector, hierarchy, connection, and structure".

| Part (Koray's term) | In plain words | On the made-up grinder page | Source |
| --- | --- | --- | --- |
| Angle (contextual vector) | The one way the page looks at its topic | Which grinder to buy, as a buyer. Not how grinders work, not how to clean one. | Oncrawl, 2021 |
| Nesting (contextual hierarchy) | Which angle sits inside which | Espresso, pour-over and French press sit inside "which grinder suits each brew" | Holistic SEO, 2021 |
| Shape (structure) | The format of each part, and where it goes | The verdict in two sentences first, then a table of five differences | Oncrawl, 2021 |
| Coverage (contextual coverage) | How many angles of the question you answer | The same question asked "for" each brew, and over years of use | Oncrawl, 2022 |
| Connection (contextual connection) | Which pages it joins, and why | Up to Grinders, across to two neighbors, over to the Burr grinders collection | Holistic SEO, 2021 |
| Order of ideas (contextual flow) | How each part follows from the one before | How they break beans, then evenness, then brews, then cost, then care | Holistic SEO, 2023 |

### Angle and nesting

Koray's 2021 case study puts the angle simply: "contextual vectors are the signals to determine a content's angle". His example is the topic "earthquake", which can be handled as comparing earthquakes, predicting them or their history. Each of those is a different page.

Nesting is his "contextual hierarchy", the order in which angles sit inside each other. In his [topical authority article](https://www.holisticseo.digital/theoretical-seo/topical-authority) of February 2021, a search for a flight ticket splits into holiday, business and price comparison. On our grinder page, the three brews are parts of one question. They are not three questions standing beside "what each grinder costs".

### Shape and coverage

Shape comes through in Koray's guidelines for headings, which he calls heading vectors, in [his first Oncrawl case study](https://www.oncrawl.com/technical-seo/importance-topical-authority-semantic-seo/), from 11 February 2021. "Heading vectors start from the title tag", and "every heading's content should have the right format with lists, tables, descriptive definitions". A comparison gets a table. Steps get a numbered list.

Coverage is about angles, not about more topics. In the 2022 article Koray defines it as "the processing angles of a concept" and points to small words he calls context qualifiers: "for, in, at, during, while". "Fruits for children with insomnia" and "fruits for children with anxiety" are one concept seen two ways. For a shop owner, coverage means asking your page's question with the words your customers add to it.

### Connection and order of ideas

Connection, which the 2021 article calls topical connections, is "how the contents of a website are in which subject and how they are connected to which subject". On the grinder page that is four links, each with a reason. The sections below show where each one goes.

The order of ideas is the hardest one to see. In his [semantic search article](https://www.holisticseo.digital/seo-research-study/semantic-search) of January 2023, Koray says the verbs in your sentences "shape the contextual flow of the document": the verbs carry the meaning from one sentence to the next. My way to check it across a whole page, not his rule: each section should open on the subject the last one closed with. Read only the first sentence of every section in a row. If they tell one story, the flow works.

All six depend on knowing why the site exists, which Koray calls the [source context](https://frameworkseo.com/learn/semantic-seo/source-context-central-entity). The wider practice they belong to is [semantic SEO](https://frameworkseo.com/learn/semantic-seo/what-is-semantic-seo).

## One page has one main subject and several smaller angles

Every page in the network has one main subject, which Koray calls its macro context. Inside it are smaller angles, his micro contexts, usually one per section. Some of those smaller angles only exist to lead the reader on to a neighboring page.

He explained it in a reply to a reader under his [article on initial ranking](https://www.holisticseo.digital/theoretical-seo/ranking). The macro context "represents the main purpose of the specific web page", and "the web page still stays in the same macro context, but the micro contexts change from one attribute to another". In his [lexical semantics study](https://www.holisticseo.digital/seo-research-study/lexical-semantics) of November 2022, he added the part that matters most for a network. A macro context can hold micro contexts "in the same document, or within a connected document".

On the grinder page the main subject is which grinder to buy. The smaller angles are how each one breaks the beans, how even the grind is, which brews each suits, what each costs and how to keep each working.

### Why the main subject has to reach every heading

Google does not only rank whole pages. On 15 October 2020, at its [Search On event](https://blog.google/products-and-platforms/products/search/search-on/), Google said it could now "better understand the relevancy of specific passages", and that this would improve 7 percent of searches across all languages. One section of your page can be the result for a search the page as a whole does not target.

