Knowledge base v2.13.0 · Community edition · data through 2 October 2026
Topic · Structured data and media
Images were stressed as important on both days: one in six AI Mode searches is multimodal (Day 1), and on Day 2 Google said images and videos drive a large amount of publisher traffic and that an extracted image can appear in web search, Google Images, Discover and AI features. Google finds images through img elements in the HTML (a picture element counts only through the img inside it) and through image sitemaps, cannot index an image without a src, and does not extract CSS background images; to keep images out of results, its documentation names a robots.txt disallow or a noindex X-Robots-Tag header. Alt text should describe the image for someone who cannot see it and can make the image rank for what it names; Google's image guide calls alt the most important attribute for image metadata, while Gary Illyes ranked it below src and said Google uses only three img attributes (said at the event, not in Google's docs). Day 2 confirmed part of Day 1's slide of image signals (resolution, colour and associated text; the list itself is not in Google's docs): the surrounding text is critical context for understanding and ranking an image, which Google's image guide supports. Google supports nearly all popular formats and recommends modern ones such as WebP with small files; Gary Illyes said the AVIF format currently has ingestion hiccups (heard in two independent recordings), although Google's image guide lists AVIF as supported. Day 3 added Discover and Search Console: Google's Discover slides tie click-through to image quality and recommend images at least 1,200 px wide, over 300,000 pixels and 16x9, enabled by max-image-preview:large, with no logos or text-heavy images. Search Console's multimodal search type, rolling out from 24 September 2026, reports searches with Lens, Circle to Search, uploaded images and Chrome's image search, without a query dimension. Google said image results among web results come from its image index, that an image query is searched as embeddings, and that images are indexed in hours to two days on average while changing the image shown with a text result takes 1-2 weeks (said at the event, not in Google's docs). The second recordings added more. On Day 1 Google said the search box is multimodal, that Lens was its answer to young users uploading pictures and expecting answers about them, and that Lens is used by over 1.5 billion people (Google's May 2025 post says every month); for visual search Google breaks an image into vectors and returns results from the index based on them, and Gary Illyes said searching by image or video was not possible before transformers (both said at the event). On Day 2 he said the media indexer attaches each image and video to the URL of the page that hosts it, that Googlebot-Image rules or a disallowed file location keep images out, that noimageindex also stops videos because their thumbnails are images (not in Google's docs), and that max-image-preview:large can make content do surprisingly well in Discover. On AI-generated images Google said the choice is up to the site owner, that Google shows them when users look for them, and that diffusion models are poor at text inside images, so sites should check them for hallucinations and regenerate where needed; Google's image metadata guide supports IPTC digital source types and C2PA details for AI-made images. In Lightning session F a community speaker showed LLM-rewritten alt text going wrong: for a cat anxiety medicine the model described only a cat and a vet, dropping the page's intent, and the site's image performance in Search Console fell; the safer script adds the page title to the image description and is tested on a few URLs first. Author’s view: visual search predates transformers in Google's own history (Google Images 2001, Lens 2017), so the remark is about today's multimodal AI search; and do not set noimageindex on video watch pages.
Things in this topic 25
Counts are claims that name the thing. All things
What to do
One in six AI Mode searches is multimodal, using voice or images.
Speaker Gary IllyesIn Day 1, 11:35 · What's new in the world of SearchEvidence slide photo
Ranking signals differ by result type: web pages (text, links, passages), images (resolution, colour, associated text), news (freshness, originality, diversity), local (location, type, rating, reviews, hours) and videos (language, text from speech).
Speaker Cherry Prommawin, Gary IllyesIn Day 1, 11:45 · How Search works and where's AI?Evidence slide photo, transcript
The Google search box is multimodal: users can search with text, speech or images.
Speaker Lino CattaruzziIn Day 1, 11:00 · Welcome and opening keynotesEvidence transcript
Google said it noticed young users uploading pictures to reverse image search and expecting results that answer something about the picture, and that Google Lens was its answer to that behaviour.
Speaker Gary IllyesIn Day 1, 11:35 · What's new in the world of SearchEvidence transcript
Used byglossary term Google Lens
Google said Google Lens is used by over 1.5 billion people (figure as said on stage on 30 September 2026; no period was given).
