Jade Bryan Jardinico decided to test a new habit. For a full week he made Gemini his main visual search tool on a Google Pixel 9 Pro XL and a Samsung Galaxy Tab S10 FE. The result surprised him. What once felt like a simple upgrade now left the older service looking limited.
Google Lens launched years earlier and quickly became a daily driver for many. Point the camera at a plant during a hike. Snap a foreign train schedule. Get instant matches pulled from the web. It handled identification, translation, and shopping links with speed. But something changed when multimodal models arrived.
“Instead of just identifying an object or scene, I suddenly wanted to ask more complex, conversational questions when using visual search,” Jardinico wrote in Android Police. The shift forced constant app switching. Lens delivered clean results. Gemini invited follow-ups. The old workflow broke.
And the differences showed immediately. Upload a photo of ingredients. Lens might return recipe cards or product pages. Gemini explains steps, suggests substitutions, then lets you adjust portions in the same thread. Ask it to count objects in a busy image. Or analyze a short video clip instead of a single frame. Lens hits walls where Gemini keeps talking.
Gemini gains ground on context and conversation
Those advantages run deeper than novelty. Gemini draws on stronger models that process nuance better. It maintains history across questions so one image leads to refined requests without starting over. Lens stays mostly transactional. It scans, matches, and stops.
Jardinico tested plant identification, artwork analysis, product searches, and more. In most cases Gemini returned richer explanations. It connected visual details to broader knowledge instead of simple reverse-image hits. Even better, users can feed it pre-recorded videos. Lens largely sticks to live camera input in many modes.
But trade-offs exist. Early Gemini attachment flows felt clunky. Accuracy on certain counts or obscure items occasionally slipped. Lens still wins for frictionless live translation or quick shopping scans. Jardinico plans to keep both. “If you prefer a conversational interface and superior contextual awareness, Gemini is easily the more powerful option,” he concluded. “I plan to continue using Gemini for most of my visual searches, while keeping Lens around for quick, frictionless tasks.”
Google itself has moved the pieces closer together. Recent upgrades brought video search to Lens, according to Tom’s Guide. Voice queries joined the mix. Project Astra prototypes, detailed on DeepMind’s site, hint at tighter integration ahead. The system already taps Lens and Maps for real-world understanding in glasses form factors and live experiences.
At Google I/O 2026 the company talked up AI agents and multimodal search improvements, per its official blog post. Gemini Live gains video understanding and screen sharing. Chrome’s AI Mode now supports follow-up questions on Lens results. The two tools no longer sit in isolation.
Still. The personal test reveals a psychological break. Once users taste sustained dialogue over an image, single-shot lookups feel incomplete. It’s not that Lens broke. Expectations simply moved.
Industry watchers note similar patterns across AI interfaces. Quick utility gives way to persistent conversation. Developers must decide whether to merge the systems fully or let them coexist with different strengths. For now many users juggle both. One for speed. The other for depth.
That split may not last. Ongoing updates to Gemini models and Search AI Mode suggest Google aims to close the gap. Yet Jardinico’s weeklong experiment shows the gap already changed user habits. Google Lens built the foundation. Gemini is rewriting what comes next.
Power users will test the edges. A photo of circuit boards. A complex chart. A street scene with multiple signs. Each reveals where raw recognition meets reasoned explanation. And right now the balance tilts toward the newer approach. Not everywhere. But enough to make the older service feel, for the first time, insufficient on its own.
WebProNews is an iEntry Publication