Gemini Browser Research — Skill
When to Use
- Research queries requiring Gemini's knowledge and source citations
- Data-rich research (tables, statistics, comparisons)
- When user explicitly requests Gemini for research
- Complement to ChatGPT research for cross-referencing
Prerequisites
- Chrome MCP tools available (Claude in Chrome extension)
- User logged into Google account with Gemini access
- Tab group context established via
tabs_context_mcp
Phase 1: Setup
1.1 Get or Create Tab
tabs_context_mcp(createIfEmpty: true)
If no Gemini tab exists:
tabs_create_mcp()
Note the tabId.
1.2 Navigate to Gemini
navigate(url: "https://gemini.google.com/app", tabId: <tabId>)
1.3 Read Interactive Elements
read_page(tabId: <tabId>, filter: "interactive")
- Identify the prompt textbox (labeled "Enter a prompt for Gemini")
- CRITICAL: Gemini's input is a
contenteditable DIV, NOT a standard input
form_input will FAIL — always use computer(click) then computer(type)
1.4 Ensure Research Mode (Not Create Image)
- If you see "Deselect Create image" button, click it to exit image mode
- For research, you want plain text mode — no tool selected
Phase 2: Send Research Query
computer(action: "left_click", ref: <textbox_ref>, tabId: <tabId>)
2.2 Type the Query
computer(action: "type", text: "<research_query>", tabId: <tabId>)
Query Crafting Tips for Gemini Research:
- Gemini excels at structured data: ask for comparison tables, timelines
- Request source citations — Gemini often provides them
- Ask for "lesser-known details" for unique material
- Gemini handles multi-part numbered queries well
2.3 Submit
computer(action: "key", text: "Return", tabId: <tabId>)
3.1 Wait for Response
computer(action: "wait", duration: 15-20, tabId: <tabId>)
- Gemini is typically faster than ChatGPT for text responses
- Check for the blue stop button disappearing (means generation complete)
3.2 Verify Completion
computer(action: "screenshot", tabId: <tabId>)
- Response complete when input box shows "Ask Gemini 3" placeholder again
- No spinning/loading indicators
get_page_text(tabId: <tabId>)
- Returns full conversation including Gemini's response
- Note: May include sidebar content (chat history titles) — filter when saving
- Gemini responses often include source links and "Export to Sheets" buttons in text
3.4 Save Research Output
Save extracted text to project files:
Write tool → save to research/ or docs/ directory
Phase 4: Follow-Up Queries
4.1 Continue in Same Conversation
computer(action: "left_click", ref: <textbox_ref>, tabId: <tabId>)
computer(action: "type", text: "<follow_up>", tabId: <tabId>)
computer(action: "key", text: "Return", tabId: <tabId>)
4.2 Start Fresh Conversation
If topic changes significantly:
computer(action: "left_click", ref: <new_chat_button_ref>, tabId: <tabId>)
Then re-read interactive elements and proceed from Phase 2.
Phase 5: Quota Management
5.1 Gemini Rate Limits
- Gemini has usage limits (especially for free tier)
- Signs: error messages, "try again later" responses, slow/no generation
5.2 When Quota Runs Out
- Note the time of exhaustion
- Inform user with estimated reset time
- Schedule a reminder:
create_scheduled_task(
name: "gemini-research-resume",
description: "Resume Gemini research — quota should be reset",
schedule: "<reset_time>",
instructions: "Resume research from: <topic>. Pending queries: <list>"
)
- Switch to ChatGPT or Claude for remaining research in the meantime
Gotchas & Failure Modes
| Issue | Solution |
|---|
form_input fails on prompt | It's a contenteditable DIV. Use computer(click) + computer(type) |
| "Create image" mode active | Click "Deselect Create image" before typing research query |
| Refs become stale | Re-read page with read_page to get fresh refs |
| Response includes old chat history | Start a new chat for clean extraction |
| get_page_text includes sidebar | Filter output — research text starts after "Gemini said" |
| Quota exhausted | Schedule task, switch to alternative AI |
| Sources not loading | Gemini sometimes shows expandable source sections — may not appear in text extraction |
Gemini vs ChatGPT — When to Use Which
| Strength | Gemini | ChatGPT |
|---|
| Source citations | Strong | Weak |
| Structured tables | Strong | Moderate |
| Narrative/storytelling | Moderate | Strong |
| Visual scene descriptions | Moderate | Strong |
| Data accuracy | Strong (web-grounded) | Moderate |
| Speed | Faster | Slower but deeper |
| Follow-up coherence | Good | Excellent |
Best practice: Use both in parallel for cross-referencing. Gemini for facts/data/sources, ChatGPT for narrative structure and cinematic framing.