October 2, 2026
People now start Google searches with a photo or screenshot. Here is how to optimize images for multimodal AI search, from visual intent and best practices to the new Search Console multimodal filter.
TL;DR:
Multimodal search starts with an image, photo, or screenshot instead of typed words, through Google Lens, Circle to Search, image uploads, and Chrome's Search this image option.
Optimize in three layers: make each image easy to crawl, describe what it shows plainly, and place the facts a searcher needs right beside it.
Since September 24, 2026, Search Console reports this traffic as Web: multimodal, in both the Search results and Generative AI features reports.
Google shows no query text for multimodal visits, so judge results by page, country, device, and date, and track ChatGPT, Gemini, and Perplexity separately.
Someone photographs a lamp in a friend's living room, circles it on their phone, and Google returns pages that sell it. No keyword was typed, and the page that earned the visit was never optimized for a phrase the shopper did not use.
That behavior now has its own line in Search Console. Images used to help a page rank. In multimodal search they are the question, and how you prepare them decides whether your brand becomes the answer.
You should optimize images for AI search because an image is now both the way many searches begin and the evidence AI systems use to answer them. A picture that a machine cannot identify, match to nearby text, or trust will not be chosen, however well the rest of the page reads.
Multimodal search means one search can mix input types, such as a photo, a screenshot, typed words, or a spoken follow up. Our guide to multimodal AI search across text, images, video, and voice covers the full picture.
Three reasons the work pays off:
Discovery without keywords: A shopper who circles a lamp never picks a search term. The image and the words around it do the matching, so keyword targeting alone cannot win that visit.
Real scale: Google's own updates put Lens at roughly 25 billion searches a month in early 2025, about five billion more than the figure it cited in October 2024. It also said Circle to Search was available on more than 300 million Android devices by July 2025.
One effort, many surfaces: Google says the same image guidance applies to thumbnails beside text results, Discover, and Google Images. Image work is one layer of search engine optimization, not a separate project.
The logic carries into generative engine optimization. An image is one more asset that can earn a citation, and its description, placement, and page context decide whether an AI system feels safe using it. That is AI visibility for pictures.
Four Google features drive multimodal image search today: Google Lens, Circle to Search on Android, image uploads in Google Search, and the Chrome right click option called Search this image. Google names all four in its Search Console announcement, and the results can appear as classic web results or inside its AI features.
Each one starts a search in a different moment:
Google Lens: Point a camera at an object or open a saved photo, and Lens identifies products, text, plants, and landmarks, then returns matching pages.
Circle to Search: On Android, a person circles, highlights, or taps anything on screen, including pictures inside other apps, without leaving what they were doing.
Image uploads: Anyone can upload a photo through the camera icon in the search box, on desktop or mobile, and ask about it.
Search this image in Chrome: Right clicking a picture on any web page starts a search from it, so any image, including yours, can be the first step.
The AI layer sits behind these entry points. When an image reaches AI Mode, Google has described identifying the objects in it and then running several related searches about the whole scene and about each object, a technique it calls visual search fan out. A photo of a living room can therefore surface pages about the lamp, the rug, or the sofa, depending on which one your page makes easiest to identify.
Every entry point ends the same way: a page whose images and surrounding text make a confident match.
Visual intent is the goal behind a search that starts with a picture. Google describes a multimodal query as one that uses an image, photo, or screenshot, and it works out what the person wants from the picture itself plus any words added. Nobody types the intent, so your images and pages have to make it easy to infer.
The type of input offers a useful clue. Three patterns show up most often:
A photo of a real object: The person wants a name, a source, or a place to buy, as in what is this plant or where can I get this chair. Pages that name the item plainly, show it from several angles, and list variants and availability suit these searches.
A screenshot from another app: It might capture a product in a social post, a chart, or an error message, so the goal ranges from finding an item to understanding a problem. Pages that put the explanation or product details beside a matching visual serve these best.
A saved or downloaded image: The person is usually hunting for the original source, a larger version, or similar options. The strongest match is an original image with a clear caption and credit, on the page where it first appeared.
Typed words sharpen the picture. Add a color to a sofa photo, or ask a follow up question in AI Mode, and the intent narrows from identify to compare. Write captions and product copy with the attributes people tend to add: color, size, material, and use.
