Area

Serving and ranking

How Google understands queries, retrieves and ranks results, tests and updates its systems, fights spam, and builds the results page, Discover and its features.

9 topics · 294 claims

Topics in this area 9

35 claims · 5 sessions
Rich results
Day 1 only told attendees to check and focus on rich results; Day 2 supplied the detail: structured data makes pages eligible for rich results such as review stars and recipe filters, which Google says can bring more qualified visitors. Google pointed to case studies on its developer site, including a May 2018 one in which Rakuten Recipes got 2.7 times more search traffic and 1.5 times longer sessions after adding recipe markup; the speaker called it old but expects the link to hold. The Search gallery shows the supported features, from broad page-level types (breadcrumbs, article) to vertical ones (recipes, events, products), and the Rich Results Test shows which features a page triggers and whether its markup is valid. Features come and go: Google removed several little-used ones in 2025, and a structured data manual action removes rich result eligibility without affecting web ranking. Google's generative AI guide names rich result eligibility as a reason to keep using structured data, which AI features do not require. Day 3 placed rich results in serving: Google builds search features as the last step after ranking, most of them from what it extracted during indexing, while rich results rely on extra data from the site owner, usually JSON-LD structured data; AI Overviews and AI Mode are standard search features, not rich results, and need no structured data. Google highlighted review snippets (star ratings, which it called a powerful trust signal for users, said at the event; self-serving reviews of a site's own business are not eligible), product rich results, which it called essential for e-commerce, and Article markup, highly recommended though not required for Top Stories, which can also mark paywalled content. The site name is an attribution feature that owners can influence, and Google said it may never use structured data from a site it does not trust. A Day 1 community demo treated warnings in Google's structured-data test as non-critical, which matches Search Console Help: critical issues make an item invalid for rich results, while non-critical ones only limit its appearance. Google's Rich Results Test help says the tool tests either a page's full URL or a pasted code snippet, and that all page resources must be reachable by an anonymous user, so pages behind a firewall or a password need the code mode or must be exposed, for example through a tunnel; a Day 1 community demo used both modes.
Day 1Day 2Day 3
74 claims · 7 sessions
Query understanding
Google rewrites pretty much every query before looking it up: it detects the language (falling back on location and browser settings for brand-only queries), drops stop words, recognises entities and expands the query with synonyms, so a short query can become a much longer internal one. Its synonyms are learned from search behaviour rather than catalogued by linguists, so words people use interchangeably count as synonyms, while 'siblings' such as Canon and Nikon are kept apart, and spellings with and without diacritics are generally treated as the same. Google's How Search Works page confirms the synonym and spelling systems; stop words, siblings and the rewriting detail were said at the event and are not in Google's docs. Fan-out queries from the AI features pass through the same query understanding, which also predicts which result types a query gets, using historical interaction data for ambiguous queries such as 'orange'. Query processing deliberately mirrors indexing, as Day 2 said of word segmentation, and the diacritics point extends Day 2's statement that Google usually understands words with or without accents; Google's summary slide added that typos and plurals need no attention because Google rewrites them automatically. Day 1's second recording put this at the start of serving: Cherry Prommawin said serving begins by interpreting the query, cleaning it up, detecting its language and expanding it; Gary Illyes contrasted 1990s searchers who had to 'speak machine' with today's natural-language queries, and said MUM helps interpret the context of query words. On Day 2 the non-Latin-script talk added that users may type the same Persian query in their own script or in Latin letters with the same intent, as Google's 2023 post on multilingual searches describes for Hindi.
Day 1Day 2Day 3
16 claims · 4 sessions
Retrieval: finding candidates in the index
After query understanding, Google looks up the expanded query in the index, which stores posting lists: for each word, the URLs associated with it. A document must contain the query's words, or related concepts in the case of embeddings, to be retrieved, and because a query can match millions of pages Google already orders the candidates at retrieval using per-document signals from indexing: language, then country, then quality. Google called quality the most important of these retrieval signals and said the same signal is used again in ranking (said at the event; Google's docs list quality among ranking signals but do not describe it as deciding retrieval). Author’s view: a page judged low in quality may never reach ranking, so quality is a precondition, not a final tweak. This builds on Day 2, which described the posting lists, retrieval through vector embeddings and the per-document signals, quality, country and language among them, attached to each document in the index. Gary Illyes said on Day 1 that for visual search Google breaks an image into vectors and retrieves results from the index with them, and on Day 2 that descriptive text around a video helps Google rank and retrieve it (both said at the event).
