You're reading a dense research paper. A paragraph uses terminology you don't recognize. You want to ask an AI to explain it. The fastest path today: select the text, Ctrl+C, Ctrl+T, type the LLM's URL, wait for the page to load, click the input field, Ctrl+V, press Enter. Seven steps, maybe fifteen seconds, and by the time you're done, you've lost the thread of what you were reading.
That friction is the reason most people don't use AI tools as often as they'd benefit from them. The barrier isn't capability — ChatGPT, Claude, Gemini, and Perplexity are all remarkably good at explaining concepts, summarizing text, debugging code, and answering questions. The barrier is access latency: the gap between "I have a question" and "the AI has my question." Every extra step in that gap kills casual usage.
This guide covers the current state of browser-AI integration — what tools exist, where they fall short, and how Wilee's approach eliminates the friction entirely.
The Current Landscape of Browser-AI Tools
There are several ways people currently use LLMs from within a browser. None of them are frictionless.
Dedicated tabs. The simplest approach: keep a ChatGPT or Claude tab pinned and switch to it when needed. This works, but it requires tab-switching, manual pasting, and context loss. You leave the page you're reading, which breaks your flow. For quick questions, the overhead outweighs the benefit.
Browser sidebars. Chrome's built-in Gemini sidebar and Edge's Copilot sidebar offer AI access without leaving the page. The tradeoff is screen real estate — sidebars consume horizontal space, which matters on smaller screens and when reading wide-format content. They also don't always carry selected text into the prompt automatically; you often still need to paste.
Extension popups. Several AI extensions add a toolbar icon that opens a chat popup. These are better than tab-switching but still require multiple clicks: click the icon, wait for the popup to load, paste or type your query. The popup also tends to cover part of the page you're trying to reference.
Bookmarklets and custom scripts. Power users sometimes create bookmarklets that grab selected text and open an LLM URL with the text embedded as a query parameter. This is the most efficient pure-browser approach, but it requires technical setup, doesn't work on all LLM services (some don't accept URL-based prompts), and breaks when the service changes its URL structure.
Right-click context menu extensions. A few extensions add "Send to ChatGPT" options to the right-click menu. This is functional but slow — the right-click menu in Chrome with several extensions installed can have 20+ items, and scanning a long vertical list for the right option adds time and cognitive load.
The common thread: every existing approach adds friction between "I selected interesting text" and "the AI is processing my request." Some add three clicks of friction. Some add five. None add zero.
How Wilee's LLM Integration Works
Wilee reduces the interaction to one gesture: select text, hold the right mouse button, flick toward your preferred LLM, release. A new tab opens with the selected text pre-loaded into the LLM's input field. The entire motion takes under a second once you've done it a few times.
Here's what happens technically:
- You select text on any web page
- You hold the right mouse button — Wilee's radial menu appears
- Because text is selected, the menu shows text-context actions, including your configured LLM targets
- You flick toward the LLM you want (e.g., Claude is up-right) and release
- Wilee opens a new tab at the LLM's URL with your selected text passed to the input field
The new tab loads with your text already in the prompt. You don't paste anything. You don't click any input field. The LLM is ready to process your text the moment the page loads.
Supported LLMs
Wilee supports four LLM services out of the box:
- ChatGPT (OpenAI) — the most widely used general-purpose LLM. Strong at conversational explanations, creative writing, and code generation. Good default for most users.
- Claude (Anthropic) — excels at nuanced analysis, long-form text, and careful reasoning. Handles academic papers and technical documentation particularly well.
- Gemini (Google) — integrated with Google's knowledge graph and search. Useful when you want answers that reference current information or Google's ecosystem.
- Perplexity — a search-focused LLM that cites its sources inline. Best for fact-checking, research questions, and situations where you want to verify the AI's claims.
You don't have to use all four. In Wilee's settings, you choose which LLMs appear in your radial menu. If you only use Claude, only Claude shows up. If you want all four, each gets its own wedge. The principle is that your menu stays clean — only actions you actually use.
Use Cases by Profession
The "select and send" pattern is simple, but it maps onto surprisingly varied workflows depending on what you do.
Software developers. You're reading documentation and a code example doesn't behave as expected. Select the code block, flick to ChatGPT: "Why does this throw a TypeError?" You're debugging a stack trace — select it, flick to Claude for a step-by-step breakdown. You're reviewing a pull request and a function looks suspicious — select it, flick to ask "What are the edge cases here?" The radial menu turns any code on any page into an instant prompt. No copy-paste, no tab switching, no context loss.
Researchers and academics. You're reading a paper and hit a paragraph dense with unfamiliar terminology. Select it, flick to Claude: "Explain this in plain language." You find a methodology section that's unclear — select it, flick to Perplexity to find related papers that explain the same method. You're comparing findings across two studies — select a key claim, flick to ask "What's the current consensus on this?" The speed matters because academic reading involves dozens of micro-questions per session. If each one costs 15 seconds of tab-switching, most of them go unasked.
