Claude and Gemini: text and documents versus the Google ecosystem

ChatGPT is dealt with. Now the two that get talked about less, unfairly so: each of them has something the most popular one simply doesn't.

There's a lot of detail ahead, so hold one question in your head: where in your week do long texts live, and where do mail and spreadsheets live? That's what decides which of the two suits you.

Claude: the one that doesn't spoil your text

The first thing people notice when they move to Claude is that it writes differently. Fewer breezy turns of phrase, less "in today's world", fewer conclusions nobody asked for. What comes out reads more like something an attentive person wrote than a press release.

The second thing is how much it holds in its head. Models have a thing called context: the amount of text a model can see at once within one conversation. Claude's is large enough for a whole book, a thesis or a stack of contracts, and you can ask questions across all of it without chopping anything into pieces.

Here's what that looks like in practice. A lawyer uploads five lease agreements at once and asks where the terms diverge. The model doesn't read them one after another, forgetting the last — it sees all five simultaneously and answers: "in the third and the fifth the notice period is two months, in the rest it's one." Doing that comparison by hand would take half a day.

The third thing is code. Among programmers Claude is often the favourite, and a lot of vibe-coding tools are built on it — that's when a website or app gets assembled from a text description, with no programming. You won't need this tomorrow, but it's worth knowing: if you ever try to build something of your own, Claude is likely to be under the bonnet.

Gemini: the one already sitting inside your mail

Gemini's strength isn't how it writes but where it lives. It sits inside Google, and that changes the nature of the work.

Picture a marketer who needs a summary of a month of correspondence with a client, plus a draft reply. In any other model they'd be copying emails into the chat one by one. In Gmail they simply ask Gemini for the summary — it looks at the thread itself, because it's in the same mailbox. No copying, no tabs.

Same story in documents and spreadsheets: edit the text in place, build a formula, split a list into columns. On top of that Gemini is closest to fresh information — Google Search stands behind it, not only what the model once learned.

Worth saying separately: Gemini's free tier is the most generous of the three. Google's infrastructure shows — it's cheaper for them to give away.

Six situations and who to give them to

Theory is theory, but people choose for a specific job. Here are six common situations — see which ones are yours.

• An eighty-page contract and you need to understand the risks → Claude, it sees the whole document at once.
• Reply to an email without leaving your inbox → Gemini, it's already there.
• Rewrite an article so it sounds alive → Claude, that's its strong suit.
• Find out what ticket prices are doing right now → Gemini, it has search behind it.
• Put together a simple website with no code → Claude, most such tools run on it.
• A summary of thirty emails from this week → Gemini, it can reach your mail.

The pattern is simple: Claude wins where there's a lot of text, Gemini wins where you need your own data inside Google, or freshness.

Two beliefs that stop people trying

You'll hear both in every conversation about AI, and both narrow the choice for no reason.

"Claude is for programmers." Code is one of its strengths and, for most people, not the main one. Journalists, lawyers, editors and translators love it precisely for ordinary text.

"Gemini is just search with a chat on top." Search is only part of it. The main thing is that it works inside your mail and your documents, where other models have no way in at all.

The catch: Claude's free tier is strict

In fairness: free Claude runs out sooner than the others. On a busy day you'll hit the message limit, and it's annoying at exactly the moment the work got going.

So here's the practical approach while you're getting acquainted. Try Claude in short, dense runs: one task, one conversation, no warm-up. Save the long multi-hour experiments for Gemini and ChatGPT, where the ceiling is higher. That way you see what each is good at without spending an evening bumping into a wall.

How to set up the missing accounts

The exercise ahead needs all three. After the last lesson you most likely have ChatGPT already — that leaves two to set up, two minutes each, and neither asks for a card.

ChatGPT: open chatgpt.com, click Sign up and choose «Continue with Google» — no password to invent.

Claude: open claude.ai, click Sign up and choose the Google option. No password to invent.

Gemini: open gemini.google.com. If you've ever signed into Gmail or YouTube on this device, there's no sign-up at all — it opens already under your account.

All three are free and none of them asks for a card. Install the apps only from the Anthropic and Google publishers: the stores are full of fakes, and they charge for exactly what's free here.

A test drive in one evening

Don't take reviews on trust, including ours. Fifteen minutes with your own hands are worth more than ten other people's comparisons.

Take the same material — a real email, report or chapter of yours — and give both models an identical job. For instance: "cut this in half, keep every number, tone businesslike but warm."

Then compare on three things. Accuracy: nothing lost, all the numbers still there? Voice: does it sound like you or like a robot? Obedience: were all three conditions met, or only two? Usually a favourite emerges right about here — and it isn't always the one the reviews praised.

Do this now

Open claude.ai and gemini.google.com and give them both one real job from your week — one you have to do anyway.

Save both conversations. The next lesson is about money, and your own impression will be worth more there than our words: deciding whether a subscription is worth twenty dollars is much easier once you've seen what these models do to your own work.