How to Optimise Images and Videos for AI Responses? Preparing for Multimodal Search

Glossary of Terms for 2026

To dominate search, you must master the vocabulary of the new web. Here are the core concepts defining modern optimization:

• Multimodal Search happens when someone uses their camera, voice, and text all at once to find what they need. Think of it as a search that works the way humans actually communicate, not just typing keywords into a box.
• GEO (Generative Engine Optimization): It's the practice of structuring content so AI engines like ChatGPT, Gemini, and Perplexity actually cite you in their responses. Traditional SEO got you on page one. GEO gets you inside the answer itself.
• Share of Model (SoM): It measures how often AI models choose your brand as their source of truth. I track this for clients using tools like GetCito, and the competitive insights are fascinating.
• Entities are how search engines understand the world now. They're not matching your keywords anymore; they're identifying distinct concepts like people, brands, and places within their Knowledge Graph.

When AI Can't Find You, You Don't Exist

Stop optimizing for strings. Start optimizing for things. In 2026, keywords are dead; Entities and Intent rule. In 2026, your customers are searching with their cameras, their voices, and their intent.

When I explain this to clients, I tell them: "Google doesn't see the word 'Apple.' It sees Apple the company, Apple the fruit, and Apple Records as three completely different things

If your content strategy relies solely on traditional text-based SEO, you are effectively invisible to the 60% of high-intent traffic now originating from multimodal queries. When a user snaps a photo of a sneaker and asks Gemini, "Where can I buy this nearby?", traditional rankings don't matter. Only Multimodal Authority matters.

Let me give you a real example from last month. I watched my 67-year-old mother use Google Lens for the first time. She photographed her neighbor's gardening gloves and asked out loud, "Find these with better grip for arthritis." Within seconds, she had three options filtered by her exact needs. She never typed a single keyword.

That's the user behavior we're optimizing for now.

What is Multimodal Search and Why Should You Care?

Multimodal search lets you interact with AI using text, images, voice, or video together. Instead of separate silos, platforms like Google Gemini, ChatGPT, and Perplexity understand multiple inputs at once. This matters because it delivers faster, richer answers, making online discovery more intuitive, human‑like, and context‑aware.

And that’s it.

Multimodal AI understands what you saw and what you said. It matches the design, filters the color and features, checks the price, and shows you exactly what you’re looking for. No endless scrolling. No “close enough” options. No guessing what keywords might work.

Just see it, say it, and get the right answer.

Why This Matters for Your Business

The numbers tell a compelling story:

• Product reviews and comparisons represent approximately 25% of cited videos in AI Overviews
• Pages with FAQ schema are 3.2 times more likely to appear in Google AI Overviews
• AI-referred sessions jumped 527% between January and May 2025

The Core Components of Multimodal Search

To optimize for the future of search, you need to understand how the machine thinks. It’s no longer just matching text to text; modern AI reads, sees, and listens simultaneously.

Here is the breakdown of the four "senses" AI uses to understand your content:

1. Computer Vision (The Eyes): This is how AI sees. It looks closely at every image and video frame to understand what’s in front of it, products, logos, shapes, and even the setting. So when someone uses Google Lens on your product photo, the AI relies on clean, well-lit images and proper metadata to correctly recognize what it’s looking at. Blurry or poorly lit visuals? That’s like asking the AI to see without its glasses.
2. Natural Language Processing (The Ears): Whether someone types or speaks their query, AI interprets conversational intent. Voice searches like "What are the best waterproof boots for Alaska winters?" require content that answers naturally, not just keyword-stuffed pages.
3. Semantic Fusion (The Brain): This is where the magic happens. AI combines text, visuals, and audio into unified, context-rich responses. Your job is to create content that connects these elements seamlessly.
4. Retrieval-Augmented Generation (The Researcher): AI pulls real-time information from the web to ground its answers in current data. Fresh, authoritative content wins.

Quick-Reference: The Multimodal Optimization Checklist

Use this table to align your strategy with how AI actually processes data.

Visuals That Speak: Image Optimization for the AI Era

In 2026, image optimization isn’t about gaming an algorithm anymore. It’s about teaching AI how to see. Modern AI doesn’t just store images in an index it interprets them, connects them to meaning, and decides whether they’re relevant.