We cover the brand side of this in our guide on how to optimize your brand's presence in multimodal AI search . The practical point here is different. Search Console hides query text for image searches, so the pages that receive multimodal clicks are your best evidence of what people wanted.
To optimize images for multimodal AI search, make the subject unmistakable, describe it the way a searcher would, and put the next answer beside it. Those three moves strengthen both the visual match and the text match that AI systems rely on when choosing what to show and cite.
Work through five steps:
Pick images that carry answers: Start with pages that already earn visits, and with product shots, diagrams, comparison charts, and screenshots.
Make the subject unmistakable: Use one clear subject per image, several angles, and at least one shot in context. Keep labels, logos, and model numbers legible in the frame, since Lens can read text inside pictures. In a scene with many items, make each one separately identifiable.
Describe it like a searcher would: Alt text should name the subject plus the details that set it apart, such as color, material, and model. Captions add what alt text should not: why the image matters or what it demonstrates. Skip keyword stuffing, which Google warns can make a page look like spam.
Answer the next question on the same page: After identification comes price, size, availability, reviews, or how to use the thing. Put those facts in plain text within a scroll of the image, never only inside the picture.
Signal your preferred image: When a page holds several pictures, indicate which one represents it through primaryImageOfPage, an image on the page's main entity, or an og:image tag. Choose a relevant, high resolution shot without a logo, heavy text, or an extreme shape.
Recommended Tool
Don't guess whether ChatGPT, Claude, Gemini, or Perplexity can access your content. Analyze your site in seconds.
Check AI Visibility Try Prompt Finder
No signup required • Instant results
For the deeper mechanics, our walkthrough of how to optimize your images for AI search explains why these signals matter to AI systems.
Consistency multiplies the effect. Use the same product name in the alt text, caption, page heading, and structured data, so systems see one entity instead of four. That habit protects AI visibility because every signal points to the same thing.
The best practices for image optimization are technical and repeatable: use real HTML image elements, serve supported formats that stay sharp and fast, submit an image sitemap, keep image URLs stable, and add structured data where it applies. Google applies this guidance across Search, Discover, and Google Images, so it covers multimodal search too.
A working checklist:
Use real img elements: Google discovers pictures through the src attribute of an img element, even when it sits inside a picture element. A product photo set as a CSS background never enters the image index.
Submit an image sitemap: It flags pictures Google might otherwise miss, and it can point to files hosted on other domains, such as a CDN. Verify the CDN domain in Search Console, and reuse one URL per image so Google can cache it.
Pick supported formats and keep a fallback: Google Search reads JPEG, PNG, GIF, WebP, AVIF, SVG, and BMP, so match each file extension to its real type. If you use srcset or picture, keep a working src for tools that ignore them.
Balance sharpness and speed: Crisp images look better as result thumbnails, yet pictures are usually the heaviest part of a page. Compress them, size them to their slots, and test with PageSpeed Insights.
Test on a phone: Camera led searches usually start on mobile, so check how landing pages load, crop, and read on a small screen.
Add structured data where it fits: Supported types can earn image badges and rich results, and the image property is required for that. Follow Google's general structured data guidelines or the markup will not qualify.
Run this list after every redesign or CMS change, since a template update can quietly turn real images into background images.
To view multimodal performance, open Search Console, go to Performance, choose Search results, click the Search type filter, and select Multimodal under Web. The same filter works in the Generative AI features report. Google announced it on September 24, 2026 and said it is rolling out globally.
The filter groups every search that starts from a picture, whether it came from a camera, a circled item, an upload, or a Chrome right click. Web text based covers queries typed into the standard search bar, while Web multimodal covers web results where an image was used as part of the search.
Here is a simple way to read it:
Set the filter and range: Pick Multimodal and a range that spans several weeks. The filter is new, so history is short.
Annotate the change: Search Console supports custom chart annotations. Mark September 24, 2026, because comparisons across that date measure the report as well as your performance.
Read pages, countries, devices, and dates: Clicks, impressions, click through rate, and position still appear. Sort by Pages to see which URLs win visits from images.
Compare with text based traffic: Switch the filter for the same pages. Pages with strong multimodal clicks but weak text clicks reveal visual demand you were not targeting.
Repeat for AI answers: Apply the same filter inside the Generative AI features report.