Speaker Gary IllyesIn Day 1, 11:35 · What's new in the world of SearchEvidence transcript
Gary Illyes said searching by image or by video, as Search offers it today, was not possible before transformers were invented.
Speaker Gary IllyesIn Day 1, 11:35 · What's new in the world of SearchEvidence transcript
For visual search, Google breaks an image down into vectors, sends the vectors to the index and returns results based on them.
Speaker Gary IllyesIn Day 1, 11:45 · How Search works and where's AI?Evidence transcript
Images were stressed as important.
In Day 1 · session not recordedEvidence notes
Google's May 2025 Search announcement says more than 1.5 billion people use Google Lens to search what they see every month.
“Google Lens, which more than 1.5 billion people are using to search what they see every month”
Publisher Google blog (20 May 2025)Annotates Day 1, 11:35 · What's new in the world of Search
Used byglossary term Google Lens
Visual search predates transformers in Google's own history: its Search timeline dates Google Images to 2001 and Google Lens to 2017, the year Google Research published the Transformer, so read the stage remark as being about today's multimodal AI search.
Author Ibrahim AnjroAnnotates Day 1, 11:35 · What's new in the world of Search
With one in six AI Mode searches being multimodal, original images with descriptive file names, alt text and captions feed AI answers as well as image search.
Author Ibrahim AnjroAnnotates Day 1 · session not recorded
The max-image-preview rule sets the maximum size of a page's image previews: none shows no preview, standard a default-sized one and large the largest possible, for example in Discover.
Speaker John MuellerIn Day 2, 10:30 · Controlling indexingEvidence slide photo, transcript
Used byrequirement DEV-IDX-04glossary term max-image-preview
To be as visible as possible in Google, John Mueller's closing slide recommended the robots rules max-image-preview:large and max-snippet:-1.
“To be as visible as possible in Google, use: max-image-preview:large, max-snippet:-1”
Wording checked against the slide or recording
Speaker John MuellerIn Day 2, 10:30 · Controlling indexingEvidence slide photo, video, transcript
Used byrequirement DEV-IDX-04
John Mueller said Google Search does not always show an image thumbnail for a result, and when it does, the thumbnail is usually limited to the size of the result entry.
Speaker John MuellerIn Day 2, 10:30 · Controlling indexingEvidence transcript
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.
Speaker John MuellerIn Day 2, 10:30 · Controlling indexingEvidence transcript
The noimageindex rule tells Google not to index any of the images on the page, and John Mueller said he could not see why a site would want that.
Speaker John MuellerIn Day 2, 10:30 · Controlling indexingEvidence transcript
Used byrequirement DEV-IDX-09
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.
Speaker Gary IllyesIn Day 2, 13:30 · Finding the gold nuggets: structured data, media, and more!Evidence transcript
For images, Google's feature extraction takes the img element with its src and other attributes, including inline images, and passes them on to Google's image indexing service.
Speaker Gary IllyesIn Day 2, 13:30 · Finding the gold nuggets: structured data, media, and more!Evidence transcript
Gary Illyes said images and videos drive a large amount of traffic to publishers.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Google extracts images and videos from the page's document object model (DOM), and finds images with a fairly standard HTML parser that looks for img elements.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-01
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.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-03
Google supports the picture element only because a picture element must contain an img element, and that img element is what Google extracts and passes to its media indexer.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-01
One of the most reliable ways to make sure Google finds an image is to include it in an img element in the HTML; Gary Illyes named image sitemaps as the other method.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirements DEV-IMG-01, DEV-IMG-04
Google does not extract CSS background images.
