The 4 questions that catch an AI cover letter

Recruiters can spot a ChatGPT cover letter in three sentences. These are the four exact things they look for — and how to fix each one in your own draft before you send.

Last updated: May 2026

I asked four recruiters, two hiring managers, and one VP of Engineering the same question: how do you know within 30 seconds that a cover letter was written by ChatGPT? They each gave a slightly different version of the same four answers. The answers are below.

If you've ever used ChatGPT for a cover letter and gotten no replies, the four things below are almost certainly why. The fix isn't to stop using AI — the fix is to grep your own draft for these four patterns and rewrite them out before you send. Each one takes 30 seconds.

Question 1: Does the opening sentence start with “I am writing to express my strong interest”?

This is the single most common AI tell. ChatGPT reaches for it because it's the most common opening across the millions of cover letters in its training data. Recruiters have read it ten thousand times. They cut the letter in 8 seconds.

The variants are all equally bad: “I am writing to express,” “I would like to express,” “I am thrilled to apply for,” “I am excited to submit my application for.” Same pattern, same tell.

The fix is the opposite of what most candidates do, which is to spend more time on the opening sentence. Spend less. Cut the entire first sentence and replace it with whatever your actual second sentence was — the one that has substance. If your second sentence isn't strong enough to be the first, the letter has bigger problems than the opener.

Better openers usually do one of three things: name a specific recent thing the company did, reference a specific person there by name in context, or open mid-thought with a strong claim. Example from a real letter that got a phone screen at a fintech:

Three months ago I sat through your CTO's QCon talk on the migration off the legacy payment processor. The way she framed the rollback architecture — instrument first, ship second — is the principle I've been trying to get my own team to adopt for the past year.

That opener requires research the model can't do for you. But the research takes 10 minutes per application, and the response rate roughly doubles on letters that have it.

Question 2: Are there three-word adjective stacks anywhere in the letter?

“Dynamic, fast-paced environments.” “Results-driven, detail-oriented professional.” “Innovative, scalable, cost-effective solutions.” Three abstract adjectives in a row, none of them concrete.

Real writers don't reach for three abstract adjectives. They reach for one specific noun. AI reaches for stacks because the model's training data is full of corporate writing that uses stacks, and the stacks add fluency without adding information.

The fix: open your draft, search for any comma between two adjectives, and rewrite. Each three-adjective stack should become either one specific noun (preferred) or a concrete example.

Before: “I thrive in dynamic, fast-paced, results-driven environments.”

After: “The last team I worked on shipped a new product every six weeks. That cadence is the one I want to be on.”

One specific concrete fact > three abstract adjectives. Every time.

Question 3: Is there a “proven track record” claim with no specific proof?

The phrase “proven track record” without the actual proof is one of the highest-confidence AI tells. ChatGPT writes it because it sounds professional and confident. Recruiters grep for it because it's a tell that the rest of the letter is unsubstantiated.

The pattern: “I have a proven track record of [vague category]” followed by no specific evidence. Examples:

• “I have a proven track record of driving cross-functional collaboration.”
• “A demonstrated history of exceeding expectations.”
• “Consistently recognized for delivering high-impact results.”

None of these mean anything. They sound like they do, which is the problem.

The fix: if you wrote “proven track record of X,” you have two choices. Either follow it with one specific number, project, or outcome that proves X — or delete the phrase. There is no third option that doesn't sound like AI.

Better: “Last year I led the migration of our analytics pipeline from Redshift to BigQuery, which cut our reporting latency from 4 hours to 9 minutes.” That sentence proves “track record on data infrastructure” without ever using the phrase.

Question 4: Does the closing read “Thank you for your consideration. I look forward to hearing from you”?

Universal closer, universally skipped. Recruiters' eyes pass over it without registering content. Worse, it signals that you ran out of things to say and reached for the standard template.

The fix: replace the closer with one sentence specific to this exact role. The strongest closers do one of two things — either offer something concrete the recruiter could use, or ask a specific question that's easy to reply to.

Examples that work:

• “If it'd be useful, I can send the deck from the migration project I mentioned — happy to walk through it on a call.”
• “Curious whether the team is solving the same problem we ran into in Q2: hot-path latency under multi-region failover. If so, I'd love to compare notes.”
• “Worth a 20-minute conversation? Happy with a no — just thought the overlap was worth asking about.”

Each one ends the letter with something the recruiter can actually respond to, instead of a generic sign-off they skim past.

The underlying pattern: AI reaches for the median sentence

All four of these patterns are versions of the same underlying problem: ChatGPT, by design, produces the median sentence for any given context. The median cover letter opening is “I am writing to express.” The median way to fill space is a three-adjective stack. The median way to claim competence is “proven track record.” The median closer is “thank you for your consideration.”

The median sentence is, by definition, the one every recruiter has read most often. Which makes it the one their eyes skip past fastest.

The fix isn't to stop using AI. The fix is to give AI specific real inputs it can't make up — a specific recent thing the company did, a specific project of yours with a specific outcome, a specific question you want answered. With specific inputs, the model produces a specific letter. Without them, it falls back on the median sentence in every position.

This is why the prompt structure matters more than the prompt wording. The free cover letter prompt on SnipPrompts has a refuse-to-write-generic gate built in — it won't produce a letter unless you give it one specific real reason for wanting this exact role at this company. That single constraint flips the model from median-sentence mode to specific-letter mode.

How to grep your own draft in 90 seconds

Before you send any cover letter you used AI for, do this 90-second pass:

1. Read sentence one out loud. Does it start with “I am writing to” or any variant? If yes, delete it and start with sentence two.
2. Search the draft for commas between adjectives. Any stack of two or three adjectives with commas between them — rewrite as one concrete noun or one specific example.
3. Search for the literal phrase “proven track record.” If it's there, follow it with one specific number/project/outcome or cut the phrase.
4. Read the last sentence out loud. If it's “thank you for your consideration” or any variant, replace with one sentence specific to this role — offer something concrete or ask a specific question.

If you fix all four, you've removed the four highest-confidence AI tells from the letter. What remains will read like a human wrote it, even if the model did most of the work.

None of this is about cheating the system or hiding the use of AI. Spell-check didn't require disclosure; AI for formatting doesn't either. The point is that the median ChatGPT output is the one recruiters specifically grep for — not because they hate AI, but because the patterns above signal unsubstantiated, generic, or template-driven writing. Your job is to make sure the letter isn't any of those, regardless of how it was drafted.

If you want more

The pillar guide on the same topic is how to use ChatGPT for cover letters without sounding like AI — longer, with more worked examples. The free prompt with the refuse-to-generic gate is the cover letter prompt. The deeper version, with the 5-question framework for pulling a specific reason out of any application, is in the bundle cover letter module.