Students. Studying involves constant small questions. What does this term mean? How does this formula derive from the previous one? Why is this historical event significant? Can you give me an analogy for this concept? Each question is worth asking but not worth seven steps of copy-paste-switch-type-wait. With a radial menu, the cost of asking drops to one gesture, which means students actually ask — and learn more as a result.
Writers and editors. You've written a paragraph and want to check the tone. Select it, flick to Claude: "Is this too formal for a blog post?" You're editing someone else's work and a sentence feels off but you can't articulate why — select it, flick: "What's wrong with this sentence?" You want three alternative ways to phrase a headline — select it, flick. The radial menu turns your LLM into a real-time writing partner that's always one gesture away.
Product managers and business analysts. You're reading a competitor's feature page and want to understand the positioning. Select a key paragraph, flick to ChatGPT: "What market segment is this targeting?" You're reviewing user feedback in a support tool — select a complaint, flick: "What's the underlying product issue here?" You're reading a quarterly report — select a data claim, flick to Perplexity: "Is this consistent with industry benchmarks?"
The Privacy Model
A reasonable concern with any tool that interacts with selected text: does the extension see or store what you select?
Here's how Wilee handles it: when you trigger an LLM action, Wilee reads the currently selected text from the page DOM, constructs a URL with that text as a parameter, and opens a new tab. The text travels from the page to the new tab via a URL — it passes through Wilee's content script momentarily during the navigation, but Wilee does not log it, store it, transmit it to any server, or retain it after the tab opens.
Wilee has no backend server that processes your text. There's no API call to Algomittens (Wilee's developer). The text goes from the page you're reading directly to the LLM service you chose, via a new browser tab. Once the tab is open, the interaction is between you and the LLM — Wilee's role is finished.
This is verifiable. Wilee's permissions are listed in its Chrome Web Store page, and the extension's network activity can be inspected using Chrome's DevTools. You'll see requests to the LLM service when you trigger an action, and nothing going to Algomittens or any third party in connection with your selected text.
Customizing Your LLM Setup
The default configuration puts available LLM actions in your text-context radial menu. But you can customize this in several ways:
Choose which LLMs appear. In Wilee's AI Actions settings, toggle each LLM on or off. Only enabled LLMs appear in your radial menu. If you only use one, your text-context menu has one LLM wedge, leaving room for other actions like Copy, Search, and Define.
Set wedge positions. Each LLM action can be assigned to a specific direction in the radial menu. If you use Claude most often, put it in the direction you flick fastest — typically the one aligned with your dominant hand's natural wrist motion. In the Gesture settings, click a wedge in the preview and assign the LLM action to it.
Use prompt templates. Instead of sending raw selected text, you can configure prompt templates that wrap the text in an instruction. For example: "Explain this code to a junior developer: [selected text]" or "Summarize this in three bullet points: [selected text]." Templates let you skip writing the same instruction every time.
Combine with other actions. The radial menu isn't limited to LLM actions. A typical power-user text-context setup might be: Copy (up), Search (right), Claude (up-right), ChatGPT (down-right), Define (left), Paste (down). All your most-used text operations in one menu, one gesture each.
The Shift from Search to Ask
For twenty years, the dominant pattern for answering a question while browsing was: open a new tab, type a search query, scan results, click a link, read the answer. LLMs have compressed this into: give the AI your question, get an answer. But the browser hasn't caught up to that shift. The tab-search-scan-click-read pattern is still the default, and most AI integration is bolted on top of it rather than replacing it.
Gesture-based LLM integration is what the "ask instead of search" pattern looks like when it's fully integrated into the browsing experience. You don't leave your page. You don't open a search tab. You don't formulate a search query. You select the thing you're confused about, flick, and the AI gets it. The answer appears in a new tab when you're ready for it, and your reading flow on the original page is uninterrupted.
This matters most for the questions that currently go unasked — the quick clarifications, the "what does this term mean," the "is this claim accurate" checks that aren't worth 15 seconds of tab-switching but are worth one gesture. Reducing the cost of asking a question increases the number of questions asked, which means more learning, more verification, and better-informed browsing.
Getting Started
Wilee's LLM integration works on the free tier. Install Wilee, select text on any page, hold the right mouse button, and you'll see LLM actions in the radial menu. Flick toward one and a new tab opens with your text loaded. No API keys, no configuration, no account required.
If you want to customize which LLMs appear, their positions in the menu, or add prompt templates, open Wilee's settings (click the extension icon in Chrome's toolbar). The AI Actions section controls everything related to LLM integration.