If you want visibility, your visuals need to speak the machine’s language.

Here’s how:

Strategic File Naming

Use descriptive, hyphenated file names that tell both humans and AI what they're looking at.

Bad: IMG_2847.jpg

Good: vintage-leather-messenger-bag-brown.jpg

When file names align with user intent, they don’t just help AI understand the image, they help users find it. Studies show descriptive filenames can improve image search click-through rates by up to 40%.

Alt Text That Actually Works

Alt text isn’t a checkbox anymore; it’s how AI understands your image. In 2026, think of it as micro-copy for machines. In about 125 characters, explain what the image shows and why it matters.

Bad: Product image.

Good: Vintage brown leather messenger bag with brass hardware and adjustable strap, ideal for daily commute.

The second version gives AI real context. It doesn’t just recognize the object it understands who it’s for and how it’s used. That’s what helps your image show up for searches like office style, work bags, or daily commute essentials, instead of getting lost under the generic label of “bags.”

Semantic HTML Structure

AI understands relationships through structure. An image dropped inside a random <div> gives very little context, but an image wrapped in semantic HTML tells a clear story.

Large Language Models rely on tags like <figure> and <figcaption> to understand how visuals and text relate to each other. When you use them correctly, you’re explicitly saying: this description belongs to this image.

This creates a programmatic bond between image and meaning exactly what AI systems look for.

Instead of generic image tags, use semantic HTML5 to bond your image to its context:

HTML


<figure>
  <img src="vintage-leather-messenger-bag.webp"
       alt="Vintage brown leather messenger bag with brass hardware."
       width="800" height="600">
  <figcaption>The 2026 Vintage Messenger features brass hardware and a reinforced strap for daily commutes.</figcaption>
</figure>

Next-Generation Image Formats

Speed is a proxy for quality. Slow images drain crawl budgets and frustrate users. Adopt modern formats like AVIF and WebP. These formats reduce file size by 30-50% without losing visual fidelity. Faster load times signal technical competence to AI systems, directly influencing your authority score.

Video Optimization: How to Capture the YouTube Citation Surge

If there is one statistic that should dictate your 2026 strategy, it is this: Since January 2024, YouTube citations in AI Overviews have jumped by 25%.

Video is no longer just for engagement; it is a primary source of data for AI. To get cited, you need to stop creating "content" and start creating "answers." Here is how to engineer your video strategy for the AI era.

Quick Audit: Is Your Video AI-Ready?

Schema Markup: The Language AI Speaks

If images and video are your content’s body, Schema Markup is the nervous system. It is the structured data that tells AI exactly what your content means, who created it, and how it connects to the rest of the web.

In March 2025, Microsoft’s Fabrice Canel stated that structured data directly helps LLMs (Large Language Models) understand web content. This isn't just SEO anymore; it’s the foundation of AI processing.

The 2026 "Must-Have" Schema Types

To stay visible, you need to prioritize the schemas that AI agents use to build their responses:

• Organization Schema (The Identity): Establishes your brand's DNA name, logo, and service areas. This feeds directly into Knowledge Panels and AI brand recognition.
• Article Schema (The Authority): Provides the editorial framework. It identifies the "who" (author) and the "when" (date), which are critical signals for E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness).
• VideoObject Schema (The Citation Magnet): You must include the duration, thumbnail, and a full transcript. This allows AI to "watch" and cite specific segments of your video.
• Product Schema (The Salesman) For e-commerce, this feeds real-time pricing, availability, and reviews into visual search tools like Google Lens.
• FAQ Schema (The MVP of AEO): This is the highest-cited schema type. Its question-and-answer format perfectly mirrors how AI assistants present information to users.

Implementation Best Practices

1. Stick to JSON-LD: It is the industry standard cleaner, easier to maintain, and the preferred format for both Google and Bing.
2. The "Mirror" Rule: Never hide data in your Schema that isn't visible on the page. If the AI detects a mismatch between your markup and your visible text, it will flag your site as unreliable.
3. Validate Constantly: Before hitting publish, run your code through the Rich Results Test. Even a small syntax error can make your entire page "invisible" to an AI crawler.

