How AI assistants are getting smarter – and why desktop still dominates AI search traffic

A recently filed Google patent outlines how AI assistants might be evolving beyond simple keyword detection, tapping into real-world context to craft more engaging and relevant responses. While filing a patent doesn’t guarantee active implementation, it offers a fascinating glimpse into where the technology may be headed.

Here are the five contextual signals identified:

1. Time, Location, and Environmental Context
2. User-Specific Context
3. Dialog Intent & Prior Interactions
4. Input Modalities (Text, Touch, and Speech)
5. System & Device Context

Breaking it down:

• Time, location, and environmental context:
  AI may generate different responses depending on the time of day, weather, or user location. For instance, instead of a generic “Have a great day,” the assistant might say, “Looks like there’s rain headed your way – don’t forget an umbrella!”

• User-specific context:
  AI could tap into user preferences, recently used apps, or prior interactions to fine-tune its responses. If a user often orders sushi and says they’re hungry, the assistant might suggest a favorite Japanese restaurant nearby.

• Dialog intent & history:
  By recognising intent and leveraging past conversations, assistants can create more coherent and proactive exchanges – anticipating follow-up questions or offering suggestions without being prompted.

• Input modalities:
  Whether the user interacts via voice, typing, or touch can influence how the assistant interprets and responds to the request.

• System & device context:
  In cases where the device has low battery or limited computational resources, the assistant may disable LLM-powered enhancements and fall back to basic responses.

Why it matters:

This highlights how AI systems are being designed to go beyond static responses, offering more adaptive, human-like interactions rooted in real-time awareness.

Takeaway:
Google’s patent showcases how future AI assistants could become deeply personalised, using contextual signals to enhance both relevance and user engagement.

A recent analysis of traffic sources for AI-generated search reveals a surprising trend: the overwhelming majority of referrals are coming from desktop devices. The lone exception? Google Search.

Key referral stats:

• Perplexity: 96.5% desktop, 3.4% mobile
• ChatGPT: 94% desktop, 6% mobile
• Google Search: 53% mobile, 44% desktop
• Google Gemini: 91% desktop, 5% mobile
• Bing: 94% desktop, 4% mobile

Why desktop leads in AI referrals:

BrightEdge suggests that mobile AI apps like ChatGPT often display in-app previews of content, requiring an extra tap to visit external websites – potentially deterring mobile users from clicking through.

On the other hand, Google Search maintains a mobile majority in referral traffic, which could stem from its long-standing integration as the default search engine on Apple’s Safari browser.

The Apple factor:

With Apple’s Safari controlling the default search experience on nearly a billion devices, any shift in search engine partnerships announced at WWDC could have seismic implications for mobile search traffic and referral strategies.

Takeaway:
Marketers should keep a close eye on Apple’s decisions around Safari’s default search engine. Meanwhile, AI search traffic continues to skew heavily toward desktop, reinforcing the importance of optimising web experiences for larger screens – even in an increasingly mobile world.

These insights underscore two big shifts:

• AI assistants are becoming more context-aware, moving toward more meaningful and human-like interactions.
• AI-driven search referrals are still tethered to desktop, a trend that could change dramatically if mobile defaults are disrupted.

As always, staying ahead means understanding how these innovations affect user behavior – and being ready to adapt. 

To keep up with our AI and SEO news, please consider subscribing to our Innovation newsletter.