The Cover Letter Standard, Seven Pass or Fail Checks Before You Attach It

This is a standard for one decision: whether a cover letter you've drafted for a specific posting is ready to attach. It's for anyone who's about to hit submit and can't tell if the letter helps or is just filler. It gives you seven pass/fail criteria, a grading procedure you can do in about fifteen minutes, and a checklist. Two people using it on the same letter should reach the same verdict and the same list of fixes.

What does a cover letter have to prove now that polished prose is free?

A cover letter now has to prove things a language model can't make up for you: one result that fits this posting, and facts a reader could check. Fluent, well-aligned prose used to be a signal of effort. Two large studies of an AI drafting tool show it stopped being one once anyone could produce it for free.

Both studies looked at Freelancer.com, which rolled out an AI proposal drafter in April 2023.

• Galdin and Silbert, "Making Talk Cheap." They built an LLM-based measure of how closely an application is tailored to its posting, scoring proposals with Meta's Llama 4 on nine criteria. Their data cover over 2.7 million applications to 61,000 coding jobs. Before LLMs arrived, employers placed a high value on more customized applications. After LLMs arrived, they didn't.
• Cui, Dias and Ye, "Signaling in the Age of AI." They studied more than 5 million cover letters across 100,000 jobs. Access to the tool made cover letters match their postings more closely and raised callback rates (a callback rate is the share of applications that get a response inviting the next step). It also cut the value of that match: after the tool arrived, the correlation between a letter's alignment and its callbacks fell by 51%. Employers turned to other signals, including workers' prior work histories.

Two more findings shape the criteria below. Workers whose writing was weaker before AI improved their cover letters the most, so the quality of the prose no longer separates candidates. And time spent editing AI drafts was positively correlated with hiring success. That's a correlation, not proof that editing causes success. One plausible reading is that editing is where people add private facts a model doesn't have. The standard rewards exactly those facts.

I have no proprietary data on cover letter outcomes. Refolk's index produced nothing usable for this guide. Every figure here comes from public surveys and the two studies above, and I've marked where a source is secondhand.

Who reads the cover letter, and when?

Hiring managers mostly read cover letters, often before the resume. Recruiters at the first screen skip them far more often. So the letter matters more for the shortlist than for the first cut.

Here's what each number does and doesn't tell you:

• Hiring managers read. Resume Genius reports that 83% read most cover letters even where they're not required, and that 45% read the letter before the resume. Among companies that don't require one, 73% "frequently" or "always" read them anyway.
• Recruiters at first screen often don't. The Novorésumé survey splits its answer by role. Across everyone, roughly 39% don't read cover letters at all, while only 22.9% of HR managers skip them. I couldn't retrieve the primary report, so treat the split as indicative.
• A strong letter can rescue a weak resume. In the ResumeLab survey, 83% said a great letter could get a candidate an interview despite a weak resume.
• Reading time isn't measured. The only figure I found is a practitioner's estimate of about 20 seconds per letter. No survey with a stated sample size measures it. A widely quoted "60 to 90 seconds" figure is unverified.
• The 72% figure is unverified. Secondary write-ups say 72% of hiring managers expect a letter even when it's marked optional, but that number doesn't appear on the primary survey page.

That 45% is why the opening proof point is the first check after the posting's instructions. For those readers, the letter frames the resume.

Is a weak cover letter worse than no cover letter?

No public data shows that a weak letter does worse than none. In the only callback data I found, a generic letter beat no letter, and a tailored letter beat both.

This comes from a ResumeGo field experiment covering more than 7,000 applications, which I've only seen reported secondhand. It was also run before the AI-era studies, so the gap between tailored and generic letters may have narrowed since tailoring became cheap. The multiples are my own calculation from the reported rates.

Two practical consequences:

• Required means send. Resume Genius reports that 60% of companies require cover letters, rising to 72% of medium-sized and 69% of large businesses. If the posting requires one, leaving it off is no longer a choice about quality.
• Your real risk isn't generic wording. It's claims nobody can back up. A letter with an inflated figure, or one that contradicts the resume, gives a careful reader a reason to distrust both documents. That's the one case where I'd leave an optional letter off: when criterion 5 or criterion 7 below can't be made to pass before the deadline.

The seven pass/fail criteria

A letter is ready to send only if it passes all seven criteria below. Each one has a pass line written tightly enough that two graders should agree, plus a note on what passing proves and how a letter can pass while still being weak.

1. Opening proof point

A proof point is one specific result, carrying a number, a named system, or a named employer, offered as evidence that you can do something the employer needs.

• Pass line: Sentence 1 or sentence 2 contains one proof point, and that proof point answers a requirement written in the posting.
• Proves: You read the posting and have done the hard part of this job before.
• Lies when: The number is real but has nothing to do with the posting, such as "grew revenue 40%" in a letter for a compliance role. That fails.

2. Posting anchoring

Anchoring means tying the letter to requirements the employer actually wrote.