Day 1Day 2Day 3
23 claims · 5 sessions
Ranking systems and signals
Google stressed that there is not one single ranking system: its slide listed spam detection, reviews, BERT, MUM, RankBrain, freshness, deduplication, crisis information and other systems, and pointed site owners to Google's published ranking systems guide, which says MUM is not currently used for general ranking. Quality is one of the most important of hundreds of ranking signals, so maximising a single signal does not work, and the systems change over time. The quality talk said PageRank is not used so much anymore, while Google's guide says it has evolved and is still part of the core ranking systems. Google said the same quality signal is also used to order candidates at retrieval (said at the event). Day 1's slides had placed BERT in indexing and RankBrain in serving, and Google's guide already said then that MUM is used only for specific applications. Author’s view: read the slide as a list of systems Google runs, not as proof that each one ranks every query, and read the PageRank remark as links weighing less among many signals, not as PageRank being switched off. Day 1's second recording added that signals calculated during indexing are stored and used both for index selection and later for ranking, that quality is one of the most important things in ranking, and that a higher crawl rate does not make a URL rank better; on Day 2 Gary Illyes said the main content is what Google considers when ranking a page.
Day 1Day 2Day 3
49 claims · 5 sessions
Testing changes and quality raters
Before a change to Search launches it is tested: in side-by-side experiments external search quality raters compare results with and without the change, then live traffic experiments start at about 0.1% of users and are judged on long lists of metrics, and only changes that benefit users become launches. In 2023 Google ran 719,326 search quality tests, 124,942 side-by-side and 16,871 live traffic experiments and made 4,781 launches, as its How Search Works page confirms; a second Day 3 slide gave rounded figures, including 800,000+ tests, which matches no published number. Raters follow the published Search Quality Rater Guidelines but cannot affect individual sites' rankings: their ratings measure how well Search works and become labels used to improve the algorithms. The guidelines define content quality by effort, originality, talent or skill and accuracy and rate any deception, such as fake author personas with AI-generated headshots, as Lowest; Google synced its helpful-content and generative AI guidance with this material on 1 October 2026. Day 2 had already pointed to the rater guidelines as the detailed source on how Google thinks about a page's main content, and Google said on Day 3 that a core update is sometimes coupled with an update to the guidelines (not in Google's docs); Google's 2019 post says broad core updates are tested with rater feedback that is not used directly in ranking. Author’s view: with 4,781 launches in 2023 and only a handful of named updates, almost every change to Search goes unannounced. The opening keynote on Day 1 said even the classic ten blue links were settled only after millions of experiments (said at the event), and a second recording of Day 2 captured Gary Illyes defining main content as any part of a page that directly helps it achieve its purpose, the same definition the rater guidelines use.
Day 1Day 2Day 3
45 claims · 4 sessions
Core updates and spam updates
Google gave three reasons why Search changes: new content formats, the growing breadth of content and content issues such as spam, and every update shares the goal of satisfying users' information needs. Core updates are broad changes made several times a year and announced on the Search Status Dashboard (two in 2026 by 3 October, next to four spam updates); they do not target or punish sites, so a drop means other pages now do better and there is no 'recovery' in the penalty sense. Google advises looking at the whole site objectively, even asking someone outside the team to review it; its docs say improvements may take several months to register, sometimes only at the next core update, and Google estimated on Day 3 that a core update reaches a site within 2 to 4 weeks of its rollout and that recovery typically takes 3 to 6 months. Spam updates target spam only: recovering means removing the kind of spam Google described, and the benefit lost from spammy links in a link spam update cannot be regained. Older named updates now live inside the core systems: Google's ranking systems guide says Panda (2011) and Penguin (2012) became part of the core ranking systems in 2015 and 2016, and the helpful content system (2022) in March 2024. A spam update is a notable improvement to Google's constantly running spam systems such as SpamBrain, which Day 2 called central to Google's spam fighting; Google estimated that a spam update affects sites within 1-2 days of its rollout, and months for batch refreshes, and its spam updates page says a compliant site may improve over a period of months. On Day 2 a community speaker reminded attendees that features and spam improvements often reach some languages first: site names launched in select languages in 2022 and in all languages in September 2023, and the October 2023 spam update improved coverage for specific languages.