“we don't support CSS background extraction”
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-01
Gary Illyes suggested a div with a CSS background image as a way to keep an image from being picked up by Google.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-06
Of the attributes the HTML standard defines for the img element, Google uses three and ignores the rest; Gary Illyes named src and alt but not the third.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
The src attribute is the most important img attribute, because without it Google does not know where the image bytes are and cannot index the image.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-01
Gary Illyes ranked the alt attribute below src in importance, while saying alt attributes are still important.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Alt text should describe the image in words for someone who cannot see it.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-02
Google uses the words in an alt attribute to understand the image, may attach them to the image at serving time, and the image can rank for concepts the alt text describes.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-02
The text around an image is critical: Google uses it as context to understand the image and to rank it, so an alt attribute alone is not enough.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-03
Besides img elements, image sitemaps tell Google about images: they are XML sitemaps that list, under a page's loc entry, the locations of the images on that page.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-04
Google supports pretty much all of the most popular image formats on the web.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-05
Gary Illyes said the AVIF image format currently has hiccups and Google may have problems ingesting it, although it should technically be supported.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-05
Use high-quality modern image formats such as WebP for a good balance of quality and compression, and keep image files small so people can enjoy them.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-05
Google's media indexer processes the images and videos that feature extraction passes to it and attaches them to the URL of the page that hosts them.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Googlebot-Image and Googlebot-Video are the Google crawlers that fetch images and videos, and robots.txt rules addressed to their user agent tokens control how Google indexes a site's images and videos.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-06glossary term Googlebot-Image and Googlebot-Video
If robots.txt disallows the location of an image or video file, Google does not index that image or video.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirements DEV-IMG-06, DEV-VID-06glossary term Googlebot-Image and Googlebot-Video
The noimageindex robots meta tag tells Google not to index the images on the page.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IDX-09glossary term noimageindex
Gary Illyes said noimageindex also affects videos on the page, because Google has to index a video's thumbnail, which is an image.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IDX-09glossary term noimageindex
Setting the max-image-preview robots meta tag to large can make content perform surprisingly well in Discover, Gary Illyes said.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IDX-04
To the frequent question about AI-generated images on a site, Google's answer is that it is up to the site owner.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirements DEV-IMG-08, DEV-SPM-05
Google serves AI-generated images and videos in search results when users are specifically looking for them.
“We will serve users AI-generated images and videos in search results if they are looking for them specifically.”
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
The diffusion models that generate images were built to generate images, not text, so they are typically poor at rendering text inside an image.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirement DEV-IMG-08
Gary Illyes showed a timeline image that Google made with an image generation model in 2025 and deliberately left unfixed as a counterexample: its text was garbled and its years jumped from 1994 to 1995 and back to 1994.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Sites that use AI-generated images or videos should make sure they work for users, check them for hallucinations and regenerate them where needed.
Speaker Gary IllyesIn Day 2, 13:50 · Using images to your advantage and Engaging Search users with videosEvidence transcript
Used byrequirements DEV-IMG-08, DEV-SPM-05
Version 20 of the Screaming Frog SEO Spider, released in May 2024, was the first that could connect the crawler to ChatGPT through custom JavaScript, which let SEOs rewrite image alt text at scale.
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
In a community speaker's case, an LLM asked to write alt text for a product image of a veterinary anxiety medicine for cats described only what it could see, a cat and a veterinarian, and dropped the stress, anxiety and medical-treatment intent of the page.
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
After shipping LLM-rewritten image alt text on a client's site, a community speaker saw the site's image performance in Google Search Console drop (no figures were given).
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
To test LLM-generated alt text before a rollout, a community speaker recommended running the crawler's custom JavaScript in Screaming Frog's List mode on a few chosen URLs and reviewing the generated alt text, instead of crawling the whole site in Spider mode.
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
Used byrequirement DEV-IMG-02
Because not every generated alt text can be checked by hand, a community speaker's safer script adds the page title to the LLM's image description, so the alt text carries the page's intent and not only what the image shows.
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
Used byrequirement DEV-IMG-02
A community speaker named the limits of the title-based alt-text script: it rewrites only the hero image, assumes the page title states the user's intent, had been tested in only one language or market, and runs on the free Gemini model.
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
A community speaker warned that articles still recommend rewriting alt text with AI, called AI a tool, and said that adopting such new AI implementations without applying existing SEO knowledge can damage ranking performance.
Speaker not identifiedIn Day 2, 14:00 · Lightning session F: MediaEvidence transcript
Google's robots meta tag specification says that with nosnippet a static image thumbnail, if available, may still be shown when that results in a better user experience.
Publisher Google Search CentralAnnotates Day 2, 10:30 · Controlling indexing
Google's documented ways to keep a site's images out of search results are a robots.txt disallow rule (for example for Googlebot-Image) or a noindex X-Robots-Tag HTTP header, with the Removals tool for emergencies.
Publisher Google Search CentralAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirements DEV-IDX-03, DEV-IMG-06glossary term X-Robots-Tag
Google's image SEO guide calls alt text the most important attribute for providing more metadata about an image, and says Google uses it together with computer vision and the page content to understand the image.
