Task → AI: a pick-the-model table for any job

Over five lessons you've learned the characters of the big three and the specialists. Now the most useful part: one table you'll keep coming back to. Task on the left, best AI on the right, free alternative right there with it.

What follows is a walk through each row: why it's that way, and when the rule breaks.

The table's rule: task type first

Don't ask "which model is best" — ask "what type is my task?" Text? A fresh fact? An image? Work inside Google? Once you've named the type, the choice is nearly obvious. Below are the eight most common types and the leader for each. Keep your three jobs from lesson one handy — you'll try them against the rows.

Text, emails, editing → Claude

Write it, cut it, rewrite it to sound human, proofread the style — that's Claude territory (remember the Claude and Gemini lesson: copywriters and editors praise its prose). Free alternative: ChatGPT — the difference is in the shades, and for emails and posts you'll barely see it. If the text lives in Google Docs, it's faster to hand the job to Gemini right in the document.

Long documents and stacks of files → Claude

A 120-page contract, a dissertation, ten PDFs at once — Claude holds a huge context more confidently than the other two. For a one-off short file, free ChatGPT or Gemini will do.

Fresh facts and sources → Gemini or Perplexity

An all-rounder on its own only remembers the world up to its knowledge cutoff, so "what came out this week" and "what does it cost today" aren't for it. When you need current and verifiable, take Gemini (fused with Google Search) or Perplexity (a link to the source under every claim — remember the specialists lesson). Both are free to use.

Code, sites and data tables → Claude and Gemini

Debug an error, write a script, build a site from a description (vibe coding — making apps out of text, no hand-written code) — among developers Claude is a frequent favourite. But summarising a thread, throwing together a formula in Google Sheets, answering an email without leaving Gmail — there Gemini has no competition: it lives inside your services, where other models can't go.

Images, music, video → specialists

A movie-grade poster — Midjourney; a song from a description — Suno; a clip — Veo inside Gemini. But for a social post, a cover or an illustration in a deck, the free generators inside ChatGPT and Gemini are plenty — don't pay until the picture has to sell something.

Studying, analysis, brainstorming → any all-rounder

Explain a topic, throw out ideas, rehearse a conversation — all three are equally good at this. Here you're not picking a model, you're picking whichever is already open in the next tab. That's the whole point of the table: in 8 jobs out of 10 any all-rounder will do, and only a couple of rows really demand a specific name.

One project, three models: what it looks like

Olga is launching a mini-course. Monday: Claude rewrites her notes into human landing-page copy. Tuesday: Perplexity gathers fresh market statistics with links — the kind you're not embarrassed to put on the page. Wednesday: ChatGPT's built-in generator draws the cover, Suno writes a jingle for the promo clip. Thursday: Gemini sends the invitations straight from Gmail and summarises the replies. Not one new subscription beyond her single one — she just handed each row to the right performer.

And don't assume this only works for people with a "project": the same principle rescues a pensioner with a puzzling electricity bill and a shopkeeper with two hundred product cards. None of them starts by choosing a model — they start with the job, and the model turns up to match.

Common myths

You'll hear these three in every conversation about picking an AI, and all three mislead.

• "I need to find one model for every occasion." Even pros keep a combo: one subscription + free tiers for what the others are good at.
• "ChatGPT is the most famous, so it's best at everything." Fame ≠ leading a particular row: long text is Claude, fresh search is Gemini/Perplexity.
• "Switching between models is hard." It's two open tabs and the same task-framing skill, which carries across all of them.

Do this now

Write down your three real jobs for this week and put a row from the table next to each — task type and model. Then take the most frequent of the three and run it on two different models, and compare the answers. Five minutes later you don't have an abstract table from a lesson, you have a personal cheat sheet for your work.