Claude Prompts for Job Applications and Cover Letters

The short version: Claude is better than ChatGPT at sounding like a human on paper, but only if you feed it your actual sentences, not a job description and a hope. Give it your real work history, one job posting, and three rounds of editing, and you'll get a cover letter that doesn't read like every other cover letter landing in that inbox this week.

Why Claude specifically, and not just "AI"

I use Claude for anything that needs to sound like a person wrote it, because it doesn't default to the bouncy, over-punctuated tone ChatGPT slides into when you're not watching it. Cover letters are a tone problem before they're a content problem. A hiring manager reading forty applications a day can spot "I am thrilled to apply" from across the room, and most AI output defaults straight to it unless you stop it.

That said, Claude isn't magic either. It will happily invent a statistic about your old employer's revenue growth if you let it, and it will write "passionate about innovation" if you give it a thin brief. The tool isn't the fix. The prompt is.

The groundwork prompt: build your materials bank first

Before you ask for a single cover letter, spend twenty minutes building what I call a materials bank. Paste your CV, two or three performance review lines if you have them, and a rough list of achievements with numbers attached, into a Claude project or a single long message. Then use this prompt:

• "Here is my CV and some notes on my achievements. Rewrite this into a clean summary of my experience, organised by skill area rather than by job title, and flag anywhere I've only given you vague claims without numbers so I can fill them in."

That last instruction matters more than people think. Claude will tell you "you say you improved customer retention but you haven't given me a figure" and that's your cue to go find the figure, not let the AI make one up.

The job-specific cover letter prompt

Once you have your materials bank, paste the full job posting and use something close to this:

• "Using my materials bank above and this job posting, draft a cover letter. Pull two or three specific things from my experience that map directly to what they're asking for, not a general summary of my career. Open with a sentence about the work, not about how excited I am. Keep it under 300 words. Write in plain British English, no corporate filler, no words like 'passionate' or 'use'."

The under-300-words instruction is doing real work here. Most AI-drafted cover letters run to 450 or 500 words because the model has no sense of a hiring manager's patience. Cap it explicitly or you'll be cutting it yourself anyway.

A worked example

Say you're applying for a marketing manager role at a 40-person SaaS company, and the posting mentions "experience running paid campaigns on a limited budget" and "comfortable reporting to the leadership team." Your materials bank has a line about running a £3,000-a-month Google Ads budget across two product lines. A weak AI-generated letter would say "I have extensive experience managing paid advertising campaigns." A prompted one, using the format above, would say something closer to "I ran paid search across two product lines on a £3,000 monthly budget and cut cost-per-lead by a third over six months by killing the campaigns that weren't converting." Same fact, completely different letter. The second version gets read twice. The first gets skimmed and binned.

Prompts for tailoring the same letter to five different roles

If you're applying for ten jobs a week, which is realistic if you've been made redundant or you're changing industries, you don't have time to start from scratch each time. Use this follow-up prompt on your base letter:

• "Here's the cover letter I wrote for a similar role. Here's the new job posting. Adjust the examples I used so they match what this company is asking for, but keep my voice and the structure the same. Don't just swap in buzzwords from their posting, change which of my achievements you're highlighting."

That instruction to swap achievements, not just vocabulary, stops Claude from doing the lazy thing, which is mirroring the job posting's language back at you. Hiring managers who wrote the posting themselves notice when their own phrases come back at them. It reads as pandering, not fit.

The uncomfortable bit about AI-written cover letters

Here's what most guides on this topic won't say plainly: the hiring manager on the other end is also using AI, probably to screen applications before a human reads them, and sometimes to write the job posting itself. So you're not writing to impress a person who'll admire your prose. You're writing to pass an automated filter and then not sound generic to the human who reads what's left. That changes the brief. It means keyword matching from the job posting still matters even though it feels cynical, and it means the "human" parts of your letter, the specific numbers, the slightly odd detail about why you want this particular company, carry more weight than they used to, because they're the bits a filter can't fake and a bored reader notices.

