Google's closing slide told site owners they should use AI, which can supercharge workflows, help with content creation and with website and customer management.
Speaker Gary IllyesEvidence slide photo, transcript
Day 3 · Friday 2 October 2026 · 16:00
Wrapping all up: AI, Search, and making sense of everything.
Speaker from the author's recording label; same recording as How long does it take to..?, with no new introduction.
Shown on screen 3
Consistent with docsD3-C700
Google's closing slide said to use AI responsibly because AI hallucinates, and, especially when creating content briefs with AI, to make sure not to add to the sea of AI slop already flooding the internet.
Speaker Gary IllyesEvidence slide photo, transcript
Used byrequirement DEV-SPM-05
- Extends D2-C941 Day 2: Sites that use AI-generated images or videos should make sure they work for users, check them for…
Consistent with docsD3-C701
Google's closing slide said AI on Google is just SEO: AI features on Google Search use exactly the same processes as traditional results, so no new acronym is needed, as none was for mobile-first indexing or structured data.
“AI features on Google Search use exactly the same processes as traditional results.”
Wording checked against the slide or recording
Speaker Gary IllyesEvidence slide photo, transcript
Things
Said on stage 28
StageD3-C678
Google said there is so much going on in Search that AI topics were covered across the event only where contextually important, and its closing message was to use AI but understand how it works so as to use it responsibly.
Speaker Gary IllyesEvidence transcript
Google described AI as an umbrella of many technologies working together, one of which is machine learning: systems that learn from large amounts of data to make informed decisions.
Speaker Gary IllyesEvidence transcript
Gary Illyes said machine learning is about 50 years old and that Google has been using it for 'probably 30 years', starting with statistical models that made the 'Did you mean' feature possible.
Speaker Gary IllyesEvidence transcript
Some of Google's search ranking features are plain machine learning models, because for some tasks simple machine learning models are more useful than large language models.
Speaker Gary IllyesEvidence transcript
Google described large language models as deep learning on internet-scale data sets, trained so that the model has an internal vector space in which concepts are mapped by context.
Speaker Gary IllyesEvidence transcript
Google distinguished predictive language models, whose primary job is to predict the next word or words in a text, from generative models that produce output from a prompt; next-word prediction is a guess, made with some accuracy, from the context given.
Speaker Gary IllyesEvidence transcript
Predictive language models have been used a lot in Search, and BERT is one: technically, in the purest sense, an LLM.
Speaker Gary IllyesEvidence transcript
Used byglossary term BERT, RankBrain and MUM
- Extends D1-C042 Day 1: BERT is used in indexing to understand each word in the context of the whole sentence rather than one word at…
StageD3-C685
Google said it thinks generative AI is extremely helpful to its users.
Speaker Gary IllyesEvidence transcript
Hallucinations can happen with any AI model, and with current training methods there is no way to get rid of them.
Speaker Gary IllyesEvidence transcript
Used byrequirement DEV-SPM-05glossary term Hallucination
- Extends D2-C462 Day 2: Google's guidance on generative AI content warns that AI output can contain hallucinations and says…
Adding grounding or retrieval-augmented generation (RAG) on top of a model reduces hallucinations but cannot eliminate them.
Speaker Gary IllyesEvidence transcript
Things
Used byglossary term Hallucination
Generative models, including image diffusion models, make things up when they lack information or because of issues in their training.
Speaker Gary IllyesEvidence transcript
Used byglossary term Hallucination
- Extends D2-C939 Day 2: The diffusion models that generate images were built to generate images, not text, so they are typically poor…
Google's Search Quality Rater Guidelines tell raters that any form of deception makes a page untrustworthy and earns it the lowest rating.
Speaker Gary IllyesEvidence transcript
Things
Quality raters cannot give a site a penalty or a manual action; their ratings are converted into labels that Google uses to improve its algorithms.
Speaker Gary IllyesEvidence transcript
Things
Used byrequirement DEV-SPM-05
- Extends D3-C160 Day 3: Google's slide said search quality raters cannot affect the rankings of individual sites.
Publishing AI output without checking it can indirectly harm a site's standing in Search, because deceptive content rated lowest by raters feeds the labels used to improve the algorithms.
Speaker Gary IllyesEvidence transcript
Used byrequirement DEV-SPM-05
Google sees fake online personas as a growing type of deception, for example a professional author bio whose headshot is an AI-generated image of a non-existent person, and pages doing this get the lowest rater rating.
Speaker Gary IllyesEvidence transcript
Used byrequirement DEV-SPM-05
Scaled content abuse is becoming a problem again: in the early 2000s pages were churned out with Perl or PHP scripts, and now the same is done with LLMs.
