How to Set Up a Workspace to Compare ChatGPT, Claude, and Gemini Side by Side

How to Set Up a Workspace to Compare ChatGPT, Claude, and Gemini Side by Side

Want to use AI at work but not sure which one fits your tasks best? The most reliable way is to open all three at once, ask the same question, and compare the results side by side. Follow this guide and you can build your own AI comparison workspace in 30 minutes.

Why Compare All Three AIs at Once

The Limits of Using a Single AI

If you use only one AI, you have no benchmark for judging whether its answer is the best one. Ask the same question to all three and you will find that answer quality, tone, accuracy, and length differ remarkably.

Before vs. After Comparison

STEP 1: Set Up Your Three AI Accounts (Time: 10 minutes)

1-1. Create a ChatGPT Account

1. In your browser, go to chat.openai.com.

2. Click the "Sign up" button in the center of the screen.

3. Sign up with your work email or a Google account.

4. After signing up and logging in, the chat screen appears.

5. Check the model at the top left — the free plan uses GPT-4o mini, and the paid plan ($20/month) gives you GPT-4o.

1-2. Create a Claude Account

1. Go to claude.ai.

2. Click the "Start for free" or "Try Claude" button.

3. Sign up with a Google account or email.

4. After logging in, the chat screen appears, and the free plan lets you use the Claude 3.5 Sonnet model.

1-3. Create a Gemini account

1. Go to gemini.google.com.

2. Sign in with your Google account (no separate sign-up needed).

3. The chat screen appears right away, and you can use the Gemini 2.0 Flash model for free.

💡 Pro tip: The free plans of all three services are enough for a side-by-side test. Each has a daily usage limit, though, so it works best to run your comparison in one focused session of 5–10 questions.

STEP 2: Set up a three-way split browser screen (about 5 minutes)

To watch all three AIs at once, you need to arrange your browser windows side by side. Choose whichever of the three methods suits you.

Method A: Split into three with built-in Windows features (recommended ⭐)

Windows 11 users:

1. Click and hold the title bar of the browser window with the ChatGPT tab open.

2. Drag it to the top center of the screen.

3. When the snap layout options appear, click the three equal columns (⬜⬜⬜) layout.

4. Place ChatGPT in the left column.

5. Select the Claude and Gemini windows for the remaining columns.

Windows 10 users:

1. Open the ChatGPT window and press Windows key + ← to snap it to the left half.

2. Open the Claude window and press Windows key + → to snap it to the right half.

3. Open Gemini in a separate window and manually resize it to fit the top or bottom of the right side.

Mac users:

1. Hover over the green button at the top left of each browser window.

2. Select “Tile Window to Left of Screen.”

3. Arrange the remaining windows the same way.

4. Alternatively, install Rectangle (a free app) to split the screen in three with keyboard shortcuts: Ctrl + Option + D → left third, Ctrl + Option + F → center third, Ctrl + Option + G → right third

Method B: Use a Chrome extension

1. Search for “Split Screen for Google Chrome” in the Chrome Web Store.

2. Click “Add to Chrome” to install it.

3. Click the extension icon and select 3-panel split.

4. Enter the URLs chat.openai.com, claude.ai, and gemini.google.com in the panels.

Method C: Use a dedicated comparison tool (most convenient)

1. Go to chatplayground.ai or poe.com.

2. With these services, you type one question into a single input box and see answers from several AIs side by side at the same time.

3. Free usage is limited, however, so we recommend Method A for a full comparison.

💡 Pro tip: With a single monitor, we recommend a three-way split on a screen of 24 inches or larger. On a laptop (13–15 inches), comparing two at a time is easier on your eyes. If you have dual monitors, the ideal setup is two AIs on one screen, and the third plus a comparison-log spreadsheet on the other.