Day 2Day 3
33 claims · 5 sessions
Spam policies and manual actions
Google defines spam as content designed to deceive users or manipulate search results. Its Day 3 slide named cloaking, doorways, scraped content, link spam and hacked content, but the speaker said scaled content abuse would replace link spam as the type worth talking about today (said at the event). Recent policy changes covered back button hijacking (announced 13 April 2026, enforced from 15 June 2026), scaled content abuse and spam aimed at manipulating AI responses, which falls under the same policies. Google reads spam reports to decide what to tackle next and may take manual action on them; Google estimated that removing a manual action after a reconsideration request takes 1-2 weeks on average, and its help says reviews take several days or weeks. Scaled content abuse, which Google's policies define as generating many pages mainly to manipulate rankings however they are created, came up on every day: separate pages for fan-out query variations on Day 1, automatically translated scraped content on Day 2 and mass-produced AI slop on Day 3. Google said some sites never get a manual action removed because their owners do not actually clean up. On Day 2 the non-Latin-script talk recalled that Google treats buying links for ranking as spam, yet observed that bought backlinks and paid editorial content still visibly influence competitive Persian, Turkish and Arabic results (an observation, stressed as not a recommendation), and on Day 1 a community speaker warned that the industry is overusing listicles for AI visibility the way it once overused links.
Day 1Day 2Day 3
24 claims · 3 sessions
The results page and search features
Google described its results page as an auction in which results of different types compete for space (a metaphor, said at the event), with text results the most common type. Query understanding decides which result types appear: explicit words such as 'pictures' signal the intent, and for ambiguous queries Google uses historical interaction data. A text result has an attribution (site name, favicon, URL), a title link and a snippet, which Google generates from its understanding of the page, and Google can add badges such as a preferred-source badge. Search features are built as the last serving step, after ranking, and most need nothing from the site owner; rich results are the exception that relies on structured data. Images appear among web results because, Google said, people began adding words such as 'pictures' to queries in 2005-2006, which led to universal search in May 2007 (said at the event); those image results come from Google's image index, and video results take thumbnails and attribution from the page or its markup. The site name is an attribution feature that owners can influence with WebSite markup on the home page. This extends Day 2: feature extraction pulls structured data, images and videos from the page during indexing, and snippets are rebuilt from the stored tokens (said at the event). Day 1's second recording added that after ordering the results Google constructs the page by deciding which search features fit the query's intent, that a majority of search features now use AI (said at the event), and that ads in Search are clearly marked as sponsored. Google's site names, introduced on mobile for select languages in October 2022, became available in all languages in September 2023.
Day 1Day 2Day 3
14 claims · 2 sessions
Google Discover
Google's Discover talk (photographed slides only) showed an eligibility gate: content with policy violations or low quality is filtered out, and safe, trusted content becomes eligible. Google's docs say indexed content that meets the Discover content policies is eligible automatically, with no special markup, and that Discover uses many of the same signals as Search. Quality is judged through E-E-A-T, where trust comes first. For eligible content, image quality drives click-through: Google recommends images at least 1,200 px wide, over 300,000 total pixels and 16x9, enabled by max-image-preview:large, and warns against logos, text-heavy images and misleading or outrageous ones. This extends Day 2, where Google said max-image-preview:large matters mainly in Discover, and Google's Discover content policies ask news sources for clear dates, bylines and author, publisher and contact information. Discover demand cannot be read from Google Trends, which counts only typed searches (Day 2). On Day 2 Gary Illyes repeated that max-image-preview:large can make content perform surprisingly well in Discover.
