All questions

Tool comparisons

Can I write my cover letter with ChatGPT?

Updated:

In short

Yes — and for the language itself ChatGPT beats any template tool. Language, though, is the easy half of a Swiss cover letter. The hard half is the reference: a letter to one named employer about one named advert. Pasting the link is usually not enough. On 24 August 2026 we opened the thirty applicant systems behind most of our adverts with a plain page request: on about one eighth of that advert volume, nothing came back but an empty shell. Paste the advert text and you get a good letter. Paste only the link and you get a letter about nothing.

That answers the question, and the useful part starts here. A chat produces paragraphs. A Swiss application is a document — your postal address, the company address, a date, a subject heading, a salutation, a closing greeting and a signature, in the words of the federal job-seeker pages on arbeit.swiss. Between the last sentence in the chat and the file an employer receives sits work nobody mentions when they say "just use AI".

So this is not a comparison of writing styles. It is a comparison of two workflows: the chat you carry everything into by hand, and an application tool attached to an index of Swiss adverts. Everything stated here about ChatGPT was read on OpenAI’s Swiss pricing page and in OpenAI’s own crawler documentation on 24 August 2026. Last checked: 24 August 2026.

How much text one chat takes at once
How much text one chat takes at onceFree12 pagesGo40 pagesPlus40 pagesPro250 pages

Maximum input for the fast models per plan, in pages of text. Read on OpenAI’s Swiss pricing page, 24 August 2026.

  • Paste the advert text, not the advert URL. Across the applicant systems we checked, about 12.5 per cent of the advert volume returned no text to a plain page request — Workday-hosted career sites came back completely empty, and several portals returned a few dozen characters of frame.
  • This is rarely a permission problem. OpenAI documents that "ChatGPT-User" is the agent behind user-initiated fetches and that, because a person triggers them, robots.txt rules may not apply. The constraint is whether the page contains any text at all before a browser assembles it.
  • The free tier is big enough for one application. OpenAI’s Swiss pricing page lists a maximum input for the fast models of roughly twelve pages of text on Free and roughly forty on Go and Plus. One advert plus a CV fits in twelve pages; three reference letters on top do not.
  • A model usually knows nothing about the employer. Of the employers in our index, 45.4 per cent have exactly one open advert and 70.6 per cent have three or fewer; the ten largest account for 21.4 per cent of adverts. Most Swiss employers have never been written about outside their own site.
  • Treat the job title as a quotation, not a writing task. In our index 31.4 per cent of advert titles carry a marker such as (m/w/d) and 21.9 per cent state a workload percentage — exactly the parts a model smooths away when it "improves" your subject line.
  • The finished document is made outside the chat: layout, page break, file name, PDF. Ours is made inside the document — but the download belongs to the seven-day free trial, CHF 15 a month afterwards. ChatGPT’s free tier costs nothing and you can copy the text straight out. That is the honest price difference.

Who to choose

The difference is not the tool, it is what the tool starts from: with us the advert is already there. You pick the job straight from our own Swiss job index — with ChatGPT in.

Choose ChatGPT if

  • ChatGPT — unbeatable at rewriting. Change the tone, cut it down, redraft one paragraph ten times — no form-driven generator does that as well.

Choose SwissJobs.app if

  • Pick the job directly — We run the job index ourselves. The advert text is already loaded, and everything else is written against exactly that text — no searching, no copying, no pasting.

  • Four tools, not one — CV, motivation letter for the same advert, a CV check against that advert, and interview preparation — one account, the same data.

  • The RAV sheet — Your job-search efforts are filled from your application tracker and printed as the official SECO form. No other tool in this comparison has it.

  • Word export, no watermark — PDF and DOCX, both clean. Many tools hand you a text file on the free plan, or stamp a watermark on the document.

  • Start without an account — The tools open without signing up. The account is asked for to save and from the second document — the first download needs none.

Feature comparison

FeatureSwissJobs.appChatGPT
Writes a motivation letterYesnot stated
Start without an accountYesnot stated
Tailored to a specific advertYesNo
Advert text from its own job indexYesNo
AI prompt edit in secondsYesnot stated
Full manual editingYesYes
PDF export, no watermarkYesnot stated
Word / DOCX exportYesnot stated
Swiss cover-letter conventionsYesnot stated
Matching CV for the same advertYesNo
LanguagesDE / EN / FR / ITnot stated
Built in SwitzerlandYesnot stated

As of 4 September 2026. "not stated" means it could not be read on their own site — not "does not exist". Our rows come from the code (lib/stripe/trialConfig.ts) and from walking the live flow, not from marketing copy.