A separate idea makes the headings above a section part of the section. In 2018 Google was granted a patent for picking answer passages, [Context scoring adjustments for answer passages](https://patents.google.com/patent/US9959315B1/en) (US9959315B1). It describes scoring a candidate answer with a "heading vector that describes a path in the heading hierarchy from the root heading" down to the passage. A patent shows what Google's engineers designed, not proof it runs today. Still, it is the clearest public reason to carry the main subject down every level.

Put the two ideas together: the path of headings above a section is how its small angle stays attached to the page's main subject. A heading like "More about it" cuts that path. "Which grinder gives you an even grind" keeps it.

> **A one-minute test:** Read each h2 (section heading) with only the h1 (the page's main heading) above it and nothing else. If the h2 still says what it is about and how it belongs to the page, the path holds. If it needs the paragraph before it, rename it.

### Mention, or cover

A small angle on your page is often the whole subject of another page. On the grinder page, the grind size for espresso is one short part. On the neighboring page, "Best grind size for espresso", it is everything. That overlap is where the two pages meet, and it decides two things at once.

![Two made-up coffee store pages: espresso grind size is a short section in one and the whole subject of the other.](https://frameworkseo.com/learn/semantic-content-network/where-two-pages-meet.webp)

*The overlap between two pages is where the link goes, and where the first page stops.*

## A link belongs where one page's side angle is another page's subject

Put each link inside the section whose small angle is the other page's whole subject. In that section, mention the subject in a sentence or two and link. Then stop. Koray calls the link's reason a contextual bridge, and the place where a page stops a topical border. The bridge and the border are the same overlap, seen from two sides.

### What a contextual bridge is

Koray defines the bridge in his [January 2023 article on expanding topical maps](https://www.holisticseo.digital/seo-research-study/topical-map). It is the connection between two side topics that explains "the navigation purpose of a possible link". It "should come from the 'source context' and 'contextual domain' intersection". In plain words, the link has to make sense both for what your business does and for the subject at hand. Without it, he writes, "the internal links can't pass relevance".

Google described the same idea from the engine's side. Its [Reasonable Surfer patent](https://patents.google.com/patent/US7716225B1/en) (US7716225B1), filed in 2004 and granted in 2010, models how likely each link is to be clicked. One example rule says a link has "a higher probability of being selected" when the linking page's topic "is related to a topical cluster associated with the target document". Like any patent, it shows what Google designed, not what runs today. A bridge puts that kind of relation into words a reader can see.

### Where a page stops

A topical border is where one page's job ends and the next page's begins. In 2021 Koray tied borders to one decision: whether "you should open a new page or not". Get it wrong one way and a page keeps growing into its neighbor's subject, so two pages compete for one search. That is [keyword cannibalization](https://frameworkseo.com/learn/topical-maps/keyword-cannibalization).

Get it wrong the other way and you open pages for every small question. Koray names his own client here. In the January 2023 semantic search article he writes that Vizem.net, the visa site from the original 2022 case study, later "distorted the topical map by opening web pages for micro-queries by diluting the Ranking Signals, and dividing PageRank". The rule for when a search deserves its own page is in [when a search deserves its own page](https://frameworkseo.com/learn/topical-maps/page-or-section).

### Three outcomes for any subtopic on a page

| The subtopic is | On this page | Link | Grinder page example |
| --- | --- | --- | --- |
| Part of this page's own subject | A full section | None needed | Which grinder gives you an even grind |
| Another page's whole subject | A sentence or two, then stop | Yes, inside that sentence | The grind for espresso, linked to Best grind size for espresso |
| Nobody's subject yet, but readers ask | A sentence or two | None yet; note it as a gap in the map | Any question the map has no row for |

The third row is what Koray calls a topical gap: "a contextual disconnectedness between different topics". In one 2023 case study note he says a new area of a map should have "at least 3 side contextual bridges" to what you already have. A gap with no bridges is probably outside your site's subject.

Bridges also run between the pages that sell and the pages that explain. On the grinder page, the cost section links over to the Burr grinders collection. How those two halves of a map support each other is in [core and outer sections](https://frameworkseo.com/learn/semantic-seo/core-and-outer-sections).

## Where a link sits and the words around it carry meaning too

A link tells a search engine three things: its words (the anchor text), the words around it, and where on the page it sits. The same anchor inside the section it belongs to, in a sentence that explains it, says more than the same words in a footer.

### The words, and the sentence around them

Anchor text has mattered since the start. In their [1998 paper on Google](https://research.google/pubs/the-anatomy-of-a-large-scale-hypertextual-web-search-engine/), Sergey Brin and Lawrence Page wrote that anchors "often provide more accurate descriptions of Web pages than the pages themselves". Google's current [link best practices](https://developers.google.com/search/docs/crawling-indexing/links-crawlable) page goes one step further: "the words before and after links matter, so pay attention to the sentence as a whole".