Your 2026 Structured Data Checklist

Don’t leave your AI visibility to chance. Ensure these five specific types are active on your high-value pages:

Google Lens Optimization: Capturing Visual Search Traffic

Visual search isn’t just the future, it’s already here. With nearly 20 billion searches happening every month on Google Lens, brands that ignore it are leaving serious traffic (and conversions) on the table. The good news? Winning in Lens isn’t complicated if you focus on the fundamentals.

The GEO Protocol: How to Engineer Content for AI Citations

Traditional SEO was about pleasing algorithms. Generative Engine Optimization (GEO) is about something bigger: maximizing your Share of Model (SoM). This metric tracks how frequently AI models prioritize your brand as the primary source of truth in generated answers.

If you want to maximize your Share of Model (SoM), your brand’s presence inside AI responses, you need to structure content so that large language models (LLMs) can parse, verify, and cite it with confidence.

Here’s how to do it.

1. The "Citation-First" Framework

AI systems don’t trust vague statements, and honestly, neither do people. What they respond to is clear, verifiable information. That’s why your content should be built with citations in mind from the start.

Instead of broad claims, lead with real data. Saying “sales are up” doesn’t tell anyone much. A clear, specific statement does the job far better, for example:
“In Q1 2026, voice-commerce sales increased by 14%”.

Details like timeframes, locations, and numbers give your content weight. They reduce ambiguity, increase credibility, and make it far more likely that both AI systems and human readers take your message seriously.

AI systems tend to trust ideas that show clear industry agreement. When you include quotes from well-known experts, you strengthen your E-E-A-T signals: experience, expertise, authority, and trust.

Think of it this way: you’re not asking AI to rely on a single viewpoint. You’re showing that your insight is shared and supported by people who actually shape the industry. That context makes your content feel credible, grounded, and worth referencing.

2. The "Inverted Pyramid" for AI

Journalists have used the inverted pyramid for decades. Now, it’s time to apply it to AI.

• Place the direct answer immediately after the H2 header.
• Keep the first 50 words tight, clear, and self-contained.

Why? AI agents often skim only the opening lines of a section when generating snippets. If your answer isn’t upfront, you risk being overlooked.

Measuring Multimodal Search Performance in 2026

You can’t improve what you don’t measure and that’s especially true as search moves beyond blue links. In 2026, discovery happens across text, voice, and visuals, which means the metrics that mattered five years ago no longer tell the full story. Forward-thinking brands are already adjusting how they define visibility.

Here’s what they’re paying attention to now.

Getting Started: Your Multimodal Optimization Roadmap

Feeling overwhelmed by multimodal search? You’re not alone. The good news is that you don’t need to tackle everything at once. Start with the essentials, build momentum, and layer in sophistication as you go. Here’s a practical roadmap to guide you.

The 3-Second "Humanity" Test

AI-generated content is everywhere, and search engines are filtering aggressively for human signals. Before you hit publish, ask yourself:

• Evidence: Did I include original photos or screenshots I created myself? (Stock photos are ignored by AI vision.)
• Experience: Did I use “I” statements to share a real encounter with the product or topic?
• Expertise: Is the author bio linked to verifiable sources like LinkedIn or speaking engagements?

If you can answer “yes” in three seconds, your content passes the humanity test.

Conclusion

We’re living through the biggest transformation in information discovery since Google’s launch. Search is no longer just text; it’s multimodal, powered by AI that interprets words, images, and video together. Traditional SEO isn’t disappearing; it’s evolving into something richer, more complex, and far more aligned with how humans naturally seek answers.

Your competitors are already moving. The brands that will thrive in 2026 and beyond are those that speak AI’s language: content that is semantically complete, structurally clear, and available across every format users engage with.

The window of opportunity is wide open. Today, only 12.4% of websites implement structured data, meaning early adopters gain outsized visibility as AI platforms reward the sites that make their jobs easier.

You don’t need to overhaul everything at once. Start small:

• Add schema to your most important pages.
• Optimize your strongest images.
• Publish one answer-focused video.

Each step compounds, building toward a comprehensive multimodal strategy that positions your brand as the authoritative source AI platforms cite.

The future of search isn’t coming, it’s already here. The real question isn’t whether you’ll adapt, but how fast you’ll move.