STEP 3: Create a prompt template for comparison testing (about 5 minutes)

3-1. The core principles of comparison

To get an accurate comparison, you must follow these rules:

• Copy and paste the exact same prompt into all three AIs

• Start each AI in a new chat (New Chat) so earlier conversation context doesn’t influence the results

• Test at the same time of day (AI updates can change performance)

3-2. Sample test prompts by task type

📧 B2B sales email test:

You are a B2B SaaS sales expert. Write a cold email for the situation below.

• Our company: a provider of cloud-based inventory management software

• Target customer: purchasing managers at mid-sized manufacturers with KRW 50 billion (≈ US$37,408,300) in annual revenue

• Goal: book a 15-minute online demo

• Tone: professional, but not stiff

• Length: under 200 characters

• Core value: 30% less excess inventory, 40 hours saved per month through automated ordering

📊 Data analysis and summary test:

Analyze the quarterly sales data below and draw 3 insights for an executive report.

• Q1: KRW 1.23 billion (≈ US$920,200) (+8% year over year)

• Q2: KRW 1.57 billion (≈ US$1,174,600) (+23% year over year)

• Q3: KRW 1.12 billion (≈ US$837,900) (-5% year over year)

• Q4: KRW 1.89 billion (≈ US$1,414,000) (+31% year over year)

  For each insight, include specific numerical evidence and an action item for next quarter. Present the results in a table.

📝 Meeting minutes summary and action item extraction test:

Read the meeting notes below and provide (1) a 3-line summary, (2) key decisions, and (3) action items by owner.

[Meeting notes]

Director Kim: New customer acquisition this month reached only 70% of target. The main cause is Competitor A’s price cut.

Manager Lee: Existing customer churn is also up 2 percentage points from last month. The CS team will submit a complaint analysis report by next Monday.

Assistant Manager Park: The marketing campaign’s CTR is 3.2%, better than the industry average (1.8%), but the landing page conversion rate is low at 0.8%. The UX needs improvement.

Director Kim: Then Park, please work with the design team on the UX improvements and build a prototype within 2 weeks. I’ll handle the pricing policy review myself.

💡 Pro tip: If you save your prompts in Notepad or Notion beforehand, you can copy and paste them into all three AIs quickly. Repeat this flow: Ctrl+A → Ctrl+C → click each AI’s input box → Ctrl+V → Enter, and you can send the same prompt to all three AIs in under 30 seconds.

STEP 4: Build a systematic comparison sheet (time needed: 5 minutes)

4-1. Create the comparison sheet in Google Sheets

1. Go to Google Sheets (sheets.google.com).

2. Click “Blank spreadsheet”.

3. Set up the columns as follows:

4-2. Five Evaluation Criteria

Score each AI's response from 1 to 5 on the following criteria:

• Accuracy — Is it free of factual errors?

• Completeness — Does it cover every item you asked for?

• Usability — Can you copy it and use it right away?

• Tone/Style — Does it match the tone you requested?

• Response speed — How long it felt like the answer took to generate

4-3. Sample Evaluation Log

💡 Pro tip: Once you've logged 10 or more tests, the best AI for each type of work starts to become clear. In real-world testing, after about 20 tests, the time spent deciding which AI to use dropped by an average of 85%.

STEP 5: Run a Real Side-by-Side Test (Time needed: 10 minutes)

5-1. Test Procedure

1. Make sure the chat windows of all three AIs are visible in the three-way split screen.

2. Start a new conversation in each AI:

   • ChatGPT: Click "New chat" at the top of the left sidebar

   • Claude: Click "New chat" at the top of the left sidebar

   • Gemini: Click "New chat" in the left sidebar

3. Copy the prompt you prepared in your notepad with Ctrl+C.

4. Click the ChatGPT input box → Ctrl+V → Enter (or click the send button ▶)

5. Right away, click the Claude input box → Ctrl+V → Enter

6. Right away, click the Gemini input box → Ctrl+V → Enter

7. Watch all three AIs generate their answers at the same time.

5-2. What to Check When Comparing Responses

Once all the answers are in, compare them in this order:

Step 1: Compare overall structure (10 seconds)

• Which AI followed the requested format (table, numbered list, etc.) best?