• Pass line: At least two requirements are quoted or paraphrased from the posting, and each is followed by a sentence of evidence.
• Proves: The match is supported by evidence, not just claimed.
• Lies when: The posting's keywords appear with nothing behind them. That's exactly what an AI drafter produces automatically, and it's the alignment whose signal value fell 51%. A requirement with no proof sentence doesn't count toward the two.

3. Name-swap test

The name-swap test checks whether the letter would still be true if you sent it to a different company.

• Pass line: Delete the company name and any mission or values line, then insert a competitor's name. At least one remaining sentence becomes false or stops making sense.
• Proves: The letter was written for this employer's actual problem.
• Lies when: Only the company name and the "I admire your mission" line change. That's why you delete those lines before swapping.

4. No restating the resume

• Pass line: Every paragraph adds a why or a how that the resume doesn't contain: how you got a result, why you made a call, or why this role follows from your last one.
• Proves: The letter tells the reader something they can't get from the resume.
• Lies when: A grader marks necessary context as repetition. Naming the employer and the result before explaining how you got it isn't restating. A paragraph fails only if it adds nothing beyond the resume line it points to.

5. Checkable specifics

• Pass line: Every claim carries a number, a named system, or a named employer. Every figure also appears on the resume, a public profile, or a portfolio.
• Proves: The facts are yours. A model doesn't know them, and a reader can confirm them.
• Lies when: The precision is invented. In a Capterra survey of about 3,000 job seekers in 12 countries, 58% used AI in their applications, and 83% of those admitted using it to exaggerate or lie. A figure you can't trace to your own records fails, however specific it looks. Adjectives like "adept", "tech-savvy" and "cutting-edge", words an Intuit university recruiter singled out, are claims with nothing to check.

6. Length and format

• Pass line: One page or less. Any length, format, or addressee instruction in the posting is followed. The file type is accepted, and the plain text is pasted wherever the form offers a text field.
• Proves: The letter will reach a human in readable form and respects their time.
• Lies when: The letter is formatted perfectly but stuck in an attachment the reviewer can't see. More on that in the next section.

7. Match the resume

• Pass line: Every title, date, and figure in the letter is identical to the resume sent with it.
• Proves: The two documents describe the same person.
• Lies when: The letter rounds up a figure or uses an older title from a previous version of your resume. That kind of drift comes from editing the two documents separately.

Drift between letter and resume usually happens because they live in separate files and get tailored at different times. Refolk writes the resume and the cover letter from the same record of your history, so a figure that changes in one changes in the other, and criterion 7 starts from a match instead of a hunt.

How long should a cover letter be, and in what format?

A cover letter should be one page or less, in a file type the employer's system accepts, with the plain text also pasted into any text field the form provides. Length preferences are well measured. How each applicant tracking system displays letters mostly isn't.

The data is from ResumeLab, 205 HR professionals. A technical recruiter quoted by The Muse puts the target as a letter that "can be comfortably read in less than a minute." Two word-count rules get repeated widely, "70% of recruiters prefer under 300 words" and an analysis claiming interview-winning letters averaged 292 words, but I couldn't find their primary sources. Where readers actually stop reading isn't established publicly. Use the page rule, and let any limit in the posting override it.

On format, an applicant tracking system (ATS) is the software employers use to collect and route applications. Only one system was documented for this guide:

• Greenhouse accepts .doc, .docx, .pdf, .rtf and .txt uploads up to 100 MB. Its job-board API takes both file uploads and plain-text fields.
• Visibility depends on permissions. An uploaded cover letter appears on the candidate profile, but whether a particular reviewer can see it depends on their access settings.
• Keyword indexing isn't documented. Greenhouse's full text search shows snippets of where search terms appear in a profile, but the documentation doesn't say whether cover-letter text gets indexed. The common claim that an ATS "does not keyword-match cover letters" is unverified.
• Other systems weren't checked. I didn't verify Workday, Lever, iCIMS or Ashby. Check the application form itself: if there's a text box, use it.

Grading procedure: how to grade a cover letter step by step

Grade the letter in this order, against the posting and the resume you're sending with it, and write down a pass or fail for each criterion as you go. The order puts the cheapest, most decisive checks first. A letter that fails the opening check usually needs rebuilding, not polishing.

Some guides put an AI-phrase sweep first. I don't. Cui, Dias and Ye link success to editing effort, not to deleting particular phrases, and a letter scrubbed of stock phrases can still fail every criterion above. The grading sheet below is the record that lets a second grader check your verdict.