“The most important attribute when it comes to providing more metadata for an image is the alt text”
Publisher Google Search CentralAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-02
Google's image SEO guide lists AVIF among the supported image formats (with BMP, GIF, JPEG, PNG, WebP and SVG), and Google announced in August 2024 that AVIF files need nothing special to be indexed.
Publisher Google Search Central, Search Central blog (30 August 2024)Annotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-05
Google's image metadata guide says Google Images supports the IPTC Digital Source Type values for algorithmically created images, such as trainedAlgorithmicMedia, and can show C2PA details in 'About this image', such as whether an image was created or edited with AI tools.
Publisher Google Search CentralAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-08
Set max-snippet:-1 and max-image-preview:large on every indexable template unless licensing requires otherwise, and check that no CMS, plug-in or CDN setting adds lower snippet or image preview limits by default.
Author Ibrahim AnjroAnnotates Day 2, 10:30 · Controlling indexing
Used byrequirement DEV-IDX-04
Hiding an image as a CSS background is a fragile way to keep it out of Google: it only stops extraction from that page, so the same image URL used in an img element elsewhere or listed in a sitemap can still be indexed; the documented robots.txt or noindex X-Robots-Tag methods are the reliable route.
Author Ibrahim AnjroAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-06
Images that should rank need an img element with a real src; hero or product images set as CSS backgrounds are invisible to Google Images and image features, so keep CSS backgrounds for decorative images you do not need indexed.
Author Ibrahim AnjroAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-01
Give important images a caption or explanatory sentence next to them rather than relying on alt text alone, since Gary Illyes ranked src above alt and called the surrounding text critical for ranking the image.
Author Ibrahim AnjroAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-03
While the AVIF ingestion hiccup Gary Illyes mentioned lasts, serve AVIF only through source elements inside picture and keep a WebP or JPEG file in the img src, which is the URL Google extracts, even though Google's documentation lists AVIF as supported; then check in Google Images that key images appear.
Author Ibrahim AnjroAnnotates Day 2, 13:50 · Using images to your advantage and Engaging Search users with videos
Used byrequirement DEV-IMG-05
Google's Discover slide said that for eligible content image quality decides click-through: a small thumbnail lowers the click-through rate and a large, high-resolution image raises it.
Speaker GoogleIn Day 3, 11:45 · Uncovering Trustworthy Experiences on DiscoverEvidence slide photo
Used byrequirement DEV-IMG-07
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.
Speaker GoogleIn Day 3, 11:45 · Uncovering Trustworthy Experiences on DiscoverEvidence slide photo
Used byrequirements DEV-IDX-04, DEV-IMG-07
Google's Discover slide said to avoid generic images, such as a site logo, and text-heavy images.
Speaker GoogleIn Day 3, 11:45 · Uncovering Trustworthy Experiences on DiscoverEvidence slide photo
Used byrequirement DEV-IMG-07
Google's Discover slide said images are subject to the same policy requirements as the content and that misleading, exaggerated or outrageous images should be avoided.
Speaker GoogleIn Day 3, 11:45 · Uncovering Trustworthy Experiences on DiscoverEvidence slide photo
Used byrequirement DEV-IMG-07
Google's Search Console talk presented multimodal search reporting as its fifth and newest launch, under a 'Hot from the oven' badge.
“5. Multimodal search reporting”
Wording checked against the slide or recording
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence slide photo, transcript
Google estimated that updating the image shown with a text result takes 1-2 weeks on average, from a minimum of days up to several weeks to months.
Speaker Gary IllyesIn Day 3, 15:45 · How long does it take to..?Evidence slide photo, transcript
Google said it shows images on the web results page, not only in image search, because from 2005-2006 it saw people appending words such as 'pictures' or 'images' to their queries, which led to the launch of universal search.
Speaker Gary IllyesIn Day 3, 13:25 · How Search results are bornEvidence transcript
Used byglossary term Universal search
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.
Speaker Gary IllyesIn Day 3, 13:25 · How Search results are bornEvidence transcript
Google can still detect intent for image queries, even though they are searched as embeddings, the speaker said.
Speaker Gary IllyesIn Day 3, 13:25 · How Search results are bornEvidence transcript
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.