Google's internal API documentation, leaked in 2024, has a record for each link called [AnchorsAnchor](https://hexdocs.pm/google_api_content_warehouse/0.4.0/GoogleApi.ContentWarehouse.V1.Model.AnchorsAnchor.html). It lists fields for the anchor words, for the words left and right of the link ("fullLeftContext" and "fullRightContext", noted as not written to the link logs), a hash of "terms near the anchor", and a flag for links within the same domain. The fields exist. Whether and how much each one is used is not known.

Koray planned that sentence on purpose. In the 2022 case study he writes that around each anchor "a planned 'annotation text' has been used to improve the precision of the purpose of the link". Usability research has a name for the reader's side of it. Nielsen Norman Group calls it [information scent](https://www.nngroup.com/articles/information-scent/): "a mix of cues that they get from the link label, the context in which the link is shown, and their prior experiences". It builds on Peter Pirolli and Stuart Card's [information foraging theory](https://psycnet.apa.org/doi/10.1037/0033-295X.106.4.643) of 1999. The writer's annotation, the reader's scent and the engine's left and right context are one sentence. Write it first, then pick the anchor words.

### The place on the page

The Reasonable Surfer patent lists where a link sits as a feature: "in a HTML list, in running text, above or below the first screenful", "in a footer, in a sidebar". It also lists "the context of a few words before and/or after the link". One of its example rules is that "links positioned closer to the top of a document have a higher probability of being selected than links positioned toward the bottom".

![A page outline marking six places a link can sit, from header to footer, and what each place tells.](https://frameworkseo.com/learn/semantic-content-network/link-position.webp)

*The same anchor words mean more inside the section they belong to, with a sentence around them.*

Koray's published rules turn this into placement. In the 2021 case studies he kept every page to at most 15 links, "a maximum of 1 link in each heading section", mostly in the main content. Links to a side topic went in "one of the last heading's paragraphs". He never used the same anchor more than three times on a page, and checked that the anchor words appear on the page being linked to. In the 2021 topical authority article he rates single in-text links as having "the highest contextual relevance between web pages", meaning single links in the text with no similar links around them. He also says plainly that these were examples from his projects, not laws.

By the 2022 case study his practice had changed. The 2021 sites had no header or footer menu at all. In 2022 a network's most important sections were linked from the header in short words and from the footer in longer, different words. No fixed sidebar of the same links repeats across every network. Both sites stayed under 150 internal links a page. How many links a page should carry, the studies on anchor wording and how to test a linking change are all in [internal links inside a topical map](https://frameworkseo.com/learn/topical-maps/internal-links-in-a-topical-map).

## Why networks go live in connected batches

Koray launches a semantic content network as a group of pages, not one post at a time. When the network is on one topic and the site has no history there, he publishes most of it at once. When it is not, he starts small and raises the pace. Either way, each new page should arrive with its neighbors and its links in place, instead of alone weeks before them.

### What Koray published, and when

The 2022 Oncrawl case study gives two very different launches. On one project he published "nearly 400 pieces of content in one day". On the other he started with "only 10-15 content suddenly" and then raised the pace steadily. His reasoning: when a network sits on one topic and one search pattern, and the site has no history on that topic, publishing most of it at once brings "faster indexing, understanding and re-ranking". On the education site there was also a seasonal event, and he published the first network 45 days before it.

In January 2024 he wrote up a smaller launch on Medium, [How to Launch a Semantic Content Network?](https://medium.com/@ktgubur/how-to-launch-a-semantic-content-network-7b527acddced) "We published only 15 new web documents," he writes, with a chart captioned "100% Click Growth for the first 28 days". The network started from the root of the topic, because the company already ranked as a "Church Management System". A network's shape, he says, "must change" with the site's topic priorities, history, PageRank and brand.

> **His numbers, not a test:** These are Koray's reported results from his own screenshots, with no control site, and some overlap with Google updates and seasons he names himself. Read them as a practitioner's pattern, not proof.

### Why a batch helps a new page

Google's own slides, made public in the US antitrust case, give a reason. A 2017 Google deck shown as court exhibit [UPX0004](https://www.justice.gov/atr/media/1322476/dl?inline) lists "The 3 Pillars of Ranking". They are body ("What the document says about itself"), anchors ("What the Web says about the document") and user interactions ("What users say about the document"). A December 2016 all-hands deck, [UPX0203](https://www.justice.gov/d9/2023-11/417516.pdf), admits that Google's understanding fails "maybe because a document is new, recently-changed, or rarely-shown".