Speaker Gary IllyesEvidence transcript
Things
Used byrequirement DEV-SPM-02
- Spam policies for Google web search
- Google Search's guidance on using generative AI content on your website
- Extends D3-C284 Day 3: Reviewing the spam slide, the speaker said link spam would probably be replaced with scaled content abuse as…
The cheaper tokens become, the more AI slop is created, and Google counts AI slop as scaled content abuse.
Speaker Gary IllyesEvidence transcript
Things
Used byrequirement DEV-SPM-02glossary term AI slop
- Extends D2-C611 Day 2: Google's spam policies define scaled content abuse as generating many pages mainly to manipulate rankings…
Google said it is going to great lengths to get AI slop out of its search results.
Speaker Gary IllyesEvidence transcript
Used byglossary term AI slop
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.
“it's not the tool that was used that is the problem, but rather how the tool was used and what for.”
Speaker Gary IllyesEvidence transcript
Things
Used byrequirement DEV-SPM-05
StageD3-C699
Google encouraged brainstorming and even creating content and images with AI, while staying mindful of what AI can do, because it sometimes lies to its users and that hurts a business and its Google rankings.
Speaker Gary IllyesEvidence transcript
AI Overviews and AI Mode are built on the Search infrastructure Google has used for 25 to 30 years and have very few processes of their own.
Speaker Gary IllyesEvidence transcript
Things
Used bystory angle A-001
StageD3-C703
Gary Illyes concluded that SEO is not dead.
“To me, that means that SEO is not dead.”
Speaker Gary IllyesEvidence transcript
StageD3-C704
Google warned that shiny new things, such as trying to work out fan-out queries, distract from the real mission of creating helpful content, since Search is just piping that connects users to information.
Speaker Gary IllyesEvidence transcript
Things
- Repeats D3-C065 Day 3: Because every system runs fan-out differently, Google advised understanding that fan-out happens but not…
Search has changed every year since its start, for example the Florida update in 2003 (before core updates existed), Universal Search and the Knowledge Graph.
Speaker Gary IllyesEvidence transcript
Things
StageD3-C706
Google said that each time Search changed someone declared SEO dead, but that new features just mean new opportunities.
Speaker Gary IllyesEvidence transcript
Google presented Web Guide as an alternative to AI Mode: it groups the search results into categories with a short summary for each, while the classic text results remain.
Speaker Gary IllyesEvidence transcript
Things
Used byglossary term Web Guide
StageD3-C708
Google does not tell lightning-talk and poster speakers what to talk about; they bring their own ideas.
Speaker Gary IllyesEvidence transcript
- Extends D1-C227 Day 1: Google received close to 200 submissions for the community lightning talks of the Deep Dive, reviewed them…
StageD3-C709
Gary Illyes said he hoped the community lightning talks and posters confirmed Google's on-stage message, and judged from them that the vast majority of SEOs have a very good handle on what is happening in Search.
Speaker Gary IllyesEvidence transcript
What Google's documentation says 4
Google's Search Quality Rater Guidelines (September 2025) list fake owner or content creator profiles, such as made-up author profiles with AI-generated images, as deception, and say pages using deception of any type should be rated Lowest.
Publisher Google Search Quality Rater Guidelines (PDF, 11 September 2025)
Things
Used byrequirement DEV-SPM-05
Google's Search Quality Rater Guidelines say no single rating can directly change how a page appears in Google Search; ratings measure how well search works and give examples of helpful and unhelpful results to improve it.
Publisher Google Search Quality Rater Guidelines (PDF, 11 September 2025)
Things
Google's July 2025 blog post introduced Web Guide as a Search Labs experiment that uses a custom version of Gemini and a query fan-out technique to group web links by aspects of the query.
Publisher Google blog (24 July 2025)
Things
Used byglossary term Web Guide
Google updated its generative AI content guidance on 1 October 2026 with information from the Search Quality Rater Guidelines, to keep its documentation in sync with the presentations used at its developer events.
Publisher Google Search Central
Things
Analysis by the author 1
AnalysisD3-C693
Show only real authors with real photos and verifiable profiles; never invent author personas or use AI-generated headshots, which Google's raters treat as deception.
Author Ibrahim Anjro
Used byrequirement DEV-SPM-05
Links to other sessions 17
- Stage D3-C680 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Gary Illyes said machine learning is about 50 years old and that Google has been using it for 'probably 30 years', starting with statistical models that made the 'Did you mean' feature possible.
extendsD1-C041 Day 1 · How Search works and where's AI?Statistical models have been used at Google for over 20 years, for catching spam and originally for the 'Did you mean' feature.
- Stage D3-C684 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Predictive language models have been used a lot in Search, and BERT is one: technically, in the purest sense, an LLM.
extendsD1-C042 Day 1 · How Search works and where's AI?BERT is used in indexing to understand each word in the context of the whole sentence rather than one word at a time.