POSTING: [title] at [company], [link]
LETTER STATUS IN POSTING: required / optional / not mentioned
POSTING INSTRUCTIONS (length, format, addressee):
GRADER:                DATE GRADED:

C1 OPENING PROOF POINT ............ PASS / FAIL
   Sentence #: __  Result: ______  Posting requirement it answers: ______
C2 POSTING ANCHORING .............. PASS / FAIL
   Requirement 1 (as written): ______  -> proof sentence: ______
   Requirement 2 (as written): ______  -> proof sentence: ______
C3 NAME-SWAP (company name and mission lines deleted) ... PASS / FAIL
   Competitor used: ______  Sentence that breaks: ______
C4 NO RESTATING THE RESUME ........ PASS / FAIL
   Paragraph adding nothing beyond the resume (if any): ______
C5 CHECKABLE SPECIFICS ............ PASS / FAIL
   Claim -> where verifiable (resume line / profile / portfolio):
   ______ -> ______
   ______ -> ______
C6 LENGTH AND FORMAT .............. PASS / FAIL
   Pages: __  File type: __  Text field pasted: Y / N / no field
C7 MATCHES THE RESUME ............. PASS / FAIL
   Title, date, or figure mismatches: ______

VERDICT: SEND (7/7) / FIX AND REGRADE / LEAVE OFF (optional only, C5 or C7 unfixable before deadline)
FIXES (one per failed criterion):
1.
2.

When criterion 1 fails, the fix is almost always the same: lead with the posting's requirement in its own words, then your evidence. This skeleton gets you a passing opening, and you replace every bracket with a fact from your own records.

[Requirement from the posting, in its own words] is the part of this role I have done most directly: at [employer], I [what you did] using [named system or process], which [result with a number]. [One sentence on how you did it that your resume does not say.]

If you're grading several letters a week, the slow part is steps 2 and 3: pulling requirements out of each posting and matching them to proof from your history. Refolk scores how well you actually fit each posting and drafts the letter against those requirements, so you're grading a draft that already pairs requirements with evidence instead of building that pairing from a blank page.

How cover letter grading goes wrong: false passes and false fails

Most grading errors are false passes: a letter clears a criterion on the surface and still carries the weakness that criterion exists to catch. A few are false fails, where a grader penalizes something that's actually doing its job. Each criterion has a known way to fool it, and a specific re-check.

False passes

• Opening proof point: a real number with no link to the posting. "Grew revenue 40%" leading a letter for a compliance role passes a careless grader because it's specific. Re-check: point to the line in the posting that this number answers. If you can't, it fails.
• Posting anchoring: keywords without evidence. The posting's terms show up, but no proof follows them. That's the default output of an AI drafter, and it's why alignment went up while its value as a signal fell 51%. Re-check: draw a line from each requirement to its proof sentence. Unpaired requirements don't count.
• Name-swap test: only the name and mission line change. The letter looks tailored because it names the company twice and praises its mission. Re-check: delete those lines first, then do the swap. If nothing else breaks, it fails.
• Checkable specifics: invented precision. "Cut onboarding time 37%" sounds verifiable, but nothing in your records supports it. With 83% of AI-using job seekers in the Capterra survey admitting they used it to exaggerate or lie, readers have reason to doubt precise numbers. Re-check: the figure has to appear on the resume, a public profile, or a portfolio.
• Format: a perfect file nobody sees. A clean PDF is uploaded, but the reviewer's permissions hide attachments, or the form shows only its text field. Re-check: paste the plain text into the text field whenever one exists.

False fails

• No restating the resume: context marked as repetition. A grader fails a paragraph because it names an employer and result already on the resume, even though the paragraph goes on to explain how the result happened. Re-check: does the paragraph add a why or a how? If yes, it passes.
• AI-phrase sweep: detector output treated as a verdict. Cui, Dias and Ye describe current AI detection tools as "error-prone and unreliable". Failing a letter because a detector flagged it, or passing one because it didn't, is grading noise.

How much AI involvement costs you isn't settled either. In TopResume's survey of 600 hiring managers, 19.6% said they would reject an applicant who used generative AI to make a resume or cover letter, and 33.5% said they can spot AI in 20 seconds or less. CV Genius found 80% of hiring managers viewed AI-generated application content negatively, with 57% less likely to hire. Yet 52% of hiring managers in the TopResume data said AI is acceptable for proofreading or drafting support. A review of these figures notes that reported rejection rates run from about 20% to about 49%, mostly from secondary or vendor-run sources. Vendor blogs list openers such as "I am writing to express my interest in" as giveaways, but that's vendor opinion, not survey data.

A letter scrubbed of every stock phrase can still fail all seven criteria. Facts are what pass it.

Before you attach it: the checklist, and keeping this standard current

Run the checklist below right before you upload. Every item is a statement you can confirm as true or false against the letter, the posting, and the resume.

To keep the standard current, watch the mechanisms rather than the numbers:

• The signal research. The 51% drop comes from one marketplace and one tool. If new studies extend these findings to other kinds of hiring, or contradict them, adjust how much weight criteria 2 and 5 carry.
• ATS behavior. Before relying on an attachment being seen, check the application form and the system's own help pages for accepted file types and whether a text field exists. Only Greenhouse was documented here.
• Length preferences. The page rule rests on one survey of 205 HR professionals. If a posting states a limit, the posting wins.
• Callback evidence. The case for sending a generic letter over none rests on one experiment reported secondhand and run before AI drafting. If newer data shows generic letters falling below no letter, revise the leave-off rule in the verdict line.

Keep the grading sheets. After a few dozen applications, they'll show which criterion you fail most often, and that tells you where to start on the next draft.