Day 2Day 3

Across days 43

Stage D2-C452 Day 2 · What is Structured Data and why we need it on the internet.
Search results have moved from ten blue links to feature-rich, media-rich results because users wanted more.
extends
Stage D1-C158 Day 1 · Welcome and opening keynotes
Google said that even the classic ten-blue-links layout was settled only after millions of experiments, as part of using as much data as possible for product decisions.
Stage D2-C521 Day 2 · What is Structured Data and why we need it on the internet.
Structured data makes pages eligible to appear as rich results, Google's structured data talk said in its recap.
extends
Stage D1-C122 Day 1 · session not recorded
Check and focus on rich results for Google.
Stage D2-C687 Day 2 · Deciding what goes in the index?
Index selection uses the signals calculated earlier in indexing for each document it has to select or discard.
extends
Stage D1-C205 Day 1 · How Search works and where's AI?
Signals calculated for a page during indexing are stored in the index and used both to decide whether the page gets indexed and, later, for ranking.
Stage D3-C006 Day 3 · Making sense of users' queries
Google's first step in understanding almost any query is to detect its language, which tells Google roughly what content the user wants: a query in German suggests German content, a query in English English content.
extends
Stage D2-C654 Day 2 · Calculating (some) signals
In ranking, country and language signals help Google serve users the right content for their country and language.
Stage D3-C007 Day 3 · Making sense of users' queries
Query language detection works poorly when someone searches only for a brand name, such as Facebook or Google, because the query does not show which language the user wants results in.
extends
Stage D1-C215 Day 1 · How Search works and where's AI?
Serving starts with interpreting the query, which includes cleaning it up, detecting its language and expanding it.
Stage D3-C011 Day 3 · Making sense of users' queries
Google named Thai as a language that makes query understanding more complex because it does not separate words with spaces; the speaker added, hedging with 'apparently', that Thai uses spaces to separate sentences.
extends
Stage D2-C320 Day 2 · Understanding what's on a page
Text in languages written without spaces, such as Thai and Chinese, would end up in the index as long strings that might never be searched for, so Google segments it into words with statistical models built from other web content in that language.
Stage D3-C013 Day 3 · Making sense of users' queries
Google's query processing deliberately mirrors indexing: a query is transformed into something that can be matched against the index, and stop word removal is part of that transformation.
extends
Stage D2-C321 Day 2 · Understanding what's on a page
For languages written without spaces, such as Thai and Chinese, Google uses exactly the same word segmentation when indexing a page as when interpreting the user's query, because otherwise the query could not be matched against the index.
Stage D3-C013 Day 3 · Making sense of users' queries
Google's query processing deliberately mirrors indexing: a query is transformed into something that can be matched against the index, and stop word removal is part of that transformation.
extends
Stage D2-C737 Day 2 · How does the index look like?
A search query is broken into words with the same segmenter or tokenizer that Google used to build the index.
Stage D3-C015 Day 3 · Making sense of users' queries
Google said the difference between words and entities can be seen in Google Trends, where a term can be searched as words or as an entity (Trends calls these a search term and a topic).
extends
Stage D2-C761 Day 2 · Google Trends
The Google Trends Explore page, which Google called the heart of Trends, shows search interest in a query or topic and how it changes over time.
Stage D3-C048 Day 3 · Making sense of users' queries
Google generally treats spellings with and without diacritics as synonyms behind the scenes, for example a German 'ü' written as 'ü', as 'ue' or left out.
extends
Stage D2-C600 Day 2 · Focusing on Internationalisation and Localisation
Google usually understands a query word whether it is written with or without diacritics (accents).
Stage D3-C059 Day 3 · Making sense of users' queries
Google treats fan-out queries generated by the LLM the same way as queries typed by users, so understanding how normal queries work explains fan-out queries too.
extends
Stage D1-C172 Day 1 · Welcome and opening keynotes
Google said the Gemini model lets Search understand the user's intent, and query fan-out then adds further queries to the first one to enrich the quality of the answer.
