Comparing Tools

There is no single best AI tool — only the best one for a specific task, with a specific feature, on the day you ask. Claude offers Free, Pro, and Max tiers and an Opus/Sonnet/Haiku family. ChatGPT spans Free, Plus, and Pro tiers with built-in Canvas, GPTs, and data analysis. Gemini ships Free, Pro, and Ultra tiers and embeds into Gmail and Docs.

Because each vendor ships new models on its own schedule, a verdict you write today expires in weeks. So this lesson teaches a repeatable method, not a frozen leaderboard — three parts you re-run in five minutes whenever a tool updates: a fair harness, a scoring rubric, and a personal decision matrix from your own results.

A fair test means every tool gets the same chance. Open three tabs, start a fresh chat in each, and hold five things steady: the prompt (identical text, copy-pasted), the model (a comparable tier in each), the context (no memory carried over), the attachments (same file, or none), and the starting state (web browsing off everywhere, or on everywhere).

Pick the model deliberately: in each dropdown, choose the most capable model your plan offers — reasoning against reasoning, not flagship against lightweight. Claude exposes Opus, Sonnet, and Haiku here. Then make the prompt demanding enough to separate them:

> From the three customer-churn findings below, produce: a one-paragraph
> summary; a table ranking them by urgency with a reason for each; and
> two actions for the highest-urgency one. Under 250 words, British spelling.

That one prompt tests comprehension, format control, prioritisation, brevity, and a spelling rule at once.

"It just felt better" is not data. Score every response 1 to 5 on five dimensions so the feeling becomes a comparable number:

Score the churn prompt this way and totals usually land close, but the shape differs: one tool nails instructions, one format, one speed. Weight the dimensions for your work first — a lawyer doubles accuracy, a marketer creativity.

One prompt is a data point; five categories is a pattern. Run one per category through all three tools (same spreadsheet for analysis):

> RESEARCH: The consensus on intermittent fasting for over-50s — three findings, each with a caveat.
> WRITING: A 120-word LinkedIn post announcing a new accessibility feature. Warm but professional. No hashtags.
> ANALYSIS: From the attached sales CSV, find the three slowest months, explain the cause, recommend a promotion.
> CODING: A Sheets formula flagging any row where column D is over 20% above column D's average. Explain in plain words.
> CREATIVE: Five unexpected metaphors for "learning AI tools" — no cars, journeys, or magic wands.

Each stresses a different muscle — synthesis, voice, reasoning over data, technical-but-explained output, open invention. Record a winner per category. Research usually favours whichever tool has live browsing and analysis whichever runs the file; writing and creative come down to taste.

Often the smarter model loses because the other tool has a feature that fits the job — the model writes the answer, but the feature decides whether you can use it. Check each picker before judging:

A feature flips a verdict when it changes what you can do: a tool that runs Code Interpreter on your spreadsheet beats one that only describes it, and drafting inside Gmail beats a better answer you copy-paste in.

Turn your scores into a one-glance map. Combine your category winners (Step 4) with the feature advantages (Step 5) into one matrix:

Pin it where you will see it. But it has a shelf life: new model versions ship every few weeks, and one can move a tool from third to first. So re-run the bake-off whenever a vendor ships a new flagship, and quarterly regardless. Because you kept the identical prompt, each re-run is comparable — revealing trend lines no review site can give you about your own work.