Asking ChatGPT “write me a cover letter for a banking internship” produces something that looks like a cover letter. That's exactly the problem. Recruiters see 200 letters per week — they detect raw ChatGPT output in 5 seconds: too-smooth phrasing, repeated vocabulary, suspiciously balanced structure, zero real specificity. A raw AI letter isn't a bad letter. It's an invisible letter.

This guide shows you how to use ChatGPT (or Claude, or Mistral) to go from invisible to interview-landing. You'll find: why 80% of prompts fail, the structure of a working prompt in 5 steps, 6 copy-paste prompts for different use cases, a concrete before/after, the 5 signals recruiters detect in AI letters, and the recommended workflow to combine ChatGPT, your own writing, and a human mentor.

TL;DR — the 60-second version

Bad prompt: “Write a cover letter for an M&A internship at Lazard.” Output: generic, trashed. Good prompt: “You are a former M&A associate at Lazard. Here is my CV (pasted). Here is the offer (pasted). Write a 4-paragraph letter, 320 words max, with this structure: P1 about a recent Lazard deal, P2 with two numerical proofs from my profile, P3 about what I bring to the team in 6 months, P4 dated availability. Avoid clichés ‘industry leader’ and ‘rigor.’” The result is 3x better. But — important — it's still a draft. Rewrite P1 and P3 yourself. Those are the two paragraphs ChatGPT always gets wrong.

Why 80% of ChatGPT prompts fail

The problem isn't ChatGPT. The problem is candidates ask it to write for them without giving the ingredients to produce something usable. Three recurring mistakes.

Mistake 1: prompt too short

“Cover letter banking internship” — 4 words. ChatGPT has zero context on you, the company, or the specific role. It produces the average letter from its training set. Result: generic, interchangeable with 5 other letters.

Mistake 2: no role / no constraints

Without “You are [X],” ChatGPT defaults to a vaguely-formal English-teacher voice. Without “320 words max,” it writes you 600 words. Without “avoid these clichés [list],” it puts “industry-leading” and “rigor” in the first three sentences automatically.

Mistake 3: no raw material

“Write me a letter” without pasting your CV or the offer. ChatGPT invents. You get a fictional letter with skills you don't have and references to projects that never happened. If you send it as-is, you're lying inadvertently — detectable in the interview's first minute.

The good prompt in 5 steps

A prompt that produces a usable letter contains five blocks, in this order.

Step 1 — Give a role (system prompt)

“You are a former [role] at [company]. You've hired X interns over the last 5 years. You know the format that recruiters at [company] actually value.”

Why it works: ChatGPT adapts its tone to the identity you give it. A “former M&A associate at Lazard” produces different text than a “career coach.” More specific = better grounded in the reality of the role.

Step 2 — Paste the context (your CV + the offer)

Three context blocks to provide:

The 3rd block is critical. Without it, ChatGPT invents a generic motivation. With it, it integrates it into P1 and the letter becomes personal.

Step 3 — Give the structure

ChatGPT defaults to 3 long, badly-calibrated paragraphs. Impose the “Vocacia canvas” in 4 paragraphs:

Target total: 270-340 words. Without this cap, ChatGPT overflows.

Step 4 — List things to avoid

ChatGPT has detectable linguistic tics. Ban them explicitly.

“Avoid these phrases: ‘industry-leading,’ ‘strong culture,’ ‘rigor,’ ‘team player,’ ‘passion for,’ ‘dynamic,’ ‘solid skills,’ ‘awaiting your favorable reply,’ ‘I am writing to express.’ Avoid bullet lists — use full sentences. Avoid the words ‘synergy,’ ‘collaborative,’ ‘impactful.’”

Step 5 — Specify output format

“Write the letter in formal but simple English. No emojis. No asterisks for bold. No introduction from you — give the letter text directly, ready to paste into a PDF. End with ‘[First Last]’ as signature.”

6 copy-paste prompts that work

Six prompts ready to use for the most common cases. Adapt the brackets [X]to your situation. These prompts are the base — you'll then need to rewrite P1 and P3 yourself (see below).

Prompt 1 — Investment banking internship

You are a former M&A associate at [Bank]. You've hired 12 interns in 5 years. Help me write a cover letter for an M&A internship.