Stage D3-C059 Day 3 · Making sense of users' queries
Google treats fan-out queries generated by the LLM the same way as queries typed by users, so understanding how normal queries work explains fan-out queries too.
extends
Stage D2-C727 Day 2 · How does the index look like?
Google said that when AI Overviews or AI Mode run a query fan-out, the generated queries are sent to Google's Search index and documents come back with their snippets, which then feed the AI-generated answer (part of this passage is unclear in the recording).
Stage D3-C075 Day 3 · Making sense of users' queries
At retrieval, Google splits the query into words, applies query understanding and expansion, and matches the words and their expansions against the posting lists.
extends
Stage D2-C736 Day 2 · How does the index look like?
In posting-list retrieval, the posting lists of the query's words are intersected, which yields an unranked list of candidate URLs.
Stage D3-C075 Day 3 · Making sense of users' queries
At retrieval, Google splits the query into words, applies query understanding and expansion, and matches the words and their expansions against the posting lists.
extends
Stage D2-C738 Day 2 · How does the index look like?
At retrieval, Google looks up the posting lists of the query words that are actually important rather than of every word in the query.
Stage D3-C077 Day 3 · Making sense of users' queries
The first condition for retrieving a document is that the query's words, or its concepts in the case of vectors or embeddings, are in the document or related to it.
extends
Stage D2-C740 Day 2 · How does the index look like?
Besides posting lists, Google can retrieve documents through vector embeddings: parts of documents are associated with embeddings, which form a vector space used for retrieval.
Stage D3-C079 Day 3 · Making sense of users' queries
To order candidates at retrieval, Google uses signals collected during indexing, and the first two are language and country.
extends
Stage D2-C649 Day 2 · Calculating (some) signals
Country and language are among Google's most important signals and have been used since Google's early days.
Stage D3-C079 Day 3 · Making sense of users' queries
To order candidates at retrieval, Google uses signals collected during indexing, and the first two are language and country.
extends
Stage D2-C722 Day 2 · How does the index look like?
Each document in Google's index has pretty much all the signals calculated for it attached, for example quality signals plus the page's country and language, according to an illustration the speaker called an approximation of the real structure.
Stage D3-C080 Day 3 · Making sense of users' queries
At retrieval, Google tries to match results to the user's language wherever possible: someone searching in Spanish does not necessarily want results in Italian.
extends
Stage D2-C654 Day 2 · Calculating (some) signals
In ranking, country and language signals help Google serve users the right content for their country and language.
Stage D3-C082 Day 3 · Making sense of users' queries
Country is the second retrieval signal: a user searching from Switzerland wants cheese from Switzerland, not from Germany, and a user in Spain is poorly served by results targeting a South American country.
extends
Stage D2-C649 Day 2 · Calculating (some) signals
Country and language are among Google's most important signals and have been used since Google's early days.
Stage D3-C083 Day 3 · Making sense of users' queries
Google called quality the most important of the signals used to order candidates at retrieval: a URL of high quality is more likely to be retrieved from the index for specific queries.
extends
Stage D2-C722 Day 2 · How does the index look like?
Each document in Google's index has pretty much all the signals calculated for it attached, for example quality signals plus the page's country and language, according to an illustration the speaker called an approximation of the real structure.
Analysis D3-C135 Day 3 · How Google thinks about Quality
The quality talk's slide listed MUM among Google's ranking systems, while Google's ranking systems guide says MUM is not currently used for general ranking in Search; read the slide as a list of systems Google runs, not as proof that each one ranks every query.
extends
Docs D1-C130 Day 1 · How Search works and where's AI?
Google's ranking systems guide says MUM is not currently used for general ranking in Search, only for specific applications such as COVID-19 vaccine searches and featured snippet callouts.
Stage D3-C147 Day 3 · How Google thinks about Quality
Google's quality talk said that at any moment thousands of Search experiments are probably running.
extends
Stage D1-C158 Day 1 · Welcome and opening keynotes
Google said that even the classic ten-blue-links layout was settled only after millions of experiments, as part of using as much data as possible for product decisions.