Here is my reading of those slides, not something they say about batches. Line a new page up against the three pillars. Published with no live page linking to it, it has a body on day one and nothing else. Published with its batch, it also has anchors from the first time Google crawls it (reads it): links from its hub and neighbors, each with a sentence around it. The slide's anchors are mostly what other sites say, but the leaked link record above has a flag for links within the same site, so your own links count as anchors too. A batch gives a new page two of Google's three pillars on its first day. User interactions come later either way. The slides predate BERT and MUM, Google's later language models, so take them as Google's thinking in 2016 and 2017.

![A new page published alone versus in a batch, scored on Google's three ranking pillars: body, anchors, user interactions.](https://frameworkseo.com/learn/semantic-content-network/batch-launch.webp)

*Clicks take time either way. A batch is how a new page gets its anchors on the first day.*

### Sizing a batch by hand

The first point is Koray's. The rest are my rules of thumb built on his launches.

- **Start from the page that already has strength.** Koray's 2024 advice is to grow from the page with more PageRank (link strength), or from a subject where your brand already shows up in Google's search suggestions.
- **Take a hub and every page that bridges to its siblings.** If a page's links all point to pages in the same batch or already live, it belongs in the batch.
- **Never link to a page that is not live.** Add the link on the day its target goes live.
- **Finish what you start.** The education site's third network was never published. In 2023 Koray wrote that this left the site open to being seen as "a half-authority source".

Which batch goes first across the whole map is a separate decision, covered in [publishing order](https://frameworkseo.com/learn/topical-maps/publishing-order).

## One page from a made-up map, annotated from top to bottom

Here is the row from the topical map lesson, "Burr vs blade grinders", built as one page of the network. The store and the page are made up. Every part of this lesson is marked where it happens, from the breadcrumb to the launch.

![A made-up page, Burr vs blade grinders, annotated top to bottom: main subject, angle, headings, links and where it stops.](https://frameworkseo.com/learn/semantic-content-network/annotated-page.webp)

*Every part of the network is visible on one page: what it covers, what it only mentions, and where each mention becomes a link.*

### Where each part shows up on the page

| Part of the page | What it does | Lesson idea |
| --- | --- | --- |
| Breadcrumb: Home espresso, Grinders | Links up to the hub at the top of the page | Connection, position |
| Title and h1: which one to buy for coffee at home | States the main subject and the buyer's angle | Macro context, angle |
| Opening | The verdict in two sentences, then a table of five differences | Shape |
| How a burr grinder and a blade grinder break the beans | Starts with the cause of every other difference | Order of ideas |
| Which grinder gives you an even grind | Opens on the last section's subject | Order of ideas, micro context |
| Which grinder suits espresso, pour-over and French press | Holds the three brews as parts, not as separate h2s | Nesting, coverage |
| For espresso | Two sentences and a link to Best grind size for espresso, then stops | Bridge and border |
| What each grinder costs to buy and to keep | A sentence on who should buy a burr grinder, linking to Burr grinders | Bridge to the page that sells |
| How to keep either grinder working (last section) | Two sentences and a link to How to clean a grinder | Side topic last, border |
| Launch | Live with its two neighbors, once Grinders and Burr grinders are live | Batch |

> **What the page leaves out on purpose:** There is no grind size chart and there are no cleaning steps. Both are other pages' whole subjects. A page that tried to hold them would get longer and weaker, and would compete with the two pages built to answer them.

Notice what the build did not change. The four decisions in the map row all survived: same search, same page type, same links, same launch timing. The build only decided how the page reads and where each of those links sits.

## Check a page against its network by hand

Take one page and its row in the map. Answer 14 yes-or-no questions, one for each idea in this lesson. Every "no" points to a fix on that page, a change to a neighboring page, or a gap in the map.

1. Can you say what the whole page is about in one sentence, and does the h1 say the same thing?
2. Does the page take the one angle its map row gives it?
3. Read each h2 with only the h1 above it: does it still say what it is about?
4. Does every h3 sit inside the h2 above it?
5. Does each section have the shape its question needs?
6. Have you asked the page's question with the words your readers add to it?
7. Does each section open on the subject the section before it closed with?
8. Does every section serve the page's main subject?
9. Does each link sit inside the section whose angle is the linked page's whole subject?
10. Where a section starts covering another page's subject in full, did you stop?
11. Is every link inside a sentence that says why a reader would click?
12. Are the links in the main content, with side topics toward the end?
13. Does the page link to every page its map row names, and to nothing the map does not justify?
14. Will every page it links to be live on the day it goes live?

