Google’s
AI Mode Transforms Search with Canvas, PDF Uploads and Live
Lens
Google Search has never looked quite like this before. On 29 July
2025, the company announced a sweeping set of updates to AI Mode that add live
video search through Google Lens, a persistent Canvas planning workspace, the
ability to upload PDFs and images directly into queries, and a new contextual
AI panel inside Chrome. Taken together, the changes mark the sharpest single
step yet in Google’s transformation of Search from a list of links into an
interactive, multimodal assistant capable of engaging with the physical
world, uploaded documents, and ongoing projects.
Search Live Brings the Camera Into the Conversation
The most dramatic new feature is Search Live with video input.
Built on Google’s Project Astra multimodal research, it allows users to
stream their phone camera feed directly into AI Mode, ask conversational
questions about whatever the lens sees, and receive Gemini-generated answers
in real time. The interface is accessed through a redesigned Google Lens app
that now displays a dedicated “Live” tab alongside the existing
Search and Translate options.
The practical use cases are wide-ranging. A student can point a
phone at a biology diagram and ask for a step-by-step explanation. A
traveller can aim the camera at a foreign-language menu and get an instant
translation with context. A home-repair amateur can show the camera a broken
pipe fitting and receive guided instructions without typing a single word.
According to 9to5Google,
Search Live with video began rolling out the same week to users enrolled in
the AI Mode Labs experiment in the United States, available on both Android
and iOS.
The capability builds on the voice input Google introduced to AI
Mode the previous month and represents a significant acceleration of the
company’s Project Astra ambitions, which aim to make AI models aware of a
user’s immediate physical environment rather than just their typed queries.
For context on how multimodal AI is reshaping search behaviour, LiveAIWire’s
analysis of multimodal search in education explores the same
technological thread.
Canvas Turns One-Off Searches Into Persistent
Projects
Canvas is the feature that most directly challenges AI-native
tools such as Notion AI and ChatGPT Projects. It appears as a dynamic side
panel inside AI Mode where users can build and organise information across
multiple sessions. Rather than answering a question and disappearing, Canvas
retains the output and allows follow-up refinements over
time.
Google described the use case in concrete terms: plan a
multi-destination trip, ask AI Mode to draft an itinerary, click the
“Create Canvas” button, and then return later to add hotels, adjust
dates, or extend the journey. Each follow-up refreshes the Canvas
automatically. A student preparing for an exam can build a structured
revision guide across an entire study week, adding new topics and prompts at
each session. The feature is designed to make AI Mode a long-form
collaborator rather than a one-shot answer machine.
As noted by TechRadar,
Canvas represents a shift in how Google thinks about the search session
itself: not a moment but a project, not a query but a workflow. Canvas was
set to roll out in the coming weeks to Labs users in the United States, with
broader availability to follow.
PDF and Image Uploads Add Documents to the Search
Context
Until now, AI Mode worked primarily with text queries and camera
images already stored on a device. The new upload capability lets desktop
users drag PDF documents directly into the AI Mode interface and ask detailed
questions about the content. The model then cross-references the document
against the wider web, producing answers supported by both the uploaded file
and live search results.
The most immediate beneficiaries are students and researchers. A
lecturer’s slide deck, a technical specification sheet, or a lengthy
financial report can be submitted to AI Mode, which will then answer
questions, generate summaries, and surface contextually relevant external
sources. Image uploads were already available on mobile and have now been
extended to desktop browsers. Google confirmed that support for additional
file types, including files stored in Google Drive, will follow in the months
ahead.
The development positions AI Mode as a direct competitor to
document-centric AI tools, bringing that functionality into the world’s most
widely used search engine and removing the need to switch between separate
products. The implications for the SEO and publishing ecosystem are
significant, as explored in LiveAIWire’s
earlier coverage of AI Mode’s impact on SEO and third-party news traffic.
Chrome Integration Adds an AI Lens to Any
Webpage
The fourth update concerns how AI Mode is accessed rather than
what it does. Google is adding an “Ask Google about this page”
option to the Chrome address bar dropdown. Clicking it activates Google Lens
in the browser, generates an AI Overview in a side panel, and will soon
surface a “Dive Deeper” button that opens a full AI Mode interface
for extended follow-up. The feature works across webpages, PDF files viewed
in the browser, and other content displayed on screen.
For professionals and researchers who spend hours reading
documents or following technical threads, the integration removes friction
from the process of cross-referencing, clarifying, and building on what they
read. It also tightens the feedback loop between passive consumption and
active inquiry, nudging the browser closer to an AI-enhanced reading
environment where any visible content can become the subject of a
query.
What These Changes Mean for Users and Publishers
Each of the four updates shares a common logic: Google is
embedding AI into moments where people engage with the world around them, not
just the moments when they type a question. A live camera feed, an uploaded
document, a webpage being read in Chrome, and a multi-session planning
workspace are all contexts that traditional search never reached. AI Mode, as
it develops, is designed to reach all of them.
For publishers and content creators, the pattern is worth watching
closely. A search experience that analyses uploaded documents and answers
questions from camera feeds generates fewer outbound clicks than one that
serves a ranked list of links. The shift toward an answer engine continues to
accelerate, and these new capabilities make that direction clearer than any
previous announcement. The question for the wider ecosystem is whether the
richer experience users gain justifies the distribution costs that content
producers bear. That question does not yet have a settled answer, but the
terms in which it must be asked are becoming sharper with every product update
Google releases.
For those watching the Google Workspace integration angle, the
addition of Google Drive file support to AI Mode is also notable, as it
connects Search directly with productivity tools used by hundreds of millions
of people globally. What began as a search experiment is beginning to look
like an ecosystem-wide rearchitecting of how information is found, used, and
retained.
The pace of these changes also raises questions about the boundary
between search and software. When Search can hold a camera to the world,
remember a project across weeks, and read any document you upload, it begins
to resemble something closer to an operating system for information than a
traditional web directory. For readers following Google’s broader strategy,
LiveAIWire’s
coverage of Gemini reading Google Docs aloud illustrates how the same
integration logic is playing out across the company’s product suite
simultaneously.
About the Author
By Stuart Kerr, Technology Correspondent, LiveAIWire