You’ve read two completely contradictory pieces of advice back to back. One says: mirror the exact keywords from the job posting or the software will never surface your application. The other says: sound human, be specific, don’t write like a robot filling in blanks. Both are right. Nobody tells you they’re describing two different readers, and your one letter has to satisfy both of them in the same 400 words.
That’s the actual problem — not “how do I beat the ATS” and not “how do I sound human,” treated as separate questions. It’s how you write one document that clears a keyword filter first and then convinces an actual person second, without the version that wins round one losing round two.
Why “optimize for ATS” and “sound human” are usually treated as opposites
Most advice picks a side. The ATS-optimization camp tells you to mirror exact phrasing from the job description, front-load keywords, and keep formatting simple enough for parsing software to read correctly. The human-appeal camp tells you to write like you talk, lead with a specific story, and avoid anything that sounds templated or algorithmic. Followed to their extremes, each piece of advice actively undermines the other. A letter stuffed with exact-match keywords from the posting reads stilted and robotic to a human. A letter written purely for voice and specificity might use “led cross-functional initiatives” where the posting says “managed stakeholder alignment” — different words, same meaning to a person, invisible to a keyword filter.
The reason this gets treated as a binary is that most guides only think about one stage of the process. The truth is your letter passes through two very different readers in sequence, and they’re evaluating for completely different things. A LinkedIn Talent Solutions survey found that 94% of hiring managers say a personalized cover letter influences their hiring decision when the role requires strong communication skills — meaning by the time a human is reading your letter, the personalization is what’s actually deciding the outcome, not the keyword match that got you there. The ATS’s job is to reduce a pile of hundreds down to a shortlist. The human’s job is to pick from that shortlist. Optimizing only for the first reader gets you filtered in and then quietly passed over. Optimizing only for the second gets you filtered out before a human ever sees your writing at all.
There’s a second layer to this that’s changed fast: AI-written cover letters are now common enough that hiring managers are actively screening for them, which raises the cost of writing something that reads like it was optimized by a formula. Recent industry data found that a large share of hiring managers have already encountered AI-generated content in applications, with a substantial portion specifically flagging AI-written cover letters as a negative signal rather than a neutral one. That matters here because heavy keyword-stuffing and generic AI-generated prose tend to produce the exact same telltale flatness — sentences that technically say something relevant without actually sounding like a specific person wrote them. The two problems (bad ATS strategy, bad human impression) often have the same root cause and the same fix.
The fix isn’t splitting the difference by writing something half-optimized and half-human — that produces a letter that’s mediocre at both jobs. It’s sequencing the two goals correctly: match the software’s requirements structurally and mechanically, and reserve your actual voice for the content sitting inside that structure.
A quick example of what this looks like when it goes wrong: I once reviewed a cover letter for a friend applying to a data analyst role that had clearly been optimized purely for keywords — it name-checked “SQL,” “data visualization,” and “cross-functional” in nearly every paragraph, but read like a checklist rather than a person talking. It almost certainly cleared the parser. It also almost certainly bored the hiring manager within the first two lines. We rewrote it around one specific project — a dashboard she’d built that cut a weekly reporting task from four hours to twenty minutes — and worked the same required keywords into that story instead of stacking them on their own. Same keywords, same qualifications, completely different letter. She got the interview.
The practical tool: build the letter in two passes
Pass one — structure for the parser. This is entirely mechanical and doesn’t touch your voice at all:
- Use a standard, single-column layout — no text boxes, tables, or graphics the parser might skip.
- Pull 3-5 exact phrases directly from the job posting (not synonyms) and use them naturally somewhere in the letter — “cross-functional collaboration” if that’s their phrase, not your reworded “worked across teams.”
- Save and submit as a .docx or PDF built from a text-based document, not a scanned image or design file.
- Keep your contact info and the role title in plain text near the top, not embedded in a header or footer some parsers skip.
