ChatGPT vs Claude vs Gemini: pick your daily driver

On Day 2 you learned that the briefing, not the tool, sets the quality of the answer. Today you put the same briefing in front of all three assistants, taste the differences for yourself, and choose the one that becomes your home base — without firing the other two.

1Why all three feel 90% the same

If you ran yesterday's landlord email through a second assistant, you already know today's first secret: the answers came back in a different voice, but both were specific, sendable, and built from your facts. That wasn't luck. ChatGPT, Claude and Gemini are built on the same generation of underlying technology — systems trained on enormous amounts of text that have learned to hold a conversation, draft, explain, plan and rework. Whatever one of them can do in everyday use, the other two can do at least respectably.

That means everything you've learned so far transfers completely. The back-and-forth conversation from Day 1 works in all three. The Briefing Formula from Day 2 — Role · Goal · Context · Format · Tone — works in all three. "Before you answer, ask me up to 3 questions" works in all three. Iterating in the same thread instead of starting over works in all three. Your skill is portable; nothing you practice today gets wasted if you later switch.

It also means you can stop worrying about a question that paralyzes a lot of beginners: "What if I pick the wrong one?" You can't, not in any way that matters. For 90% of what you'll do — drafting, explaining, planning, thinking out loud — all three are excellent, and all three have free plans good enough to run this entire course. The companies release new versions constantly and the names change every few months, which is exactly why this course teaches capabilities and habits, not version numbers. A version number is trivia. Knowing which assistant to hand which job is a skill — and that's the remaining 10%.

2The 10% that matters: meet your three staff

On Day 1 you met the Confident Intern — brilliant, tireless, occasionally confidently wrong. Here's the upgrade to that picture: you don't employ one intern. You employ three, and they arrived with different résumés. The differences aren't about which is "smarter" this month — they're about capability and ecosystem: what each one can do beyond chat, and what each one is connected to.

ChatGPT — the all-rounder. The most popular of the three, and the one most people have already tried. It's a strong conversationalist, it generates images right inside the chat (with a small free daily budget, and it's particularly good at putting readable, correctly spelled text into an image — useful for posters and cards), and it has a capable voice mode for talking instead of typing. It also includes a small free taste of Deep Research — a handful of lighter-weight runs a month — enough to follow tomorrow's lesson without leaving ChatGPT. The free plan caps how much you can send before it switches you to a lighter model for a while (exact numbers in the cheat sheet below). One thing to know if you are in a market where ChatGPT shows ads: they should be labelled "Sponsored". Treat any Sponsored block as paid placement, not your intern's advice. Notice the label before you trust the answer.

Claude — the writer-analyst. Many people's pick for drafting, editing and document work: it tends to write prose people are happy to send. Its standout muscle is files — the most generous free file handling of the three (exact numbers in the cheat sheet below), and it can create files for you too, handing back a finished document or spreadsheet instead of a wall of chat text. It also has Artifacts: a side panel where Claude builds documents, tables and even small interactive tools next to the conversation, so your work product doesn't get buried in the scroll. Two honest limits: Claude generates no images and no video, full stop, and its deep, multi-step Research mode is paid — on the free plan Claude searches the web but doesn't run long investigations.

Gemini — the best free deal, and the Google citizen. Google's assistant gives away the most on its free plan: Deep Research — a mode where the assistant plans an investigation, reads dozens of sources and returns a structured, cited report (tomorrow's whole lesson) — is included free with a monthly allowance. It also often gives a generous free image allowance (exact limits change and live inside the app), which makes it a good place to practice freely. And because it's Google's, it can plug into your Google life: with your permission, it can reference your Gmail and Drive from inside a chat — the deepest Google integration of the three; neither rival ties into Gmail and Drive this natively on a free plan. Gemini also has Gems — saved assistant setups you'll meet properly on Day 9.

Read those three paragraphs again and notice what's happening: nobody is "the best." ChatGPT is the strongest generalist. Claude is the strongest with words and documents. Gemini hands you the most for $0 and knows your inbox. The question was never which one wins — it's which job goes to whom.

