What a Cover Letter Is Still Read For When AI Writes Them All

Is a cover letter still worth the hour?

Assume the reviewer spends less time on your letter than they used to, and that most of the letters in the stack now read alike. A generic letter is worse than no letter there, because it costs the reviewer's attention and returns nothing they didn't already expect. A specific letter is different: it is worth the hour when you have a real, particular reason for wanting this job at this employer and nowhere else on the application to put it.

• Worth writing. You have a specific reason, a real connection to the role, or something you'd do differently that the resume can't carry.
• Not worth writing. You're filling the field because it's there, and the letter would say nothing a form couldn't say more efficiently.

What decision is happening on the employer's side right now?

Some employers are deciding whether to keep the cover letter field at all, and a few have already dropped it. The standing advice there is blunt: keep the cover letter only where prose is itself the deliverable, communications, brand, policy, technical writing, and drop the field everywhere else for one short, specific question.

That decision is being made without you in the room, which is exactly why it is worth knowing about before you spend an evening on a letter nobody asked for. The letters that survive this shift are the ones the employer's own guidance already predicts: a real answer to a real constraint, capped in length, read the same way for every applicant. A letter that reads as generic gives that kind of review nothing to hold on to, since the standing guidance treats an open-ended essay field as the thing to remove, not the thing to weight more heavily once it stays on the form.

Keep the letter only when you have a real reason

Write the real reason in the first two lines, before any scene-setting. Name the employer specifically, name the part of the role that pulled you toward it, and say what you already understand about the problem the role exists to solve. Everything after that line should be evidence for the claim you just made, not a second summary of your resume in prose.

• First two lines. The one true, specific reason. No throat-clearing before it.
• Everything after. One piece of evidence for that reason: something you built, read, or noticed about the employer that a generic applicant wouldn't have.
• The cut test. Delete any sentence that would survive a find-and-replace of the company's name.

This matters more now, not less. Once a letter's wording matches a job post closely enough for any tool to manage, matching stops being informative. In a working paper studying a large online labor platform, the correlation between how closely a cover letter matched the posting and getting a callback fell 51 percent after an AI writing tool went live, and the correlation with getting an offer fell 79 percent ¹. Employers shifted weight toward signals that are harder to produce on demand: on the same platform, the correlation between callbacks and a worker's record of past work rose 5 percent ¹.

What survives in the two lines a resume can't hold?

A resume is a list of facts in a fixed shape. A letter is the one place in a standard application built for connective reasoning: why this employer specifically, what you would do first if hired, what you already understand about the problem. None of that fits naturally into a bullet, and none of it is optional filler once you actually have it to say.

• Why this employer, specifically, not "your mission of innovation."
• What you'd do first, concretely, if you started Monday.
• What you already understand about the problem, demonstrated rather than claimed.

The underlying failure is the same one behind why an AI-written resume ends up sounding like everyone else's: a prompt with nothing distinctive in it returns the average of every letter the model has read. A letter with a real, particular reason in it can't average, because nothing else in the training data has your actual reason in it.

The evidence on using an AI tool to draft the letter itself is thin, and what there is fades. When one large freelance platform added a one-click AI drafting tool, using it went with a callback rate 51 percent higher than before the tool existed, off a low base ². The authors report that result as significant at the 10 percent level but not the 5 percent, and it tapered off after two months ². An edge everyone can buy stops being an edge.

Skip the letter when the employer already told you to

Some employers have already answered this question for you by removing the letter field and asking one or two short, specific questions instead. Replacing the cover letter with one scored question is the standing guidance on the other side of the desk, and it changes what you should actually do: answer the question that's actually asked, at the length asked for, instead of pasting in a letter built for a field that no longer exists.

• The field says "cover letter." Write one, following the two-line rule above.
• The field asks a specific question instead. Answer that question and stop. A pasted letter under a mismatched prompt reads as not having read the posting.

Whichever field you're filling, the same habit applies. On the platform where AI drafting was measured directly, more than three in four AI-generated letters were submitted within a minute of being generated, essentially unedited; within the same worker, more editing time went with a higher chance of winning the job, though the authors are explicit that people who edit more are not a random group ³. The draft is a starting point either way. Say the thing only you would say, then let the tool tighten it.

That leaves what to say about the draft, if anything. In an experiment with 85 US supervisors who had hiring experience, a letter stating that generative AI had prepared it was trusted less than one that said nothing about how it was written ⁴. A companion study, run on a freelance advisor rather than a job applicant, found that being exposed by someone else cost more trust than disclosing had ⁵. Both measured stated trust, not a hiring decision. The workable rule is to follow whatever the employer published about AI use on applications, and to answer specifically if you are asked.

See a sample report >