- Stage D3-C686 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Hallucinations can happen with any AI model, and with current training methods there is no way to get rid of them.
extendsDocs D2-C462 Day 2 · What is Structured Data and why we need it on the internet.Google's guidance on generative AI content warns that AI output can contain hallucinations and says AI-generated metadata, including structured data and image alt text, should be fact-checked before publishing and the markup validated.
- Stage D3-C688 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Generative models, including image diffusion models, make things up when they lack information or because of issues in their training.
extendsStage D2-C939 Day 2 · Using images to your advantage and Engaging Search users with videosThe diffusion models that generate images were built to generate images, not text, so they are typically poor at rendering text inside an image.
- Stage D3-C690 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Quality raters cannot give a site a penalty or a manual action; their ratings are converted into labels that Google uses to improve its algorithms.
extendsD3-C160 Day 3 · How Google thinks about QualityGoogle's slide said search quality raters cannot affect the rankings of individual sites.
- Stage D3-C694 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Scaled content abuse is becoming a problem again: in the early 2000s pages were churned out with Perl or PHP scripts, and now the same is done with LLMs.
extendsStage D3-C284 Day 3 · What are quality updatesReviewing the spam slide, the speaker said link spam would probably be replaced with scaled content abuse as the spam type worth talking about today.
- Stage D3-C695 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
The cheaper tokens become, the more AI slop is created, and Google counts AI slop as scaled content abuse.
extendsDocs D2-C611 Day 2 · Focusing on Internationalisation and LocalisationGoogle'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-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.
extendsDocs D2-C611 Day 2 · Focusing on Internationalisation and LocalisationGoogle'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.
- D3-C700 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Google's closing slide said to use AI responsibly because AI hallucinates, and, especially when creating content briefs with AI, to make sure not to add to the sea of AI slop already flooding the internet.
extendsStage D2-C941 Day 2 · Using images to your advantage and Engaging Search users with videosSites that use AI-generated images or videos should make sure they work for users, check them for hallucinations and regenerate them where needed.
- D3-C701 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Google's closing slide said AI on Google is just SEO: AI features on Google Search use exactly the same processes as traditional results, so no new acronym is needed, as none was for mobile-first indexing or structured data.
extendsD1-C038 Day 1 · How Search works and where's AI?AI Mode and AI Overviews use the same crawling and the same index as Search. At serving they add grounding on the Search index and query fan-out.
- Stage D3-C708 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Google does not tell lightning-talk and poster speakers what to talk about; they bring their own ideas.
extendsStage D1-C227 Day 1 · Lightning session A: Automation and AIGoogle received close to 200 submissions for the community lightning talks of the Deep Dive, reviewed them over about a month, and gave each selected community speaker seven minutes on stage.
- Stage D3-C680 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Gary Illyes said machine learning is about 50 years old and that Google has been using it for 'probably 30 years', starting with statistical models that made the 'Did you mean' feature possible.
repeatsStage D1-C209 Day 1 · How Search works and where's AI?Google launched the 'Did you mean' feature around 2001-2002 using a statistical model, which Gary Illyes counted as AI because it is a form of machine learning.
- 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.
repeatsStage D3-C190 Day 3 · How Google thinks about QualityGoogle's quality talk said quality problems should be treated as quality issues, not as AI versus human content, because a lot of good AI-assisted or AI-written content exists.
- D3-C701 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Google's closing slide said AI on Google is just SEO: AI features on Google Search use exactly the same processes as traditional results, so no new acronym is needed, as none was for mobile-first indexing or structured data.
repeatsD1-C050 Day 1 · How Search works and where's AI?Google's answer to 'SEO is dead, long live GEO' is not to worry about the name: good SEO is good GEO and AEO.
- Stage D3-C702 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
AI Overviews and AI Mode are built on the Search infrastructure Google has used for 25 to 30 years and have very few processes of their own.
repeatsD1-C038 Day 1 · How Search works and where's AI?AI Mode and AI Overviews use the same crawling and the same index as Search. At serving they add grounding on the Search index and query fan-out.
- Stage D3-C702 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
AI Overviews and AI Mode are built on the Search infrastructure Google has used for 25 to 30 years and have very few processes of their own.
repeatsStage 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-C704 Day 3 · Wrapping all up: AI, Search, and making sense of everything.
Google warned that shiny new things, such as trying to work out fan-out queries, distract from the real mission of creating helpful content, since Search is just piping that connects users to information.
repeatsStage D3-C065 Day 3 · Making sense of users' queriesBecause every system runs fan-out differently, Google advised understanding that fan-out happens but not overfocusing on individual fan-out queries or on how to rank for them.