Pricing side by side

What you paySwissJobs.appChatGPT
Monthly priceCHF 15 — all six tools, one subscriptionnot stated
Yearly priceCHF 95 a year (= CHF 7.92 a month) — the same six toolsnot stated
Free trialOur seven days are the full product, not a limited preview.7 days, full access, cancel any timenot stated
What works without payingWe print the limit, because that is the number a comparison usually leaves out.Tools open without an account; free account: 5 documents/day, 10 per rolling weeknot stated
CurrencyCHFnot stated
BillingWatch for four-week billing periods anywhere: that is 13 charges a year, not 12.Calendar month, cancel any timenot stated

As of 4 September 2026. "not stated" means it could not be read on their own site — not "does not exist". Our rows come from the code (lib/stripe/trialConfig.ts) and from walking the live flow, not from marketing copy.

What ChatGPT genuinely does better here

The strength belongs first, because nobody disputes it. A language model writes more freely than any template product. It strips two subordinate clauses out of a laboured paragraph, it turns "I am highly motivated" into a statement with content the moment you supply the content, and it changes language without effort. If you need the same letter in French for an application in the Romandie, you have a usable draft in two minutes — one that still deserves a read from somebody who actually speaks it.

It is also endlessly patient, and patience is where most letters fail: not on the first draft but on the fourth revision. From the Plus plan upward, the pricing page lists projects, scheduled tasks and custom GPTs. Park your CV and your standing instructions in a project once and you stop re-explaining yourself every time. That is a real advantage over any tool where each letter starts from nothing, and we name it because comparisons written by tool vendors usually leave it out.

Stop where phrasing turns into claiming. A model fills gaps with whatever sounds plausible in that position: if the draft says you ran the shift plan for fourteen people, it says so for the rhythm of the sentence, not because your reference letter says it. Every number in the letter has to come from you — and so does every statement about this one employer.

Why the link on its own rarely works

The obvious workflow is: find the advert, copy the address, drop it into the chat, ask for a cover letter. We wanted to know what actually arrives at the other end, so we measured it on 24 August 2026. We took the thirty applicant systems our adverts point to most often — together 44 per cent of the index — and requested three different adverts from each as a plain page request, the way any program does when it opens an address.

The result: on 87.5 per cent of that advert volume the advert text came back. On 12.5 per cent it did not. Career sites running on Workday returned a completely empty page; two applicant portals returned a few dozen characters of scaffolding; the cantonal jobs portal of Fribourg returned its menu and the job title but no advert text. Those pages assemble themselves in a browser, and whatever opens them without one holds an empty shell.

What this is not is a block. OpenAI’s own documentation says the ChatGPT-User agent is "used for certain user actions in ChatGPT and Custom GPTs" and that, because such actions are initiated by a user, robots.txt rules may not apply. Permission is rarely the obstacle. The obstacle is that there is no text on the page that was fetched — and a model does not report this as an error. It writes the letter anyway, out of what it knows about jobs of that kind in general, which is the definition of a generic letter.

The remedy costs one gesture: open the advert, select the text, paste it. Then the model works from the vacancy instead of from an impression of it. And if you found the job on SwissJobs.app, the step disappears, because the advert text is already sitting in the document.

How much you can actually paste

A second objection circulates and it is mostly wrong: that the free tier is too small for job applications. OpenAI’s Swiss pricing page gives a maximum input for the fast models of roughly twelve pages of text on Free, roughly forty on Go and Plus, and roughly 250 on Pro. A Swiss advert rarely runs past two pages and a CV rarely past three. For one letter about one advert, the free tier is enough, and saying otherwise would be scaremongering.

It gets tight the moment you add the material that makes a letter checkable: two or three reference letters, an earlier application, the advert on top. Then the ceiling stops being theoretical, and what falls out of the window falls out silently. The same page lists "limited messages with uploads" and "limited memory and context" for Free — the free tier is thriftier about holding things than about writing them.

On the Go plan, the page carries its own footnote worth reading before subscribing: this plan may include advertising. It has no bearing on whether a letter turns out well. It has some bearing on what you are signing up for, and it appears in few comparisons.

Practically: one letter per chat, advert text first, CV second, nothing else. A model does not improve because it sees more. It improves because it sees the right thing.

