Multimodal search on Google: voice, images and SEO

For many years, searching on Google meant typing a few words into a search box.

That’s changing.

Now a person can type, speak, upload an image, use video or combine different forms of interaction to explain what they need. Google calls this behaviour multimodal search. And the arrival of artificial intelligence is accelerating this evolution.

For SEO, the change is important. Because if people no longer search with words alone, a strategy based exclusively on keywords can fall short. If you’re starting from scratch, begin with what SEO is and why it remains the foundation of everything.

What is multimodal search?

Multimodal search is one in which the user uses more than one type of input to perform a search or interact with the search engine.

It can be text. It can be voice. It can be an image. It can be video. Or a combination of several formats.

For example, a traditional search might be:

«trail running shoes»

A multimodal search might consist of taking a photo of a pair of trainers and asking:

«Are these suitable for trail running?»

The context changes completely. Google no longer has only a phrase. It also has an image that provides visual information. Artificial intelligence allows those signals to be combined to better understand what the user is trying to achieve.

Google is already seeing more searches that go beyond text

In 2026 Google published a specific analysis of how people are using AI Mode in the United States.

One of the most striking data points is that more than one in six AI Mode searches uses a non-text input. In other words, they use modalities such as voice, images or video.

What’s more, Google indicates that image-input searches are growing at a rate above 40% month on month since the launch of AI Mode.

These data reflect the behaviour observed in the United States and shouldn’t automatically be interpreted as a global figure. You can see other data in why AI Mode searches are three times longer.

This gives us an important clue: search is ceasing to be exclusively an activity based on typing keywords.

Voice also changes the way we search

Voice search isn’t new. Google has been working with voice assistants and spoken searches for years.

What changes with AI is the ability to understand more natural questions with more context.

When a person types, they can use a very short query. When they speak, they tend to express themselves differently. For example:

«What Italian restaurants are near me that have vegetarian options?»

The query contains intent, context and constraints. Conversational search makes these differences even more relevant. The user can start with a question and then continue:

• «And which one is closer?»
• «Does it have good reviews?»
• «Can I book for tonight?»

We’re no longer dealing with an isolated keyword. We’re dealing with a conversation. You can see how this behaviour evolves in how search intent is changing with AI Mode.

Image searches are especially important

Images provide something text can’t provide on its own: they let the user directly show the object, place or situation they’re interested in.

Imagine someone finds a piece of furniture they like. Instead of trying to describe it with words, they can take a photo. They can then ask:

• «Where can I find something similar?»
• «What decoration style is this?»
• «What dimensions should it have for a 12 square metre room?»

The image becomes part of the search intent. And this has a direct consequence for SEO: a website’s images can also become an entry point to the content.

Google recommends working on image SEO

Google maintains specific documentation on how to help its systems discover and index images. Its recommendations include:

• Use relevant images.
• Make them easy to discover and index.
• Optimise the pages where they appear.
• Use appropriate HTML elements.
• Provide context around the images.

Google also explains that images can appear across different search surfaces, including Google Images, search results and Discover. If you want to reinforce that part, review how to improve CTR on Google or how to optimise meta title and meta description.

Multimodal search changes search intent

When a person uses text, they normally have to translate what they want to find into words. With an image they can show directly what they’re looking for. This reduces the need to describe it.

For example, someone can photograph a plant and ask:

«What plant is this and why are its leaves turning yellow?»

The image provides additional information. The question provides intent. The combination of both generates a much richer query.

For SEO, this means we need to think about content capable of answering visual questions, not just textual queries. This approach connects with SEO and GEO content that ranks and converts.

From a keyword to a complete problem

Traditional SEO has worked for years with concepts such as main keyword, secondary keywords, search volume, intent, CTR or average position. They all remain useful. You can review the basics in what a keyword is and how to choose keywords with criteria.

But multimodal search adds another dimension. The user can arrive at Google with a problem and not know exactly how to describe it. They can show an image. They can ask a question. They can add information by voice. And they can continue the conversation.

This means a good content strategy has to cover complete needs, not just combinations of words.