Stage D3-C170 Day 3 · How Google thinks about Quality
Google's quality talk pointed to page 21 of the Search Quality Rater Guidelines for its definition of content quality by effort, originality, talent or skill and accuracy, noting that the document is updated from time to time.
extends
Stage D2-C311 Day 2 · Understanding what's on a page
Gary Illyes pointed to Google's Search Quality Rater Guidelines as the detailed source on how Google thinks about the main content of a page.
Stage D3-C212 Day 3 · How Google thinks about Quality
Google's quality talk said Google clarified that traditional spam techniques aimed at manipulating AI responses also violate its spam policies, and that Google can take action against them.
extends
Docs D1-C129 Day 1 · How Search works and where's AI?
Google's guide says creating separate content for every variation of how people might search, including fan-out queries, primarily to manipulate rankings or AI responses violates its scaled content abuse policy. It adds that its AI systems can understand a page's relevance even without an exact match to the query.
Slide D3-C224 Day 3 · Uncovering Trustworthy Experiences on Discover
Google's Discover slide recommended large images at least 1,200 px wide, with more than 300,000 total pixels and a 16x9 aspect ratio, enabled by the max-image-preview:large setting.
extends
Stage D2-C089 Day 2 · Controlling indexing
max-image-preview:large matters mainly in Discover, where it allows a large image that draws people's attention, so the rule can make a page more visible than leaving it out, John Mueller said.
Docs D3-C287 Day 3 · What are quality updates
Google's spam updates page says its automated spam detection systems run constantly, and a notable improvement to them, such as to the AI-based SpamBrain system, is called a spam update and listed with Google's ranking updates.
extends
Stage D2-C670 Day 2 · Calculating (some) signals
SpamBrain is central to Google's spam-fighting efforts and has been improved many times since its launch.
Stage D3-C311 Day 3 · How Search results are born
Google handles an image used as a search query much like a text query interpreted as an embedding: the image is broken down into vectors (embeddings) that are then searched for in the index.
extends
Stage D2-C740 Day 2 · How does the index look like?
Besides posting lists, Google can retrieve documents through vector embeddings: parts of documents are associated with embeddings, which form a vector space used for retrieval.
Stage D3-C316 Day 3 · How Search results are born
Google generates the parts of a text result, such as title link and snippet, from its understanding of the underlying web page, even when the site owner provides nothing extra.
extends
Stage D2-C724 Day 2 · How does the index look like?
The snippet shown for a web result is reconstructed from the tokens stored in Google's index: Google knows the position of each token in the document and rebuilds the snippet from those positions.
Stage D3-C317 Day 3 · How Search results are born
Google can add elements to a text result, such as a 'highly cited' badge or a preferred-source badge.
extends
Slide D1-C027 Day 1 · What's new in the world of Search
Sites that users mark as a preferred source are more visible in Top Stories and are labelled in AI Mode and AI Overviews.
Stage D3-C319 Day 3 · How Search results are born
Image results shown among web results come from Google's image index and are roughly the same images that Google Images shows for the same query.
extends
Stage D2-C525 Day 2 · Using images to your advantage and Engaging Search users with videos
An image Google has extracted can appear almost anywhere Google shows results, including Discover, image search, web search and AI features, so its potential reach is immense.
Stage D3-C323 Day 3 · How Search results are born
Most of Google's search features need nothing extra from the site owner; Google generates them from what it extracted from the page during indexing.
extends
Stage D2-C446 Day 2 · Finding the gold nuggets: structured data, media, and more!
The 'gold nuggets' that Google's feature extraction step pulls out of a page's HTML are structured data (such as JSON-LD), images and videos.
Stage D3-C324 Day 3 · How Search results are born
Rich results differ from other search features because Google builds them from extra data that site owners provide, usually structured data and usually in JSON-LD format.
extends
Stage D2-C521 Day 2 · What is Structured Data and why we need it on the internet.
Structured data makes pages eligible to appear as rich results, Google's structured data talk said in its recap.