The employer nobody has written about

The sentence that sinks most cover letters is the sentence about the company. Ask a model for it and you get a paragraph about innovation and market position, which is roughly true for a household name. But this market is not made of household names. Of the employers in our index, 45.4 per cent have exactly one open advert and 70.6 per cent have three or fewer, while the ten largest together account for 21.4 per cent of all adverts. The typical Swiss advertiser is a firm with one or two vacancies that nobody outside its own website has ever described.

The practical consequence: do not expect company knowledge from the model. Give it the one reliable source that exists — the advert. What stands there has been checked by the employer: the location, the tasks, the workload, the department, sometimes the person who will call you back. What is not in the advert does not belong in the letter unless you read it somewhere yourself.

The reverse test is just as useful. If a draft contains a claim about the employer that you cannot trace to a line in the advert or on the company site, delete it. It is decoration, and it is the sentence that becomes expensive in the interview.

The subject line is a quotation

There is one line where a good model reliably does the wrong thing: the subject heading. It writes. "Fachspezialist Payroll 80–100% (m/w/d)" becomes "Application for the payroll specialist role", which reads better and identifies nothing. In our index 31.4 per cent of titles carry a marker such as (m/w/d) and 21.9 per cent name a workload in per cent — and those are the first parts to disappear when a draft is polished.

Inside a company with several open roles, that line decides which pile your letter lands on. Copy the title exactly, workload and reference number included. It is the one line in the letter where copying beats writing.

The salutation follows the same rule. If the advert names a person, use the name. If it does not, the general salutation is correct — an invented name is far worse than an impersonal one, and a model will invent it happily when you ask for a "more personal" version.

From chat window to letter

Once the text stands, the unglamorous part begins. The federal job-seeker pages list the parts of a cover letter as your postal address, the company address, the date, the subject heading, the opening salutation and the closing greeting with your signature. A chat gives you some of those, usually in a form whose spacing collapses when it is pasted into a word processor. Then come the page break, the typeface, the file name and a PDF that stays on one page.

That step is where we differ. The letter is a document, not a conversation. The job description stays stored with the document and is sent to the AI again with every edit, so it does not fall out of memory when you change the tone on the third pass. The letter can be linked to one of your CVs so that dates and register match, and what you end up with is a file rather than a column of text.

The price belongs in the same paragraph. Writing and revising work with a free account; downloading the finished file belongs to the seven-day free trial, CHF 15 a month after that. ChatGPT’s free tier costs nothing and you can select the text and take it with you. If you are content to handle layout and file yourself, we save you nothing — that is where the chat wins outright.

Who each one suits

Use ChatGPT if you are writing one or two applications, you already have the advert text in front of you, and you are comfortable in Word or Pages. Use it above all if your difficulty is linguistic — if you know what you want to say and cannot land it in German, French or English. And use it if you want to keep asking in the same window: about the occupation, the sector, the interview. No application tool replaces that.

Use us if you apply to Swiss adverts regularly. The copying disappears because the advert is already here; the letter stays attached to the advert instead of scrolling away in a conversation; and the result is a file in Swiss letter form rather than a block you still have to format. By the third application in one week, that is the entire difference.

One check applies to both routes before you send. If the letter states a fact you could not repeat in an interview without hesitating, it does not belong there. The tool decides how fast the text appears. Whether it is true is decided by you.

Employer and job-title shares come from our running count of the Swiss adverts open on SwissJobs.app, as at 24 August 2026; we publish shares, never absolute index figures. For the reachability check we opened, on 24 August 2026, the thirty destinations carrying the most adverts (together 44 per cent of the index) with three different adverts each as a plain page request, and recorded whether the advert text came back; a destination counts as unreadable when at least two of its three requests returned no text. That measures what a page returns to an ordinary request — not what ChatGPT does with it internally. Plan details, input limits and the advertising note come from OpenAI’s Swiss pricing page, and the description of the ChatGPT-User agent from OpenAI’s crawler documentation, both read on 24 August 2026. The list of letter components comes from arbeit.swiss (SECO).

Cover letter for Switzerland: from this advert, free

Sources

Related questions

← All questions

What our job index says about the Swiss market

Computed live from our own index, not quoted from a study. Shares only, as of today.

Language the advert is written in

Deutsch
60%
English
23%
Français
13%
Italiano
3%

Of adverts that state a language requirement, the share asking for

Deutsch
70%
English
43%
Français
21%
Italiano
3%

19% posted in the last 7 days · Largest markets: Zürich 18% · Bern 10% · Genève 5% · Basel 5%