Resume Keywords From a Job Description: Find and Place Them

Resume keywords are the words a posting uses for its tools, its subject matter, and its core responsibilities. Pull about ten of them, then place each by type: tools in your skills section and inside one bullet that shows use, domain nouns in your summary and bullets, and the posting's own phrasing on the single bullet that describes that responsibility.

Screening software and recruiters both search a pile of applications using the posting's own vocabulary, so work described in your private shorthand can go unnoticed even when it is an exact match. Mirroring their language is not a trick. It is translation.

The three kinds worth extracting

Not every noun in a posting is a keyword, and treating them all the same is what produces resumes that read like a search query. There are three types, and each belongs somewhere different on the page.

1. Hard skills and tools

Named software, languages, platforms, certifications, frameworks, methodologies. Salesforce, Python, Braze, GAAP, Kubernetes, CPA, Figma, HACCP. These are the most searchable terms in the document because they are unambiguous. A recruiter filtering a pile types a tool name, not a paragraph.

Mirror these exactly, including spelling and version. "Google Analytics 4" is not "GA". "Microsoft Excel" and "Excel" are safe as each other, but "advanced spreadsheets" is neither. When the posting uses an acronym you would normally spell out, give both forms once: "search engine optimisation (SEO)".

2. Domain nouns

The subject matter the role is about. Retention, churn, accounts receivable, clinical trials, cold chain, underwriting, incident response, shrinkage, onboarding. These signal that you have worked in this problem space, not merely with these tools, and they separate two candidates with identical tool lists.

Mirror them closely but not slavishly. If the posting says "win-back" and you have always said "reactivation", use theirs once and keep yours where it reads better. The goal is that a reader recognises their own world in your document.

3. The posting's exact phrasing for a responsibility

This is the type most people skip, and it carries the most weight with a human. Postings contain whole phrases describing what the job actually is: "own the end-to-end messaging calendar", "partner with the data team", "manage a queue of high-volume claims". Lift one or two verbatim onto the bullet that describes the same work in your history.

Used once, in the right place, that produces the small jolt of recognition that makes a reader slow down. Used four times, it reads as copy-paste and does the opposite. One per role is plenty.

Where each kind goes on the page

That second appearance matters as much as the first. Recruiters read the skills list to filter and the bullets to believe you.

A real posting, and the terms pulled from it

Here is an excerpt from a posting, of the length you would actually be working with.

Read it twice and sort what you find into the three buckets. Anything appearing in both sections goes to the top of the list. Repetition is the posting telling you what it is worried about.

That is eleven usable terms from four sentences, and you have not written anything yet. Notice what was left out: "wider marketing team", "hands-on", "three or more years". Those are framing, not vocabulary. If a term would not plausibly be typed into a search box or read as a skill, it is not a keyword.

The same experience, before and after

Below is one person's real history, written the way most people write it the first time — accurately, in their own vocabulary, and invisibly. The numbers are theirs; use your own.

Nothing here is false. But not one of the eleven extracted terms appears, so a search for "Braze" or "retention" or "cohort" misses this candidate entirely, and a human skimming learns only that they sent some emails.

The work is identical. What changed is the vocabulary carrying it and which facts got promoted. The responsibility phrase appears once, on the first bullet. Braze and SQL sit inside sentences showing what was done with them, not only in a list. And the domain nouns now match the words a recruiter for this role would search.

A keyword in a list is a claim. The same keyword inside a sentence about something you delivered is evidence. Aim for evidence.

The skills line underneath does the filtering work: Braze, Iterable, SQL, Looker, Segment, Figma, Google Analytics 4. Plain, comma-separated, no rating bars or graphics — those carry no extractable text at all, which our guide on the ATS-friendly resume format covers in detail. For the wider process of reordering and cutting around these terms, see tailoring a resume to a job description.

How many keywords, and how often

Eight to twelve terms per application is a sensible working range, weighted toward whatever the posting repeats. Each needs to appear at least once. Almost none need to appear three times.

• Once is the default. Repetition does not make a term count for more in any system worth worrying about, and it makes the page worse to read.
• Twice is fine for the top two or three terms: once in the skills list or summary, once inside a bullet. That is coverage, not repetition.
• Three times means you are writing for software. Read the page aloud. If a phrase makes you wince on the third pass, a recruiter noticed it on the second.

Match the verb to the term, too. "Familiar with Braze", "used Braze", and "owned our Braze instance" describe three different levels of responsibility, and picking the accurate one is the whole game. Our guide to resume action verbs covers choosing without overclaiming.

Why stuffing and hidden text both fail

Two shortcuts circulate constantly. Both are worse than doing nothing.

Keyword stuffing is pasting a block of terms with no sentence around them, or forcing the same phrase into six bullets. It fails on both ends. A human opens the file and sees padding, which reads as someone with little to say. A recruiter searching the database gets a hit, opens the document, finds no context for the term, and moves on. You spent the hit and delivered nothing.

Hidden text — white-on-white keywords, a two-point font block at the foot of the page, terms tucked behind an image — is not hidden from software at all. That is the part people get backwards. A parser extracts every character regardless of colour or size. Select-all and copy shows it in seconds, and application systems routinely display the parsed text back to a recruiter. It is visible to exactly the audience it was meant to fool.

A keyword you cannot back up is a liability

The simplest test for any term before it goes on the page: could you talk about it for sixty seconds without inventing anything? Not present on it, just answer a follow-up.

If you list Braze and the screening call opens with "what did you build in Braze", a vague answer is worse than never having listed it, because you have now shown that your resume overstates things. The interviewer applies that discount to every other line. One padded skill quietly devalues the twelve true ones beside it.

When your exposure is partial, say so precisely instead of dropping the term. Precise partial claims are credible and easy to defend:

• "Built segments in Braze; set up the campaigns, not the integration"
• "SQL for reporting and ad-hoc analysis, not for modelling"
• "Ran A/B tests in-platform; statistical design came from the data team"
• "Six months on Iterable before we migrated; two years on Braze since"

Every one of those survives a follow-up question, and each reads as someone who knows the boundary of their own work. That is a good thing to signal in the first five minutes of an interview.

Common questions