Speaker Gary IllyesIn Day 3, 13:25 · How Search results are bornEvidence transcript
Product annotations on Google Images results also come from product structured data and do not need Google Merchant Center.
Speaker Gary IllyesIn Day 3, 13:25 · How Search results are bornEvidence transcript
Used byrequirement DEV-SHP-01
Multimodal search reporting in Search Console launched about a week and a half before the 2 October 2026 talk.
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence transcript
Google said searching with images is convenient for users and drives traffic to websites.
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence transcript
Multimodal searches reported in Search Console include searches with Google Lens, Circle to Search on Android, images uploaded to Google Search and Chrome's right-click search on an image.
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence transcript
Used byrequirement DEV-MON-07glossary term Multimodal search type
To see multimodal traffic in the Search Console Performance report, select the multimodal search type instead of the default text-based one.
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence transcript
Used byrequirement DEV-MON-07glossary term Multimodal search type
In a live demo, Google Lens on an Android phone identified a photographed object and listed websites selling it, with deep links that allow buying it online.
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence transcript
In a demo, asking Chrome about an image on a web page opened a side panel of visual matches, one of which was on Google's developers site and led on to more information about Search Console; Google described this as the full flow of the multimodal traffic now reported.
Speaker Ariel KroszynskiIn Day 3, 13:55 · Inside Search Console: What’s New & How to Use ItEvidence transcript
Google indexes images in hours to two days on average, and the fastest it can reliably index an image is under 150 seconds.
Speaker Gary IllyesIn Day 3, 15:45 · How long does it take to..?Evidence transcript
Google tries to index images from news sites much faster than usual, which lowers the average image indexing time (the reading 'news sites' is likely but not certain).
Speaker Gary IllyesIn Day 3, 15:45 · How long does it take to..?Evidence transcript
The image shown with a text result changes only after the image itself is indexed or reindexed and then goes through a further processing step involving embeddings.
Speaker Gary IllyesIn Day 3, 15:45 · How long does it take to..?Evidence transcript
Google's Discover documentation says to specify the large image with schema.org markup or the og:image meta tag, which can influence the thumbnail Discover chooses, and that a vertical image cropped to 16x9 must keep its important details.
Publisher Google Search CentralAnnotates Day 3, 11:45 · Uncovering Trustworthy Experiences on Discover
Used byrequirement DEV-IMG-07
Google's announcement of 24 September 2026 says multimodal search data appears both in the Performance report for Search results and in the generative AI performance report, can be exported, and began rolling out globally that day.
Publisher Search Central blog (24 September 2026)Annotates Day 3, 13:55 · Inside Search Console: What’s New & How to Use It
Used byrequirement DEV-MON-07glossary terms Generative AI performance report, Multimodal search type
For Discover, set max-image-preview:large on article templates, give every article a relevant image at least 1,200 px wide in og:image or schema.org markup that is neither the site logo nor text-heavy, and show clear bylines and dates.
Author Ibrahim AnjroAnnotates Day 3, 11:45 · Uncovering Trustworthy Experiences on Discover
Used byrequirement DEV-IMG-07
Google's Performance report help says the queries dimension is not available for the multimodal search type, because these searches mostly use images, so judge image-led traffic by page, country and device, and make product and article images distinctive enough for Lens to match.
Author Ibrahim AnjroAnnotates Day 3, 13:55 · Inside Search Console: What’s New & How to Use It
Used byrequirement DEV-MON-07
Gary Illyes said searching by image or by video, as Search offers it today, was not possible before transformers were invented.
Google said it created the Transformer architecture (the paper 'Attention Is All You Need', the T in ChatGPT) and published it to move the industry forward, competitors included.
Gary Illyes said images and videos drive a large amount of traffic to publishers.
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.
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.
With one in six AI Mode searches being multimodal, original images with descriptive file names, alt text and captions feed AI answers as well as image search.
Freshness is a signal for queries that deserve fresh results ('query deserves freshness'): when a breaking event hits a city, such as possible closure of Barcelona's airport, users want really fresh results, not results from two weeks ago.
Ranking signals differ by result type: web pages (text, links, passages), images (resolution, colour, associated text), news (freshness, originality, diversity), local (location, type, rating, reviews, hours) and videos (language, text from speech).
