How Lensia Works: Sidebar Chat and Quick Capture

Lensia is an AI assistant that lives inside Chrome, and most of the time you use it you are talking to the sidebar chat: a panel that opens next to whatever page you are on and stays there while you work. You ask a question, read the answer, ask a follow-up, switch models partway through, and scroll back to something you worked on last week. When you just want a quick answer about something already on your screen, there is a second surface, a floating panel you pull up with Ctrl+Shift+X. Drag a box around a problem and the explanation appears on top of the page in about three seconds. Two surfaces, one AI. This post walks through both so you know exactly what is happening when you use them. Install Lensia free and follow along.

A quick word on using it well: Lensia is a study aid. It is built to help you understand concepts, see how a problem is solved step by step, generate practice questions, and check your own work after you have attempted it. Using it during a graded assessment where AI is restricted is academic dishonesty, so check your school's academic-integrity policy before you decide where any AI tool belongs in your workflow.

The whole extension installs at around 12 MB. Most rival study extensions weigh 50 to 90 MB, which adds up when Chrome is already juggling thirty tabs. Lensia runs on the Chrome Extensions platform and stays out of the way until you call it.

The sidebar chat: your main workspace

The sidebar is where most of the real studying happens, and it is on every plan. You can open it three ways: click the Lensia icon in your Chrome toolbar, click the small Lensia widget that sits in the bottom-right corner of any page, or use the sidebar keyboard shortcut. A chat opens in Chrome's own side panel, docked next to the page so your notes and the conversation stay visible at the same time.

What makes it the main surface is that it holds a whole conversation instead of one answer:

Conversation history sticks around between sessions. Scroll back to last week's chemistry thread and pick up where you left off, or search an old explanation the night before an exam.
Bookmarks let you save the answers you know you will want again, so they are not buried three days back in the thread.
A model picker sits at the top of the chat. Start a quick algebra question on a fast model, then switch to Claude Sonnet 5 mid-thread when a statistics concept needs more careful reasoning. Same conversation, different brain.

Because the sidebar keeps the thread open, it is the right surface for working a topic out rather than grabbing a single fact. Paste a problem, read the walkthrough, then type "why does step two work?" and keep going until the idea actually clicks. For a study routine built around this back-and-forth, read the study smarter guide.

What paid tiers add inside the sidebar

The sidebar itself is free. Two upgrades make it stronger:

Web search and fetch URL. On a paid tier the AI can pull a definition, formula, or fact from a live web source while it explains, and you can hand it a specific link to read. Handy when your material references something your notes do not cover.
Agent mode on Pro. Agent mode lets the AI chain several steps inside the sidebar before it answers: read the lecture notes open in your tab, take a fresh screenshot, look something up, then walk you through the result. It is the difference between "answer this" and "figure this out for me."

The floating panel: fast answers without leaving the page

The sidebar is where you settle in; the floating panel is for when you do not want to settle in at all. You are reading a page, one thing stops you, and you want it explained without copying text into a chat or losing your place. That is the panel's whole job, and it runs on every plan.

Press Ctrl+Shift+X and Lensia puts a capture overlay on the current tab. Drag a box around the problem or passage, and a worked explanation floats up right on top of the page, which stays fully visible underneath. The panel is draggable, so you can park it next to the thing you are studying.

The panel has two modes:

Capture Mode is the default. Ctrl+Shift+X, drag, read. Best when the question is already on your screen.
Ask Mode is a switch on the same panel, on every plan including Free. Flip it on and the panel becomes a custom-question box: type your own prompt and, if you want, attach up to three images from your screen or device, then send. Use it for questions a single screenshot cannot hold, side-by-side image comparisons, or when you would rather write the question in your own words. It is a natural fit for "why is this step true?" after you have made your own attempt.

Two more panel shortcuts worth knowing:

Ctrl+Shift+F repeats the last capture area without re-drawing the box, so a page full of similar practice problems takes one keypress each.
Ctrl+Shift+E hides the panel completely, leaving no on-screen trace. Press it again to bring the panel back.

What happens to your screenshot

When you capture with the floating panel, a secure scanning engine reads the text off the screenshot right there in your browser: the question, any answer choices, labels, and other visible text. If it reads the text cleanly, Lensia sends only that text to the AI and the image never leaves your device, which is faster and lighter on data. More on that in privacy and AI tools.

If the text is hard to read (handwriting, a low-resolution PDF, a diagram with labels baked into the picture, or an equation saved as an image), Lensia falls back to sending the picture itself to a vision-capable model that reads it the way you would, by looking. That path costs a little more data and about a second more, but you still get a reliable explanation. You do not configure any of this; the fallback runs on its own. If one capture takes slightly longer, image mode probably handled it.

The model lineup and Lensia Max

Both surfaces share the same models, the same reasoning toggle, and the same usage budget, so what you learn on one carries to the other.

Quick models: GPT-5.4 Nano, Gemini 3.1 Flash Lite, DeepSeek V4 Flash, GPT-OSS 120B, Qwen3.5 Flash. Fast and light, good for vocabulary review, quick recall, and warming up on easy practice.
Standard models: GPT-5.4 Mini, Gemini 3 Flash, Claude Haiku 4.5, Grok 4.3, DeepSeek V4 Pro. The default, and a good balance of speed and depth across most homework.
Advanced models (Basic and up): GPT-5.4, GPT-5.5, Gemini 3.1 Pro, Gemini 3.5 Flash, Claude Sonnet 5, Claude Opus 4.8. Heavier models for the hard problems where a thorough explanation is worth the extra half-second.
Lensia Max is the auto-router. Leave the model on Lensia Max and it picks a fitting Quick or Standard model per question, and takes 15 percent off the usage cost while it does. It is the least-effort way to study.

Reasoning mode is a single on/off toggle that tells the model to slow down and check its own work before answering. There are no Low, Medium, or High levels; it is on or off. Leave it off for definitions and recall, turn it on for multi-step math or logic where one early slip wrecks everything after it.

Custom instructions: teach the AI who it is talking to

Custom Instructions save three short blocks the AI reads on every capture and every sidebar message: a nickname it calls you, a companion style (terse and direct, warm, professorial, whatever you actually like to read), and an about you block describing what you are studying and where you want help. Turn it on once and the explanations start sounding written for you instead of for nobody in particular.

How Lensia explains different question types

Neither surface is locked to one format. The goal is always the same: see how the answer is reached so you can check your own attempt and learn the material.

Multiple choice: the AI walks each option, explaining why one holds up and the others fall apart, so you can compare its reasoning to yours.
True or false: it says which way the statement resolves and gives the one-sentence reason, so the rule sticks.
Short answer and fill-in-the-blank: it shows what a strong response looks like and why.
Math and calculations: it shows the formula, the intermediate steps, and the final result with units, so you can find exactly where your own work went sideways.
Essay prompts: it gives a structured outline you can study to see how an argument is built before you write your own.
Diagrams and image questions: image mode lets the AI read a chart, graph, or diagram directly and explain what it shows.
Coding problems: it explains correct code and the logic behind it, and can point out bugs in code from your screenshot so you learn to spot them.
Science: it handles both concept review and worked problems across biology, chemistry, and physics.

For a subject-by-subject walkthrough, read the complete AI homework helper guide.

Try it yourself

Install Lensia from the Chrome Web Store, sign in with Google in the extension to activate (Google sign-in only), and open the sidebar on a real problem you are stuck on. Then try the floating panel on the next question you hit while reading. Most students are running both within a minute. The help center covers any setup questions, and when you are ready to build it into a routine, the study smarter guide shows how.