Turn findings into fixes: Run the optimization checks above on your top multimodal pages, then remeasure next month.
AI Visibility Live preview
Check whether AI engines can reach, read and recommend your pages.
Two limits matter. First, Google says query text is not available for this traffic because these searches mostly use images, so the Queries dimension is switched off. Second, the Search Analytics API listed no multimodal search type when practitioners checked it in late September, so exports and dashboards that request standard web data may blend both. Verify any tool's numbers against the interface.
Search Console also covers Google only. It cannot show whether ChatGPT or Perplexity used your image, or what Gemini says about your brand. Verseodin tracks brand citations across ChatGPT, Gemini, and Perplexity, and Gemini results are the nearest indicator of what Google's AI features say, so read both sources side by side.
No. Beyond shopping, people also photograph plants, spare parts, landmarks, recipes, charts, and error messages. Publishers with original diagrams, software companies with interface screenshots, and local businesses with storefront photos can all receive image led visits. If people might point a camera at what you offer, your images matter.
Google explains that multimodal searches mostly use images rather than text, so query text is not available and the Queries dimension is turned off for this filter. You still get pages, countries, devices, and dates. Use the landing pages as clues to intent, and pair them with citation tracking for the AI side.
Make your products and brand assets easy to identify in images, use one consistent name across alt text, captions, and structured data, and keep price, availability, and specs in plain text beside them. Then measure two things: Search Console's multimodal filter for Google, and citation tracking across ChatGPT, Gemini, and Perplexity for the wider AI picture.
Not automatically. Practitioners who reviewed Google's API reference soon after launch found no multimodal option among the search types, so a standard web request appears to return both kinds of traffic together. Confirm how your tool pulls data, treat September 24, 2026 as a break in your trend lines, and rely on the interface for the split until the API catches up.
No. Typed queries remain a core way people search, and crawlability, helpful content, and clear structure still decide who gets found. Multimodal search adds another way in. Treat image work as an extension of search engine optimization, and reuse the same foundations for AI search optimization across Google and other engines.
Multimodal AI Search and Images: Why Should You Optimize Images for AI Search
Google Features That Enable Multimodal Image Search
Understanding Visual Intent: Queries That Use an Image, Photo, or Screenshot
Optimizing Images for Multimodal AI Search and AI Visibility
Image Optimization Best Practices for Web and Multimodal Search
How to View Your Multimodal AI Search Performance in Search Console
Frequently Asked Questions
Summarize with ChatGPT Summarize with Claude Summarize with Perplexity
S
Satvik Mishra
Co Founder of Verseodin
Satvik Mishra is the Co Founder of Verseodin, an AI visibility platform that tracks brand citations across ChatGPT, Gemini, Claude, and Perplexity. He writes about generative engine optimization strategy and what actually works for brands trying to earn visibility in AI powered search.
Newsletter
Get the latest AI Visibility insights, GEO research, product updates, and SEO strategies delivered straight to your inbox.
Continue improving your AI visibility.
Discover what users are asking AI about your industry and uncover prompt opportunities.
Open Prompt Finder
Analyze whether AI search engines can discover and cite your website.
Run Free Report
Tutorials 12 min read
Twelve factors shape a Brand Visibility Score in AI search, from mention and citation frequency to share of voice and engine differences. Learn why scores differ, how to diagnose what limits yours, and how to read changes.
Read article
Tutorials 11 min read
A practical guide to brand citation gap analysis for AI search: build a prompt set, audit cited domains, diagnose 6 gap types, and rank fixes by gap score and effort.
Read article
Tutorials 11 min read
AI searchability decides whether ChatGPT, Gemini, and Perplexity can find, trust, and cite your brand. Here are the key factors, a technical audit checklist, and the visibility strategies and best practices that keep a brand easy to find, mention, and cite in AI answers.
Read article
Want more AI SEO insights?
Monthly research on AI search, GEO and citation trends. No noise.
[ partner program ]
Join our partner network, whether you're an agency, consultant, or reseller. Leave your email and we'll reach out with details.
Revenue share
Earn a cut of every referral you bring on, renewing month over month.
Co-marketing
Joint webinars, case studies, and content with the Verseodin team.
Priority support
Direct line to our team plus early access to new features.
How to Optimize Images for Multimodal AI Search and AI Visibility | VerseOdin