AI Mode is also making searches longer

Multimodal search doesn’t appear in isolation. It’s part of a broader change in search behaviour.

Google has indicated that AI Mode queries are, on average, approximately three times longer than traditional searches. What’s more, people are asking follow-up questions more frequently.

This means text, voice and images are part of the same evolution: Google is trying to better understand the context of what the user wants to achieve. If you want to dig deeper, you can read AI Mode searches are 3 times longer.

The image can be the start of the buying process

Multimodal search also has implications for ecommerce. A person can discover a product visually before knowing its name. They can have a reference image and want to find similar products. They can photograph a product in a shop and search for alternatives. Or they can show an object and ask:

«Where can I buy something similar?»

This changes the traditional search process.

Before, we could imagine a journey like: keyword → results → product → purchase.

Now it can be: image → question → exploration → comparison → decision.

For ecommerce, this makes product images and the information that accompanies them even more important. You can dig deeper into how to get organic traffic from Google with AI.

Images shouldn’t be isolated from the content

A common mistake is to think that doing image SEO means simply uploading photographs. That isn’t enough. The image must form part of a page that makes sense.

Google recommends that images be associated with relevant landing pages and that the content of those pages helps to understand the context. For example, a photo of a pair of trainers should be inside a page that correctly explains:

• What the product is.
• Who it’s designed for.
• Its features.
• Its uses.
• Its advantages and limitations.

The image can attract attention. The content must solve the need.

Text remains fundamental

The fact that search is multimodal doesn’t mean text has lost importance. In fact, Google recommends that important content on a page be available in text format. Images and videos should complement the content. Not replace it when the information needs to be explained.

That’s why a good page can combine:

• Text.
• Images.
• Videos.
• Structured data.
• Internal links.
• Interactive elements.

This combination lets you cover different ways of consuming information. Google confirms that its generative experiences can show relevant images and videos, so visual content offers additional visibility opportunities.

Do you need to create different content for each type of search?

Not necessarily. This is a mistake worth avoiding.

It makes no sense to create a different page for every small variation of a query simply because it might appear in a multimodal search. Google recommends focusing on the real needs of users.

Creating hundreds of pages to try to cover all possible queries can end up generating repetitive content or content of little value. A good SEO content plan in 5 steps helps you prioritise what deserves its own URL and what doesn’t.

The most solid strategy is to create resources that respond well to a need and use different formats when they really add value.

How to adapt an SEO strategy to multimodal search

You don’t need to reinvent your whole project. But you can incorporate some improvements.

1. Review your images. Check that your images are relevant, useful and of good quality. Avoid using images only as decoration when they can provide information.
2. Add context. An image works better when it’s integrated into a page that clearly explains what it represents.
3. Use descriptive text. Explain the content of the image when relevant. Text remains a fundamental source of context for search systems. If you need to reinforce the structure of the content, review our H1, H2 and H3 hierarchy guide.
4. Optimise the landing pages. Don’t just think about ranking an image. Think about what the user will find when they reach the page.
5. Work on video when it adds value. Some topics are much better understood through visual demonstrations. Google also includes video in its content opportunities for its AI search experiences.
6. Answer specific questions. Multimodal searches usually combine a visual element with a question. Your content should be able to answer real questions related to what you’re showing. You can reinforce this part with the complete on-page SEO guide.
7. Reinforce internal linking. Connect images, products, articles and related pages. The user should be able to continue their research without getting stuck on a single URL.

Image SEO is gaining importance

For years, image SEO has received less attention than page ranking. Multimodal search can change this situation. An image is no longer just a visual element within an article. It can become the starting point of a search.

This is especially relevant in visual sectors:

• Fashion.
• Interior design.
• Travel.
• Restaurants.
• Automotive.
• Architecture.
• Tourism.
• Food.
• Physical products.
• Beauty.

In these sectors, an image’s ability to convey information can be especially important. It’s a principle that fits with the criteria of E-E-A-T: experience, authority and trust in Google.

What about voice searches?

Voice introduces another difference: people don’t always speak the way they write.

