← Back to GEO Academy
Use cases

How Should Brands Optimize Assets for GEO After Google I/O 2026 Expanded Multimodal AI Search?

Drawing on Google's AI Mode and AI search updates announced in the United States on May 19, 2026, this article explains how consumer, local-service, B2B, and content brands can incorporate images, video, product data, and factual website pages into GEO.

Published 07/13/2026 11 min read
multimodal AI searchGoogle AI ModeGEO optimizationbrand assets

How Should Brands Optimize Assets for GEO After Google I/O 2026 Expanded Multimodal AI Search?

On May 19, 2026 in the United States, Google continued integrating AI Mode, AI Overviews, Lens, shopping, and agentic tasks into the search experience during I/O. For readers using China Standard Time, those announcements correspond to May 20, 2026.

The implication for brands is not "publish a few more SEO articles." A more practical shift is that AI search is beginning to interpret text, images, video, structured product information, and user tasks together.

If a brand maintains only text pages, AI may understand what it is but struggle to recommend it accurately in visual, product, scenario, and comparison questions.

What multimodal AI search changes

In traditional SEO, users see a title, snippet, and webpage link, then open the page and evaluate it. In multimodal AI search, users may upload an image, state a complex need, ask AI to compare options, or even ask it to proceed with shopping or booking steps.

This changes brand assets from conversion-page decoration into important evidence for AI understanding.

For ecommerce brands, product images, specifications, prices, reviews, return policies, and inventory status can jointly affect AI shopping answers.

For local services, location images, address, service items, opening hours, reviews, and booking information can collectively shape the context of an AI recommendation.

For B2B and SaaS, product screenshots, workflow diagrams, case-study charts, sample reports, and demo videos help AI judge which scenarios a product suits.

The most common GEO problems with brand assets

First, images carry a marketing mood but no factual information.

AI needs to know which product, scenario, and function an image represents, not merely see an attractive background. Filenames, page context, alt text, captions, and adjacent copy should all explain the image.

Second, video content has no textual counterpart.

Many websites place critical capabilities in videos without explaining them on the page. AI search may not consistently interpret every detail in a video, so the page should at least provide its summary, steps, intended user, and limitations.

Third, product information is scattered.

Prices, specifications, inventory, delivery, warranties, and returns may be distributed across platforms or even contradict one another. AI shopping answers can pick up outdated or incorrect third-party information.

Fourth, case-study assets lack business context.

A client logo or event photo without the industry, problem, usage, and result definitions gives AI little basis for using the case in a recommendation rationale.

How to approach multimodal GEO

Start with an asset inventory.

List brand assets across the official website, product pages, image library, WeChat Channels, Bilibili, Douyin, Xiaohongshu, maps, media coverage, and ecommerce stores. Mark which assets explain product facts and which are merely visual decoration.

Then fill factual gaps.

Every core product should have, at minimum, a product image, usage-scenario image, specifications, intended users, pricing basis, service scope, risk boundaries, and update date.

Next, map assets to questions.

Organize assets by user question rather than site section. Examples include "Which air purifiers are suitable for an office?", "Which pediatric dental clinic in Shanghai is suitable for a first visit?", and "What does a B2B AI search visibility report look like?" Each question should map to relevant images, text, and evidence.

Finally, retest across platforms.

Google, Gemini, ChatGPT, Claude, Perplexity, Doubao, Tongyi Qianwen, Kimi, and DeepSeek may interpret the same assets differently. GEO reports should record whether answers use the right assets, not merely whether the brand appears.

Priorities by industry

Ecommerce and consumer brands should first correct product data, product images, specifications, and after-sales policies. AI shopping answers compare product facts and purchase intent together.

Local services should first correct location profiles, map information, service items, authentic images, and review responses. When users ask about proximity, suitability for children, or budget, AI combines local evidence to form a judgment.

B2B and SaaS brands should first correct product screenshots, workflow diagrams, sample reports, deployment descriptions, and case evidence. Long decision journeys require verifiable business materials.

Content and education brands should first correct author information, course outlines, sources, update dates, and image descriptions. Source credibility in AI summaries affects how the brand is explained.

How GEO Radar monitors the effect of multimodal assets

GEO Radar can help brands create multi-platform AI search question sets and observe brand mentions, recommendation position, competitor co-occurrence, and differences in description. At https://www.georadar.top, you can group multimodal questions separately, including visual shopping, scenario recommendations, product comparisons, map-based services, and video tutorials.

Retest the same questions after every asset update. Reports should not say only that a rank improved; they should record whether AI understood the correct model, price, usage scenario, service boundary, and evidence source.

The purpose of multimodal GEO is not to make AI remember one image. It is to make a brand's visual assets, product information, and factual website pages corroborate one another.

Sources for this article