← The journal
Explainer / AI tools

How to Use ChatGPT for Job Applications Without Inventing Experience

Learn how to use ChatGPT to tailor your CV and write a cover letter using experience you can back up. Includes a fictional job advert, a worked example and a copyable prompt.

By Henry Chimanzi | FrontierAI.fi

If you’re choosing between Claude and ChatGPT for writing and research, start with a piece of work you actually need to finish. A polished email is useful only if it says what you meant. A fluent research summary is useful only if you can trace its claims back to the source.

My rule is simple. For writing, choose the tool whose first draft needs fewer edits. For research, choose the one whose claims you can verify faster. You may end up using different tools for the two jobs.

Claude vs ChatGPT for writing

Imagine you need to chase an overdue invoice without damaging the relationship. A good draft should make the amount, deadline and next step clear. It should also sound like something you would actually send.

Here’s a prompt to try in both tools.

“Write an email to a client about an overdue £480 invoice. Ask them to pay by Friday or tell me when payment is expected. Keep the tone friendly and direct. Use 60–90 words. Do not add an invoice number, bank details, late fees or any other facts I haven’t supplied.”

Read each first draft aloud. Does the request get buried under polite filler? Did the chatbot turn Friday into a vague “at your earliest convenience”? How much would you change before sending it? Those questions tell you more about everyday usefulness than a general verdict on which chatbot sounds better.

If you’re writing a CV or cover letter, see our [guide to using ChatGPT for job applications](https://frontierai.fi/news/chatgpt-for-job-applications/) for a worked example of tailoring an application using experience you can back up.

Claude vs ChatGPT for summarising documents

A useful summary preserves the distinctions that affect a decision. If a report says a policy is proposed, the summary should still say it’s proposed. If a finding applies to one group of people, the summary should keep that limitation visible.

Give both chatbots the same public report or article and try this prompt.

“Summarise this document for someone who has two minutes to read it. Give me three findings, one important limitation and one question the document leaves unanswered. For each finding, point to the relevant page or section. Use only the supplied document.”

Pick two of the most important claims in each answer and check them against the original. If you have to hunt through the document to figure out where a claim came from, factor that time into your comparison.

Claude vs ChatGPT for research

A source link is the start of a check. Open it and see whether the page actually supports the claim beside it. Pay particular attention to dates, eligibility conditions and the difference between an announcement and something already in effect.

Try this with a real question you need answered.

“Answer this question using original public sources where available. For each main claim, give the source title, date and link, and explain what the source establishes. Separate confirmed facts from any inference. If you cannot verify something, say so.”

Use the same question in both tools and record whether web search was available in each session. Then check whether the links open, whether the sources say what the answer claims and whether important uncertainty is made clear.

What our Finnish test found

In a separate FrontierAI.fi test, ChatGPT scored 17/18 and Claude scored 16/18 across Finnish meeting messages, translation and customer support. Both scored full marks on the delivery notice translation. On the meeting task, ChatGPT met the word limit but repeated a question. Claude produced a polite reply that fell below the requested minimum.

That one-point gap came from three specific prompts, with one response per chatbot and no recorded model versions. The test offers a concrete example of why it pays to check the instructions as closely as the prose. You can read the exact prompts and answers in “ChatGPT vs Claude vs Mistral Le Chat for Finnish” in the FrontierAI.fi Journal.

Which should you choose?

If writing takes up most of your time, start with three messages you regularly write and see which tool saves you more editing. If research is the priority, compare how long it takes to verify each answer against the original sources. If you work in both English and Finnish, include both languages in your test.

FrontierAI.fi has published results for the Finnish comparison. The writing, document and research prompts in this article are suggested tests you can run yourself.

Sources & further reading

  1. FrontierAI.fi — Finnish chatbot comparison with original test responses Source published 7 Nov 2026
  2. FrontierAI.fi — How to Use ChatGPT for Job Applications (related guide)
  3. OpenAI — ChatGPT product overview
  4. Anthropic — Claude product overview

Sources reviewed 7 Nov 2026. Analysis and reading guidance are editorial interpretation; provider claims are not independently benchmarked.