Google's media indexer processes the images and videos that feature extraction passes to it and attaches them to the URL of the page that hosts them.
For videos, Google's feature extraction takes the video itself and the data around it, to get a better sense of what happens in the video, and passes them to Google's media indexing engine.
In a community speaker's case, an LLM asked to write alt text for a product image of a veterinary anxiety medicine for cats described only what it could see, a cat and a veterinarian, and dropped the stress, anxiety and medical-treatment intent of the page.
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.
Because not every generated alt text can be checked by hand, a community speaker's safer script adds the page title to the LLM's image description, so the alt text carries the page's intent and not only what the image shows.
The text around an image is critical: Google uses it as context to understand the image and to rank it, so an alt attribute alone is not enough.
A community speaker warned that articles still recommend rewriting alt text with AI, called AI a tool, and said that adopting such new AI implementations without applying existing SEO knowledge can damage ranking performance.
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.
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.
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.
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.
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.
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.
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.
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.
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.
Google's research also counted easy access as part of AI's ease benefit: multimodal input, such as taking pictures, and low friction lower the barrier to entry.
One in six AI Mode searches is multimodal, using voice or images.
Generative models, including image diffusion models, make things up when they lack information or because of issues in their training.
The diffusion models that generate images were built to generate images, not text, so they are typically poor at rendering text inside an image.
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.
Sites that use AI-generated images or videos should make sure they work for users, check them for hallucinations and regenerate them where needed.
Google extracts images and videos from the page's document object model (DOM), and finds images with a fairly standard HTML parser that looks for img elements.
For images, Google's feature extraction takes the img element with its src and other attributes, including inline images, and passes them on to Google's image indexing service.
The text around an image is critical: Google uses it as context to understand the image and to rank it, so an alt attribute alone is not enough.
Ranking signals differ by result type: web pages (text, links, passages), images (resolution, colour, associated text), news (freshness, originality, diversity), local (location, type, rating, reviews, hours) and videos (language, text from speech).
The noimageindex robots meta tag tells Google not to index the images on the page.
The noimageindex rule tells Google not to index any of the images on the page, and John Mueller said he could not see why a site would want that.
Setting the max-image-preview robots meta tag to large can make content perform surprisingly well in Discover, Gary Illyes said.
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.
Google's quality talk said ranking signals differ by result type: for web pages they include the text on the page, links and passages, while for news, probably, freshness, diversity and originality become more important.
Ranking signals differ by result type: web pages (text, links, passages), images (resolution, colour, associated text), news (freshness, originality, diversity), local (location, type, rating, reviews, hours) and videos (language, text from speech).
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.
Setting the max-image-preview robots meta tag to large can make content perform surprisingly well in Discover, Gary Illyes said.
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.
For visual search, Google breaks an image down into vectors, sends the vectors to the index and returns results based on them.
Put every image that should be found in an <img> element with a src attribute
Rests on 6 claims, 6 of them in this topic
Give every article a large, relevant lead image and declare it in og:image or markup
Rests on 6 claims, 6 of them in this topic
Allow full snippets and large previews with max-snippet:-1 and max-image-preview:large
Rests on 8 claims, 5 of them in this topic
Describe every meaningful image in its alt attribute
Rests on 5 claims, 5 of them in this topic
Serve modern, compact image formats with a widely supported file in the img src
Rests on 5 claims, 5 of them in this topic
Keep images out of Search with robots.txt or X-Robots-Tag, not with CSS tricks
Rests on 5 claims, 5 of them in this topic
Keep the IPTC and C2PA metadata that marks AI-generated images
Rests on 4 claims, 4 of them in this topic
Track AI feature visibility in the Generative AI performance report
Rests on 11 claims, 4 of them in this topic
Rests on 8 claims, 3 of them in this topic
Place images next to text that explains them
Rests on 3 claims, 3 of them in this topic
List important images in an image sitemap
Rests on 2 claims, 2 of them in this topic
Publish AI-generated text, images and markup only after human review, and show only real authors
Rests on 11 claims, 2 of them in this topic
Send robots rules for PDFs, images and other non-HTML files in the X-Robots-Tag header
Rests on 3 claims, 1 of them in this topic
Rests on 9 claims, 1 of them in this topic
Serve videos fast in a common container with standard codecs, and never block the files
Rests on 8 claims, 1 of them in this topic