Pass two — write for the human, inside that structure. Once the mechanical requirements are met, everything else is written exactly the way you’d write for a person, because it is:
I led the redesign of our onboarding flow after noticing a 30% drop-off in week one — the kind of cross-functional collaboration [their exact phrase] this role clearly calls for, working closely with design, support, and data to cut that number in half within a quarter.
Notice the keyword (“cross-functional collaboration”) sits inside a real, specific sentence with a number and an outcome — it’s not a standalone phrase bolted onto the letter to satisfy a filter. That’s the whole method: the keyword clears the machine, the specificity and voice around it convinces the person reading it next.
Understanding your audience is important, but the structure of the letter matters just as much. Our guide to writing a cover letter that gets you noticed walks through the key elements that help make your message clear and relevant.
Common mistakes that fail one reader while trying to please the other
- Keyword-stuffing at the expense of sentences that make sense. Cramming in every phrase from the posting produces a letter that might parse cleanly and still get discarded the moment a human starts reading it.
- Writing exclusively for voice and ignoring structure. A beautifully written letter in a two-column design template or embedded in an image can fail to parse at all — the human reader never gets the chance to be won over.
- Using synonyms instead of the posting’s exact terms. “Team leadership” and “people management” mean the same thing to a person; an ATS keyword match often doesn’t see it that way. Match their language first, then add your own around it.
- Letting an AI tool generate the whole letter unedited. It tends to produce exactly the kind of generic, keyword-adjacent flatness that fails the human test even when it technically passes the parser — the fix is using AI for a draft, then rewriting the specifics in your own voice.
- Forgetting the human reader is evaluating something different than the software. The ATS checks for relevance; the human checks for fit and genuine interest. A letter that only proves relevance without showing personality clears the first gate and stalls at the second.
FAQ
Do all companies actually use an ATS to screen cover letters? Most mid-size and large companies use some form of applicant tracking system, though how heavily it weighs the cover letter specifically (versus just the resume) varies a lot by employer. When in doubt, treat every submission as if both documents get parsed — the two-pass method costs you nothing extra either way.
Will using AI to help write my cover letter get flagged by an ATS? The ATS itself typically isn’t detecting AI-generated text — that’s a human judgment, not a parsing function. The real risk isn’t the software catching you; it’s a hiring manager recognizing generic phrasing once they’re reading it, which is why editing an AI draft into something specific and personal matters more than avoiding AI tools altogether.
Should I use the exact words from the job posting even if they sound awkward? Yes, in moderation — pick the 3-5 terms that matter most (usually the hard skills or specific methodologies mentioned) and work them in naturally rather than trying to match every phrase. Forcing in every keyword usually produces the awkward, robotic sentences you’re trying to avoid.
How do I know if my cover letter formatting is ATS-friendly? Stick to a single-column layout, standard fonts, and no text boxes, tables, columns, or graphics containing important information — if you can copy and paste your entire letter into a plain text document and it still reads correctly in order, it will almost certainly parse correctly too.
Is it worth tailoring my cover letter for every single application? For roles you’re genuinely targeting, yes — the two-pass method above only takes a few extra minutes once you have a strong base letter, since you’re mainly swapping the keyword set and the specific example, not rewriting from scratch each time.
Does the order of information matter for ATS parsing, or just for the human reader? Both, though for different reasons. Parsers generally read top to bottom and can misattribute information if it’s buried in an unusual layout, while human readers make snap judgments in the first few lines — so leading with your strongest, most specific point serves both readers at once rather than being a tradeoff between them.
Write for the second reader, format for the first
The ATS is a gate, not a judge — it decides whether your letter gets read, not whether you get hired. Treating both requirements as one sequential process, instead of one contradictory tug-of-war, is what actually lets you write a letter that clears the software and still sounds like an actual person wrote it, because it does.
Once your letter is doing both jobs well, it’s worth checking that it isn’t quietly duplicating what your resume already covers — here’s how the two documents are supposed to divide the work, so the specificity you’re building into the letter isn’t wasted repeating what the ATS already pulled from your resume.