My CV: [paste plain text]
The offer: [paste full posting]
What struck me about this bank: [the deal/article/person — 2-3 sentences]

Write the letter in 4 paragraphs, 300-340 words total:
P1 (70 words): opening on the element I gave you, no mention of my profile.
P2 (110 words): two numerical proofs from my CV (use exact numbers).
P3 (90 words): what I bring concretely to the team in 6 months. Mention at least one specific operational competence.
P4 (35 words): availability on [date] + opening to a conversation.

Avoid: “industry-leading,” “rigor,” “team player,” “awaiting your favorable reply.” No emojis, no bold. Give the letter text directly, no introduction from you.

Prompt 2 — Consulting internship (MBB / Big 4)

You are a former manager at [Firm]. You've interviewed interns for 4 years and know what distinguishes an accepted letter from a rejected one.

My CV: [paste]
The offer: [paste]
Reference on the firm: [report/publication/speaker — 2-3 sentences]

Write a letter in 4 paragraphs, 280-320 words. P1 on the report/publication mentioned. P2 with two numerical problem-solving proofs. P3 on my contribution to a client case (diagnostic phase + structuring). P4 availability on [date].

Avoid: “I love solving complex problems,” “industry-leading,” “rigor,” “passion for.” Mention “hypotheses” and “structuring” at least once each.

Give the text directly, no introduction.

Prompt 3 — Marketing / communications internship

You are a senior brand manager at [Brand]. You're hiring a marketing intern with a strong requirement on creativity and impact measurement.

My CV: [paste]
The offer: [paste]
The recent campaign that struck me: [campaign — 2-3 sentences]

Letter in 4 paragraphs, 300-340 words. P1 on the campaign mentioned. P2 with a campaign I managed + reach/engagement/conversion numbers (use my CV numbers). P3 on my social-first contribution and emerging-format scouting. P4 availability on [date].

Avoid: “dynamic,” “passion,” “synergy,” “creative spirit.” Slightly warmer tone than finance/consulting but not familiar.

Give the text directly.

Prompt 4 — Tech internship (PM / engineer)

You are a Senior Product Manager at [Tech company] who has hired 6 tech interns in 3 years.

My CV: [paste]
The offer: [paste]
The recent product launch that struck me: [feature/product — 2-3 sentences]
My side-project (GitHub link if applicable): [short description]

Letter in 4 paragraphs, 280-320 words. P1 on the product launch. P2 on the side-project + a CV experience with precise stack. P3 on my operational contribution to the product team. P4 availability on [date].

Tone: direct, no excessive formalism (tech culture). Mention my technical stack. Avoid: “passion for technology,” “impactful,” “disrupt.”

Give the text directly. End with “Best regards” (not “Sincerely”).

Prompt 5 — Law internship (lawyer / legal counsel)

You are a partner at [Firm], [specialty] department. You regularly hire LLM students for summer internships.

My CV: [paste]
The offer: [paste]
The case/publication from the firm that struck me: [case/article — 2-3 sentences]

Letter in 4 paragraphs, 300-340 words, very formal tone. P1 on the case/publication mentioned with precise analysis. P2 on my research and writing capacity (CV proofs). P3 on my contribution to jurisprudential research + case documentation. P4 availability + mention I attach CV and academic transcripts.

End with “Sincerely” or “Respectfully yours.”

Give the letter text directly.

Prompt 6 — Cover letter for an internship with no experience

You are a senior recruiter who knows how to read an intern application from someone with no professional experience yet. You value academic projects, association involvement, and side-projects as valid proofs.

My CV: [paste — can be short]
The offer: [paste]
My 3 experience substitutes:
1. [Numerical academic project]
2. [Association role with impact]
3. [Documented side-project]

Reference about the company: [precise element]

Letter in 4 paragraphs, 270-310 words. P1 on the company. P2 treats my 3 substitutes as professional experiences — concrete numbers and results. P3 on my ability to turn a project into a measured deliverable. P4 availability.

NEVER say “despite my lack of experience” or “my first professional experience.” Present my projects with confidence as valid proofs.

Give the text directly.

