When you think of a web search, you probably picture a text box. But the fastest-growing way to ask an AI for answers doesn’t involve typing at all. More than 16% of searches in AI Mode now include a voice command, a photo, or a video clip. That’s roughly one in every six queries. It’s a shift that’s quietly changing how we interact with information and it’s just getting started.
What Exactly Is AI Mode?
AI Mode is a search interface—found in platforms like Google’s Search Generative Experience or Microsoft’s Copilot—that uses generative AI to craft a direct answer instead of a list of links. You can ask a question, get a synthesized response, and even have a follow-up conversation. Historically, these queries were typed. But now, the input can be spoken, snapped, or even filmed.
The 16% figure marks a sharp rise from the single-digit percentages seen just 12 to 18 months ago. That growth isn’t accidental. It’s the result of three converging technologies: vision-language models (VLMs) that can interpret images, speech recognition accurate enough for noisy real-world environments, and the ubiquity of cameras and microphones on every smartphone.
Why People Are Pointing Their Cameras at Everything
The most striking use cases are visual. Imagine your car won’t start. Instead of typing out a vague description, you snap a photo of the engine bay and ask, “What’s this part?” Or you upload a video of a strange noise your washing machine makes and ask, “Why is this happening?” That’s multimodal search in action.
Shopping is another driver. A user might photograph a piece of furniture and ask where to buy it. Students point their cameras at math problems or historical landmarks for instant context. For people with motor or visual impairments, voice input isn’t just convenient—it’s essential.
These aren’t edge cases. They’re everyday scenarios, and they’re pushing multimodal adoption forward.
The Reality Behind the Hype
Multimodal search feels magical, but it’s worth remembering what’s actually happening. The AI doesn’t “see” the way you do. It processes pixels and audio waves through a model trained on massive datasets. That means it can misidentify objects, struggle with low-resolution images, or stumble over a heavy accent.
The “garbage in, garbage out” problem gets amplified. A blurry photo or an ambiguous voice command can lead to a confidently wrong answer. That’s a real risk, and it’s one that platforms are still working to mitigate.
There’s also a business angle. Visual search can connect directly to products—think shoppable ads. Voice search, on the other hand, often returns a single answer with no ad slots at all. This threatens the traditional click-based revenue model. And the compute cost of running vision models is steep, which could lock out smaller players and consolidate power among a few tech giants.
Privacy is another concern. Uploading a photo or video to a server means sharing more than just pixels—it can include location metadata, faces, or sensitive documents. Many users don’t realize how much they’re giving away.
The Numbers Aren’t Universal
The 16% figure is an aggregate, but the reality varies widely. In markets with high smartphone penetration like India or Brazil, the share of multimodal queries may be much higher. In desktop-heavy enterprise settings, it’s likely much lower. Age and tech comfort also play a role—younger users are more likely to reach for the camera.
And text isn’t going away. Most multimodal queries still include a text prompt or a follow-up clarification. The 1-in-6 figure means 5 in 6 are still text-only. Text remains the backbone; multimodal is the growing branch.
What’s Next?
As VLMs improve and on-device processing gets faster, expect the 1-in-6 ratio to climb. The race is on among Google, OpenAI, and Microsoft to make multimodal the default. The tools are already in your pocket—the question is how quickly you’ll start using them.
The keyboard isn’t obsolete, but it’s no longer the only way to ask. With 16% of AI Mode searches now using voice, image, or video, we’re entering an era where the question matters more than the input method. The next time you reach for your camera to identify a plant or record a strange sound, you’re part of a shift that’s redefining search itself.
Summary
- Over 16% of AI Mode searches now include voice, image, or video input.
- Growth is driven by advances in vision-language models, speech recognition, and smartphone hardware.
- Common use cases include visual troubleshooting, shopping, education, and accessibility.
- Multimodal search has limitations—AI can misinterpret images or audio—and raises privacy and business model concerns.
- The adoption rate varies by region and device, and text remains the primary input for most queries.
FAQ
Q: What is AI Mode?
A: AI Mode is a search interface that uses generative AI to provide direct answers instead of a list of links, allowing for conversational follow-ups and multimodal inputs.
Q: Does voice search count as multimodal?
A: Yes, voice is one of the modalities. The 16% statistic includes any non-text input—voice, image, or video.
Q: Is this just about Google Lens?
A: No, Google Lens is one example, but the statistic covers all AI Mode interactions across multiple platforms like Bing and ChatGPT.
Q: Are multimodal searches more accurate?
A: Not necessarily. AI can misinterpret images or audio, leading to errors, especially with low-quality input.
Q: Will text search disappear?
A: No, text remains dominant—5 out of 6 searches are still text-based. Multimodal is growing, but it’s an addition, not a replacement.