3The free-tier cheat sheet

Two quick words before the table, because they'll follow you through the whole course. A free tier is what a tool gives you for $0. A usage cap is the limit on that gift — so many messages per hour, files per day, or reports per month, resetting on a clock. Caps are always quoted "roughly," because companies move them without warning. Here's the whole landscape on one card:

Two things to take from the card. First: as of mid-2026, low-cost budget tiers exist on two of the three platforms in many countries — but this course needs none of them. Every exercise in all ten days runs on the free rows above. Second: look at the teal cells. They're spread across all three columns. That's the entire argument of today's lesson, sitting in a table: no single assistant holds the best free offer in every row, so using only one means leaving some of the best free capabilities of 2026 untouched.

4Pick a home base — and use the other two on purpose

So how do you actually live with three assistants? Not by treating them as equals — that way lies opening three tabs for every small task. You pick one as your daily driver: the assistant you open by default, where your chat history accumulates, where you'll set up memory and custom instructions (Day 9), and whose quirks become familiar. A home base compounds: the more of your work lives in one place, the more context it has and the faster every next task goes.

Which one should it be? Genuinely: whichever one you liked talking to most in today's taste test below. For everyday chat, drafting and planning, all three are so capable that fit and taste are the honest tiebreakers — does it ask good questions back, do you find its answers easy to read, do you trust its voice? There is no wrong pick, and you're allowed to change your mind later; your skills move with you.

The other two don't get deleted — they get job descriptions. Keep them bookmarked and walk over deliberately: a big question worth a researched report → Gemini's free Deep Research (Day 4; ChatGPT's free allowance works too). A pile of documents to digest, or a polished piece of writing → Claude and its file powers (Day 5). Images in volume → Gemini; an image with words on it → ChatGPT (Day 7). A video script → whichever writes best for you, usually Claude (Day 8). That routing map is exactly what the rest of this course teaches, day by day. Today you only need the principle: one home, two specialists, on purpose.

1. Open all three, in separate tabs. Go to chatgpt.com, claude.ai and gemini.google.com — one browser tab each, signed in with the free accounts you created on Day 1. If any of them asks you to log in again, use the same email or Google account as before. three chat windows, each with an empty message box at the bottom and your Day 1 chats in the sidebar.
2. Run the decision prompt in ChatGPT first. One convention for the whole course, stated once: in copy-paste prompts, anything in [square brackets] is a placeholder — replace it with your own words before you send. Pick a real, small decision you're actually weighing this week, fill in the brackets, and send:
   Copy-paste prompt · the decision testI need to decide between two options: [your real, small decision — e.g. two phone plans, two weekend plans, two gifts for the same person]. The facts: [3–5 bullet points — prices, dates, constraints, what matters to you]. Walk me through the trade-offs, then recommend one and tell me why in plain language.
   a structured walk-through of the trade-offs ending in a clear recommendation with reasons — built from your facts, not generic advice.
3. Paste the exact same filled-in prompt into Claude, then into Gemini. Same words, same facts — don't improve it between tabs, or the comparison stops being fair. Read all three answers side by side. three different answers to an identical brief. One may ask you a question back before recommending; one may use tables or headings; one may be chattier. None of this is luck — it's personality.
4. Run the explanation test in all three. Decisions show how they reason; explanations show how they teach. Pick something from your work or a hobby and send this to each of the three, again with identical wording:
   Copy-paste prompt · the explanation testExplain [a topic from your work or a hobby — e.g. how mortgage interest works, what a sourdough starter is, how padel scoring works] like I'm a smart 15-year-old. Use one analogy from cooking, and end with the 3 things most worth remembering.
   three explanations of the same idea in three noticeably different voices — different analogies, different rhythm, different endings.
5. Score deliberately, not by vibe. Give each assistant a letter (A/B/C) on three things, like the scorecard above: clarity (could you repeat the answer to someone else?), warmth (does the voice feel like a colleague or a manual?), and structure (could you act on it without re-reading?). Two tiebreaker questions: which one asked you a question back before answering — the Day 2 instinct, unprompted? And which recommendation would you actually follow with your own money?
6. Take a 60-second orientation lap in each tab. Before you close anything, find five of the seven controls you met on Day 1 in each assistant — they exist in all three, arranged differently: the message box (where you type), the model picker (usually at the top of the chat — switches between the default fast model and a slower thinking mode), the attach button (the paper-clip or + icon — you'll lean on it from Day 5), the new-chat button, and the sidebar with your chat history. Buttons move around as interfaces update; the furniture stays. the same five controls in three different arrangements — proof that switching assistants is like renting a different car, not learning to drive again.