It also means a flawlessly polished AI letter can work against you. If your LinkedIn profile and interview manner don't match the tone of a letter that reads like a copywriter wrote it, that gap shows up in the interview and it shows up badly. Use AI to get you to a strong first draft, then roughen it up with your own phrasing in two or three places. Deliberately.

Prompts for the parts people forget

Cover letters get all the attention, but two other documents matter just as much and get far less AI help:

• The follow-up email after applying. Prompt: "Write a short follow-up email for this application, to send a week after submitting, that adds one new piece of information rather than just asking if they've made a decision."
• The LinkedIn message to a connection at the company. Prompt: "Write a two-sentence LinkedIn message to someone who works at [company], referencing the role I'm applying for, that doesn't ask them to refer me outright but opens the door to a conversation."
• The thank-you note after an interview. Prompt: "Based on these notes from my interview, write a thank-you email that references one specific thing we discussed, not a generic thanks for your time."

These three documents, done badly, undo good cover letter work. Done well with the same specificity rule, they compound it.

Checking Claude's work before you send anything

Run this prompt on every draft before you submit it:

• "Read this cover letter back and tell me which sentences sound like generic AI output and which sound specific to me. Be blunt."

Claude is surprisingly good at marking its own homework when asked directly, and it will often flag its own clichés if you ask it to be critical rather than to just "improve" the letter again, which tends to make things worse, not better, because it adds more adjectives instead of removing them.

If you want a longer breakdown of exact prompt wording for different seniority levels and industries, the fuller guide to Claude prompts for job applications goes further into role-specific examples. And if your CV itself needs the same treatment before the cover letter stage, there's a dedicated piece on writing better resumes using AI prompts worth reading first, because a strong cover letter can't fix a weak CV underneath it.

What this costs you in time, realistically

Building the materials bank takes 20 to 30 minutes, once. After that, a tailored cover letter using these prompts takes about 10 to 15 minutes end to end, including the edit pass, versus the 45 minutes to an hour it takes most people to write one from a blank page. If you're applying for 15 roles, that's the difference between finishing your search in a fortnight or dragging it out over two months out of sheer dread of the blank page. The time saving is the actual point here, not the writing quality on its own, because the writing quality was always achievable, people just didn't have an hour to spare per application.

If you're applying for a technical role specifically, the framing shifts again, because technical hiring managers read cover letters looking for evidence you understand the actual problem, not evidence you can write nicely. There's a separate piece on writing a cover letter for a software engineer role that goes into what that evidence looks like in practice.

Where this fits with other AI writing work

None of this is unique to job hunting. The same discipline, feed it real material, give it a word limit, make it justify specificity, applies whether you're asking AI to draft a letter for a landlord dispute or you're using ChatGPT prompts for something entirely unrelated to work. The tool changes. The rule that bad input makes bland output doesn't.

Frequently asked questions

Can recruiters tell if a cover letter was written with Claude?

Often, yes, if you use the raw output, because generic AI phrasing has a recognisable rhythm. The fix isn't hiding that you used AI, it's feeding it specific numbers and achievements so the content itself isn't generic, then editing two or three sentences into your own voice before sending.

Should I tell an employer I used AI to write my application?

There's no obligation to disclose it, in the same way you wouldn't disclose using a spellchecker. What matters is that the content is accurate and that you can speak to everything in it confidently in an interview, because that's where an unprepared AI-assisted applicant usually gets caught out.

Is Claude better than ChatGPT for cover letters?

For tone, generally yes, Claude tends to produce less bouncy, less exclamation-heavy prose by default. For keyword matching against applicant tracking systems, the difference is smaller, and the quality of your prompt and your input material matters more than which model you pick.

How long should an AI-assisted cover letter be?

Keep it under 300 words. Most AI tools will draft closer to 450 to 500 words unless you cap it explicitly in the prompt, and hiring managers reading dozens of applications a day rarely read past the first ten lines of a long one.