Stage D3-C325 Day 3 · How Search results are born
AI Mode and AI Overviews are not rich results but standard search features: they need no structured data to function and work with the normal text results from Google's index.
extends
Stage D2-C726 Day 2 · How does the index look like?
AI Overviews and AI Mode use the same index structures and token-based snippets as classic web results, a point Google called important but not obvious.
Stage D3-C337 Day 3 · How Search results are born
Review structured data lets a site specify how its users rated something; the review snippet shows an average star rating and often the number of reviews of a product, service or piece of content.
extends
Stage D2-C454 Day 2 · What is Structured Data and why we need it on the internet.
Structured data turns the loosely structured web into structured information that powers visual search features such as review stars and recipe filters (for example by preparation time).
Stage D3-C341 Day 3 · How Search results are born
Some countries rely a lot on social proof such as reviews, Google's speaker said, recalling a point from Google's Day 2 internationalisation talk.
extends
Stage D2-C614 Day 2 · Focusing on Internationalisation and Localisation
According to consumer data shown on a slide in the talk (source not captured), US consumers judge product quality more by user feedback and reviews, while European shoppers seem to look more at brand reputation.
Stage D3-C647 Day 3 · How long does it take to..?
Google may never use structured data from a site it does not trust: once it sees markup it does not trust, it does not touch it.
extends
Stage D2-C500 Day 2 · What is Structured Data and why we need it on the internet.
Structured data that is not relevant to the page's content can be treated as abusive: Google's filters make it ineffective, and egregious cases can lead to a manual action.
Stage D3-C697 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Google's Search Quality Rater Guidelines point out that the tool used to create content is not the problem, but how it was used and what for.
extends
Docs D2-C611 Day 2 · Focusing on Internationalisation and Localisation
Google's spam policies define scaled content abuse as generating many pages mainly to manipulate rankings, with little or no value to users, no matter how they are created, and list automated translating of scraped content among the examples.
Stage D3-C056 Day 3 · Making sense of users' queries
Google's generative AI features in Search build on the traditional ways of searching, so query understanding also flows into AI Overviews and AI Mode.
repeats
Slide D1-C051 Day 1 · How Search works and where's AI?
Three reasons were given: generative AI features are built directly on the core ranking systems, query fan-out expands the original query to find related information, and generative AI features highlight content indexed by Google Search.
Stage D3-C074 Day 3 · Making sense of users' queries
Google's index uses posting lists: for each word, a list of the URLs associated with that word.
repeats
Stage D2-C733 Day 2 · How does the index look like?
For most of the tokens Google finds on the web, though not every single one, the index keeps a posting list of the URLs that contain that token.
Stage D3-C076 Day 3 · Making sense of users' queries
For retrieval, Google uses signals attached individually to each document in the index.
repeats
Stage D2-C722 Day 2 · How does the index look like?
Each document in Google's index has pretty much all the signals calculated for it attached, for example quality signals plus the page's country and language, according to an illustration the speaker called an approximation of the real structure.
Slide D3-C224 Day 3 · Uncovering Trustworthy Experiences on Discover
Google's Discover slide recommended large images at least 1,200 px wide, with more than 300,000 total pixels and a 16x9 aspect ratio, enabled by the max-image-preview:large setting.
repeats
Stage D2-C936 Day 2 · Using images to your advantage and Engaging Search users with videos
Setting the max-image-preview robots meta tag to large can make content perform surprisingly well in Discover, Gary Illyes said.
Stage D3-C311 Day 3 · How Search results are born
Google handles an image used as a search query much like a text query interpreted as an embedding: the image is broken down into vectors (embeddings) that are then searched for in the index.
repeats
Stage D1-C224 Day 1 · How Search works and where's AI?
For visual search, Google breaks an image down into vectors, sends the vectors to the index and returns results based on them.
Stage D3-C329 Day 3 · How Search results are born
Google's structured data feature guide lists the kinds of structured data Google supports with a search feature and what each can do to a site's search results.
repeats
Slide D2-C490 Day 2 · What is Structured Data and why we need it on the internet.
Google recommends using the Search gallery in its developer documentation to find the structured data features that suit a site; the gallery shows each feature and how Google uses the markup.