[Download the page check](https://frameworkseo.com/learn/semantic-content-network/semantic-content-network-page-check.csv) (CSV, 14 rows: the question, Koray's term, what a yes looks like, what to do with a no, and the source)

Two things the checklist cannot do for you. It cannot tell you whether a page says anything new. That is [information gain](https://frameworkseo.com/learn/semantic-seo/information-gain), and every section needs some. It also cannot fix a bad map. If half your answers point to missing or overlapping pages, go back to [how to build a topical map](https://frameworkseo.com/learn/topical-maps/how-to-build-a-topical-map).

You can run these questions with an AI assistant at your side. How to set one up for a page plan is in [content briefs with Claude](https://frameworkseo.com/learn/claude-for-seo/content-briefs-with-claude). On client maps we built our own process for this step. What you see here is the hand version, which is how I learned it. If you would rather start from a finished map with every page and link already planned, that is [the growth plan](https://frameworkseo.com/).

## Sources

- [Creating Semantic Content Networks with Query and Document Templates (case study)](https://www.oncrawl.com/on-page-seo/creating-semantic-content-networks-with-query-document-templates-case-study/), Koray Tuğberk Gübür, Oncrawl, 2022
- [Importance of Topical Authority: A Semantic SEO Case Study](https://www.oncrawl.com/technical-seo/importance-topical-authority-semantic-seo/), Koray Tuğberk Gübür, Oncrawl, 2021
- [Importance of Topical Authority for SEO](https://www.holisticseo.digital/theoretical-seo/topical-authority), Koray Tuğberk Gübür, Holistic SEO & Digital, 2021
- [Importance of Initial-Ranking, Re-ranking, and Historical Data for SEO](https://www.holisticseo.digital/theoretical-seo/ranking), Koray Tuğberk Gübür, Holistic SEO & Digital, 2021
- [Importance of Lexical Semantics and Semantic Similarity for SEO](https://www.holisticseo.digital/seo-research-study/lexical-semantics), Koray Tuğberk Gübür, Holistic SEO & Digital, 2022
- [How to Expand a Topical Map for Higher Topical Authority?](https://www.holisticseo.digital/seo-research-study/topical-map), Koray Tuğberk Gübür, Holistic SEO & Digital, 2023
- [Semantic Search for Semantic SEO: Understanding the Verbs of Life](https://www.holisticseo.digital/seo-research-study/semantic-search), Koray Tuğberk Gübür, Holistic SEO & Digital, 2023
- [How to Launch a Semantic Content Network?](https://medium.com/@ktgubur/how-to-launch-a-semantic-content-network-7b527acddced), Koray Tuğberk Gübür, Medium, 2024
- [How AI is powering a more helpful Google (Search On, passages)](https://blog.google/products-and-platforms/products/search/search-on/), Google, 2020
- [Link best practices for Google](https://developers.google.com/search/docs/crawling-indexing/links-crawlable), Google Search Central, 2025
- [Context scoring adjustments for answer passages (US9959315B1)](https://patents.google.com/patent/US9959315B1/en), Google LLC, Google Patents, 2018
- [Ranking documents based on user behavior and/or feature data (US7716225B1)](https://patents.google.com/patent/US7716225B1/en), Google LLC, Google Patents, 2010
- [AnchorsAnchor (leaked Content Warehouse API documentation)](https://hexdocs.pm/google_api_content_warehouse/0.4.0/GoogleApi.ContentWarehouse.V1.Model.AnchorsAnchor.html), hexdocs.pm copy of the google_api_content_warehouse package, 2024
- [Life of a Click (user-interaction), trial exhibit UPX0004](https://www.justice.gov/atr/media/1322476/dl?inline), US Department of Justice, US v. Google, 2017
- [Q4 Search All Hands, trial exhibit UPX0203](https://www.justice.gov/d9/2023-11/417516.pdf), US Department of Justice, US v. Google, 2016
- [The Anatomy of a Large-Scale Hypertextual Web Search Engine](https://research.google/pubs/the-anatomy-of-a-large-scale-hypertextual-web-search-engine/), Sergey Brin and Lawrence Page, Stanford University, 1998
- [Information foraging](https://psycnet.apa.org/doi/10.1037/0033-295X.106.4.643), Peter Pirolli and Stuart Card, Psychological Review, 1999
- [Information Scent: How Users Decide Where to Go Next](https://www.nngroup.com/articles/information-scent/), Nielsen Norman Group, 2020