• Is the response length appropriate?

Step 2: Compare accuracy of content (1 minute)

• Did any AI make factual errors?

• Did it cite the numbers accurately?

Step 3: Compare practical usability (1 minute)

• Can you copy and use it as is?

• How much more editing does it need?

5-3. Example of real comparison results

Actual result patterns from testing a B2B cold email prompt:

STEP 6: Build your AI assignment chart by task

6-1. Identify patterns from accumulated test results

After 20 or more tests, you will have a best-AI-by-task chart like the one below. (The chart below reflects general trends aggregated from many users’ test results.)

6-2. Build your own assignment table

1. Add an "AI Assignments" sheet to your spreadsheet.

2. List the 5–10 types of tasks you do most often.

3. For each task, assign the AI with the highest win rate from your three-AI test results.

4. Save this assignment table to your desktop or bookmarks bar.

💡 Pro tip: AI models are updated every 2–3 months. Once a quarter, re-test with the same prompts and refresh your table. Rankings can shift: Gemini’s Korean was weak in early 2024 but improved significantly by 2025.

STEP 7: Advanced Setup — Workflow Automation

7-1. Create a browser bookmark group

No need to open the three sites one by one every time. Here is how to open all 3 tabs with a single click.

In Chrome:

1. Open all three tabs: ChatGPT, Claude, and Gemini.

2. Right-click an empty spot in the tab bar → select "Bookmark all tabs"

3. Name the folder "AI Comparison Test" and save it.

4. From now on, right-click that folder in your bookmarks bar and select "Open all" to open all 3 tabs at once.

7-2. Enter prompts fast with a text expander

1. Install Espanso (free, espanso.org) or TextExpander (paid).

2. Register your frequently used comparison-test prompts as shortcuts:

   • ;email → types out the full B2B cold email prompt automatically

   • ;meeting → types out the full meeting-notes prompt automatically

   • ;data → types out the full data analysis prompt automatically

3. This cuts prompt entry time from 2 minutes to 3 seconds per prompt.

7-3. Set up automatic result saving

1. In each AI’s response, select all (Ctrl+A or drag) → copy with Ctrl+C.

2. Paste it into a Notion or Google Docs template you have prepared in advance.

3. Example template structure:

📅 Date: 2025-01-15

📋 Task type: Cold email

💬 Prompt: (full prompt text)

🤖 ChatGPT response:

(Paste response)

🤖 Claude response:

(Paste response)

🤖 Gemini response:

(Paste response)

⭐ Rating: Claude wins | Reason: natural tone, accurate length

Final Checklist

Check the items below to confirm your workspace is set up correctly:

• Logged in to all 3 accounts: ChatGPT, Claude, and Gemini

• 3-way split screen or bookmark group set up in your browser

• At least 3 prompts for comparison testing prepared in a notepad

• Scorecard created in Google Sheets

• Ran and logged at least 5 comparison tests

• Task-by-task AI assignment table draft complete

Frequently Asked Questions (FAQ)

"Is the free plan enough?"

More than enough. All three services allow dozens of conversations a day on their free plans. However, free plans may have limited access to the latest models (GPT-4o, Claude 3.5 Opus, etc.), so if you want to use AI seriously for work, upgrading just one — the AI you use most often — is the most cost-effective choice.

"How many comparison tests do I need to run to get a meaningful conclusion?"

At least 10 runs, ideally 20. Even for the same type of task, results vary with the details of the context, so you need a sufficient sample. With 10 tests you can identify the best AI with about 70% confidence; with 20, with over 90% confidence.

"How can the whole team use this together?"

Share the scorecard in Google Sheets with your team. If everyone records their test results in the same sheet, even just 20 tests per person adds up to hundreds of comparison data points for the whole team, letting you build a much more accurate AI assignment table.

Setting up this workspace takes 30 minutes in total. But that 30-minute investment will save you 8–12 hours of work every month and lay the foundation for raising the quality of every AI task. Start today.