A written search might be:

«best laptop video editing»

A voice search might be:

«Which laptop would you recommend for video editing if I have a budget of £1,500?»

The second query contains more information. And it can express a much more concrete intent. That’s why it’s important to create content that answers complete questions. Not because we have to repeat those questions literally, but because we need to understand what problem exists behind them. It’s one of the principles we work on in website optimisation for AI (GEO).

Multimodal search also changes keyword research

Keyword research doesn’t disappear. But it can be complemented with another type of analysis. In addition to researching keywords, we can study:

• Real questions.
• Frequent problems.
• Comparisons.
• Use situations.
• Visual needs.
• Follow-up questions.
• Objections before buying.

This lets you build more complete content. Instead of creating ten different articles for ten related keywords, it might be better to create a main resource that covers the need and several supporting pieces. This also fits with a good internal linking architecture and with the way Google is evolving search intent with AI Mode.

Google doesn’t recommend special optimisation for AI Mode

It’s important not to confuse the evolution of search with the need to use completely new techniques. Google states that fundamental SEO practices remain relevant for AI Overviews and AI Mode.

There are no additional technical requirements to appear. A page must be indexed and eligible to appear in Google Search with a snippet.

Google also recommends:

• Allow crawling.
• Use internal links.
• Provide a good page experience.
• Keep important content in text.
• Use quality images and videos.
• Keep structured data consistent with the visible content.

What’s more, Google indicates that you don’t need to create special files or add specific schema.org markup to appear in these features. If you want to understand what you should have, review how to create robots.txt and sitemap XML correctly.

Visual content should add something real

Artificial intelligence makes image creation easier. But generating more images doesn’t necessarily mean having a better SEO strategy. What matters is that visual content has a function.

A good image can:

• Explain a concept.
• Show a product.
• Demonstrate a process.
• Help compare options.
• Show a result.
• Provide information that text doesn’t convey easily.

If an image doesn’t add anything, it probably isn’t necessary to add it. Quality and relevance are more important than quantity.

How multimodal search can affect organic traffic

The appearance of new forms of search can change the path the user follows to a website. Before, they might make a written search and go directly to a result. Now they can:

• Take a photo.
• Ask what they’re looking at.
• Receive an answer.
• Ask a second question.
• Compare options.
• Finally visit a website.

The journey is longer. But it can also be much more contextualised. That’s why we shouldn’t measure SEO only by a specific position. We have to look at how our content appears across different experiences.

This connects directly with the new AI visibility reports in Search Console. You can read more about this in our article Search Console and visibility in generative AI.

What all this means for SEO

Multimodal search doesn’t mean keywords have died. Nor does it mean text has stopped being important. It means Google can receive much more information about what a person wants to achieve. And websites have to be prepared to provide information in different formats.

This evolution makes SEO closer to a real understanding of the user.

Conclusion: the search engine is no longer just a text box

Google is evolving towards a much more visual, conversational and multimodal search experience.

The data published about AI Mode shows that a significant part of its queries already use non-text inputs. More than one in six AI Mode searches in the United States uses voice, images or video, and image-input searches are growing rapidly.

For SEO professionals, this means we need to broaden our vision. Written content remains fundamental. But images, videos and visual experiences can become new discovery pathways.

The strategy shouldn’t be to create different content for each format. It should be to create genuinely useful resources and use the format that best solves each need.

Google still recommends the SEO basics. You can review them in what SEO is and why it remains the foundation of everything or in the complete on-page SEO guide. And you can reinforce the technical side with Core Web Vitals: what they are and how to improve them and with a complete SEO audit in 7 steps.

The big difference is that now the user can start that search with a word, a phrase, a photograph or their own voice. And SEO has to be prepared for all of them.

Written by

Alex Figueiras

Freelance SEO consultant in Valencia · WordPress + Elementor + Rank Math

I help local businesses and SMEs rank on Google with technical audits, on-page SEO and content that converts. More than 40 projects audited and cases measured in sectors such as renovations, training and ecommerce.

• About me
• LinkedIn
• Contact