Before / After — same candidate, two prompts

Bad prompt

“Write me a cover letter for an M&A internship at Lazard. I'm a finance student.”

ChatGPT output (typical)

“Dear Sir or Madam,

As a finance student, I am very interested in your industry-leading firm. My academic background gave me solid skills in finance and strategic analysis. I am dynamic, rigorous and have a strong team spirit. I am convinced that your firm is the ideal place for my professional development and that I could put my skills at the service of your team. I have a passion for financial transactions and would be delighted to contribute to your advisory mandates...”

Score: 0. Six clichés in five lines. Zero information about the company, the candidate, or the internship. Letter trashed in 8 seconds.

Good prompt (Prompt 1 above, completed)

“You are a former M&A associate at Lazard. My CV: [LSE Master's, Lazard 6-month internship — €350M divestiture DCF model, Eurazeo 4-month internship — 3 LBOs modeled]. The offer: [6-month M&A internship Industries department]. What struck me: the Eurosic-to-Gecina divestiture in March 2026. Write the letter in 4 paragraphs, 320 words, [full structure], avoid [list of clichés]. Give the text directly.”

ChatGPT output (with this prompt)

“Dear Sir or Madam,

Your firm's advisory on the sale of Eurosic to Gecina in March 2026 made me reconsider what I want from my Master's internship — the simultaneous execution of the carve-out and due diligence in 4 months illustrates exactly the type of mandate I want to work on.

A Master's student in finance at LSE, I'm coming off a 6-month internship at Lazard where I built the DCF and trading-comps valuation model on a €350M mid-cap industrial divestiture. I also spent 4 months at Eurazeo modeling three LBOs. These experiences taught me modeling discipline under deadline and the coordination between M&A advisor, vendor due diligence and legal counsel.

For my internship at your firm, I can take operational ownership of sector comparables and precedent-transaction models for the real-estate segment. My goal is a return offer for the 2026 analyst program.

Available starting June 1, I would be happy to discuss how I can contribute from the first weeks.

Sincerely,

Léa Martin”

Score: 7/10. Much better. But — important — this is not the final version. P1 and P3 are generically good but can be improved. That's where you take over.

The 5 signals recruiters detect in AI letters

Even with a good prompt, ChatGPT leaves traces. Senior recruiters — especially at major firms that see hundreds of applications a week — have learned to spot them. Here's what they look at.

Signal 1: too-balanced structure

An AI letter often has 4 paragraphs of near-identical length (4-5 sentences each) and sentences of homogeneous medium length (15-22 words). A human letter has irregularities: a short paragraph, a long sentence followed by a short one. Break the rhythm manually.

Signal 2: medium-high vocabulary, never simple

ChatGPT avoids simple words (“do,” “see,” “thing”) and prefers more formal synonyms (“perform,” “observe,” “element”). A 100% medium-high letter signals AI. Reinject 2-3 simple words per paragraph.

Signal 3: too-smooth transitions

“Furthermore,” “moreover,” “in addition,” “in this regard” — ChatGPT chains paragraphs with these formal transitions. Humans rarely use them in written informal application letters. Replace them with sentences that directly mention the next paragraph's topic.

Signal 4: zero verifiable concrete specificity

An AI letter says “I contributed to several strategic projects” when a human letter says “I built the DCF model on the Eurosic-to-Gecina divestiture.” If every detail is vague, the recruiter knows it wasn't written by someone who actually did the work.

Signal 5: symmetric closing

ChatGPT ends with “I am convinced my skills will make a difference” or “This opportunity would be a true springboard for me.” These phrases add nothing and signal AI. Cut them. End with a concrete action sentence (dated availability + opening).

When to use ChatGPT vs a human mentor

ChatGPT is useful for

ChatGPT is useless (or counterproductive) for

The optimal workflow: ChatGPT + human writing + mentor

Best practice combines all three steps. Here's the 90-minute sequence.

