ChatGPT, Claude or Gemini? Which AI assistant fits which task

The setup: six tasks, three assistants, identical inputs

We gave ChatGPT, Claude and Gemini six tasks from our own working day: summarise a 40-page tender, analyse a spreadsheet of revenue figures, write a polite rejection, translate an English technical article into German, repair a faulty Python script, and research a topic with sources.

Each task went into all three tools with exactly the same input, with no subsequent prompt polishing. We judged not the first impression but the effort needed to reach a usable result, the only measure that counts at work.

Some perspective: all three models sit closer together than rankings suggest. The differences show up not in right versus wrong but in how much rework a result costs.

Long documents: a clear lead for Claude

The gap was widest on the 40-page tender. Claude held details from the annex together with the deadlines at the front and pointed out an inconsistency between two paragraphs by itself. The summary was usable without correction.

ChatGPT produced a readable version but dropped two side conditions buried in the annex. Gemini summarised the opening precisely and grew noticeably coarser towards the end.

If you regularly work with contracts, studies or minutes, this is the most consequential decision, which is also why Claude leads our AI assistants & chatbots category.

Tables and numbers: ChatGPT calculates, the others estimate

On the revenue table, ChatGPT played its code execution card: it actually calculated instead of producing numbers from a language model, and showed its working when asked. The deviation from our control calculation was zero.

Gemini also reached the right totals but explained the route less clearly. Claude calculated correctly but formatted the result as prose where a table was needed.

For recurring analysis this is decisive: an assistant that calculates rather than estimates saves you the control calculation, but only if you make it show its working.

Translation: the winner is not an assistant at all

All three models translated the technical article intelligibly, and all three smoothed over terms that were deliberately narrow in the original. The comparison run with DeepL was clearly better: tighter terminology, fewer reformulations, plus a glossary for your own company language.

That is the most important finding of this comparison: for clearly bounded tasks, the specialist beats the all-rounder. More on this in our AI translation & writing tools category.

Research: without evidence, every answer is just a claim

On research, all three assistants produced plausible answers, and all three produced at least one statement that could not be verified. Gemini linked sources most reliably, ChatGPT researched more thoroughly but linked patchily.

Perplexity worked noticeably more cleanly, giving every statement a footnote. Where verifiability counts, a research assistant beats a chatbot, even if it writes less well.

Recommendation: it really does depend on the task

If you have to pick exactly one tool and work mostly with text, take Claude. If you want to handle as many different tasks as possible in one interface, take ChatGPT. If you already work in Google Workspace, Gemini delivers the largest additional benefit, because the integration saves more time than any quality lead.

For organisations with strict data protection requirements, Le Chat with EU processing is worth a look. The checkpoints to clear first are in our GDPR guide.

And the most honest recommendation: use two. The free tiers are enough to get a second opinion on important tasks, and how far that gets you is covered in our article on free AI tools.

In short: Claude for long documents, ChatGPT for numbers and breadth, Gemini for Google Workspace teams, Perplexity for sourced research, DeepL for translation.