  1. 15 minutes — company research. Find the specific element that goes in your P1 (recent deal, article, podcast, public executive intervention). The most important prompt input.
  2. 10 minutes — ChatGPT prompt (one of the 6 above). You get a draft in 90 seconds. Read it fully.
  3. 20 minutes — rewrite P1 and P3. P1: replace ChatGPT's paraphrase with your real analysis in 60-80 words. P3: replace generic phrasing with a concrete equation (what you bring to the team).
  4. 15 minutes — hunt for AI signals. Reread looking for the 5 signals above. Break rhythm, simplify vocabulary, kill smooth transitions.
  5. 20 minutes — mentor review. Ideally a mentor working in your target sector. They'll tell you what sounds credible/exaggerated, what to compress, and what's missing on the company angle. This step turns a “7/10 ChatGPT” draft into a “9/10 human” letter.
  6. 10 minutes — formatting and sending. PDF conversion, file naming, final company-name verification.

Total: 90 minutes for a review-ready quality letter. Without the mentor step, expect 70 minutes but quality capped at 7/10.

Vocacia generator vs ChatGPT: when to use which

Vocacia provides a free cover letter generator based on GPT-4 but preconfigured with:

When to use Vocacia generator: you want a quality draft in 60 seconds without writing a long prompt. Ideal if you're applying to several similar internships.

When to use ChatGPT directly: you have a particular case (very niche sector, international application, English letter with cultural specifics) where the custom prompt gives better control.

In both cases, the workflow stays the same: AI output = draft, you rewrite P1 and P3, you hunt AI signals, you get a working mentor to review.

FAQ — common questions

ChatGPT, Claude, or Mistral — which one for a letter?

All three work with the same prompt. Subtle differences: Claude (Anthropic) tends to produce stylistically smoother text but less concrete. ChatGPT (OpenAI) is more versatile and better with precise numbers. Mistral (French) captures formal-French nuances better but has less capacity to integrate international sector context. For an English internship letter, ChatGPT 4 or Claude 3.5 Sonnet are the two best picks.

Do recruiters use AI detectors?

Very few — under 10% of firms in 2026. Current AI detectors (GPTZero, Originality.ai) have 15-30% error rates and aren't considered evidence. But senior recruiters detect by eye — see the 5 signals above. The risk isn't automatic detection; it's human detection by an experienced recruiter.

Do you need to disclose ChatGPT use?

No. ChatGPT for drafting a letter is the equivalent of Word with autocorrect — a tool. The question isn't “did you use it” but “does the letter that comes out represent what you think.” If yes, no need to mention it. If not, don't send it.

Can ChatGPT also write my CV?

For the CV, the effect is riskier than with a letter — the CV must reflect literal facts (companies, dates, numbers). ChatGPT can reformulate bullet points to make them more impactful, but it can't invent information. For a student CV, see our (upcoming) student-CV template guide.

Can you use ChatGPT to prep for an interview?

For technical questions (DCF, leetcode, case frameworks), ChatGPT is an excellent drill partner. For behavioral questions and coaching your personal pitch, no — AI doesn't capture tone, rhythm, eye contact, and can't test your ability to improvise under pressure. The human mock-interview remains essential. See our guide on how to prepare for a job interview.

How much time do you really save with ChatGPT?

First letter in a new sector: ~40 minutes (you go from 3h to 90 min). On similar letters in series: less, because you already have your canvas. The real win isn't time — it's the quality of the first version, which lets you focus your review time on the paragraphs that make the difference (P1 and P3).

What if I have no mentor to review?

Vocacia is built around this need. Hourly sessions with mentors in role — Goldman/Lazard bankers, McKinsey/BCG consultants, Google/Meta PMs, etc. A 30-minute letter-review session typically costs €30-60. At that price, it's the highest-ROI investment in your entire application process.

Recap and next step

ChatGPT is an excellent tool for producing a cover-letter draft in 90 seconds. But a draft is not a final letter. The right workflow combines three steps: structured prompt (5 blocks), human rewrite (P1 and P3), mentor review by someone working in role.

Recruiters detect raw AI letters in 5 seconds. Letters “ChatGPT + human rework + mentor” are the highest-performing on the market — they combine AI productivity and the human calibration that flips a decision.

Use the free Vocacia generator to start from a preconfigured draft, or use the 6 prompts above with ChatGPT directly. Then have it reviewed by a working mentor in your target sector. For the letter content itself (structure, sector templates, mistakes to avoid), read our complete cover-letter guide.