Most keyword advice reduces to "use words from the job description", which is true and not very useful. The harder questions are which words carry weight, where they should appear, and which popular terms are pure filler. This guide answers those.
How keyword matching actually works
Applicant tracking systems do not simply count word occurrences. Most weight keywords by where they appear and how recently: a skill in your current role's bullets counts for more than the same skill listed in a role from eight years ago, and a term appearing in both your skills section and your experience reads as substantiated rather than claimed.
Two consequences follow. First, a skills section alone is weak evidence — the strongest position is a term that appears in your skills list and in a bullet showing you used it. Second, repetition beyond that adds nothing. Naming the same tool six times does not multiply anything; it just costs you space.
The four keyword types that carry weight
1. Hard skills and tools
Named technologies, platforms, and systems — Python, Salesforce, Kubernetes, SAP, Figma, Snowflake. These are unambiguous and easily matched, which makes them the highest-signal terms available. If a posting names a specific tool and you have genuinely used it, it should appear.
2. Methodologies and frameworks
Domain-specific approaches: A/B testing, dimensional modelling, contract testing, SOX compliance, design systems, incident postmortems. These distinguish practitioners from people who list tools they have only touched, because they imply how you work rather than what you have installed.
3. Certifications and qualifications
AWS Solutions Architect, CISSP, PMP, CPA, CFA. These are frequently used as hard filters, and their absence can eliminate an application regardless of everything else. Write both the acronym and the full name, since postings vary in which they use.
4. Role and domain vocabulary
The language of the specific function: "cohort retention" for analysts, "error budget" for SRE, "underwriting" for insurance, "revenue recognition" for accounting. This is the vocabulary that signals you have actually worked in the field, and it is the category most candidates under-use.
The keywords that waste space
These appear on a large share of CVs and earn close to nothing:
- "Hard-working", "detail-oriented", "team player", "results-driven". Unverifiable, universally claimed, and never used as screening criteria. Every applicant asserts them, so they carry no discriminating information.
- "Microsoft Office", "email", "internet". Assumed for any professional role. Listing them signals a thin skill set rather than a broad one.
- "Agile", "Scrum", "SDLC" on their own. Not useless, but so common in technology CVs that they rarely differentiate. If you list them, attach evidence — running a specific process change is worth more than the label.
- "Responsible for". Not a keyword at all, but it consumes the highest-value real estate on the page: the opening words of a bullet, which is where scanning attention lands.
- Buzzwords without substance — "synergy", "leverage", "thought leader", "rockstar". These actively damage credibility with human reviewers while contributing nothing to matching.
Where to place keywords
Placement matters more than frequency. In rough order of value:
- Professional summary. The first thing read by both parser and human. Three or four sentences naming your specialisation, level, and strongest result will naturally contain your most important terms.
- Current or most recent role's bullets. Weighted most heavily by recency, and the most convincing place for a skill to appear because it comes with context.
- Skills section. Clean, scannable, grouped by category. This is where matching does most of its literal work.
- Job titles. If your internal title was "Growth Ninja", write the functional equivalent — "Growth Marketing Manager (internal title: Growth Ninja)" — so the role is recognisable.
The synonym problem
The most common invisible failure is having the right experience under the wrong label. Some frequent mismatches:
- "Client relationship management" versus "stakeholder management"
- "Amazon Web Services" versus "AWS"
- "Machine learning" versus "predictive modelling"
- "People management" versus "line management" versus "team leadership"
- "Business intelligence" versus "analytics" versus "reporting"
Where you genuinely have the experience, adopt the posting's phrasing. Where a term is standard in your industry but written differently in the posting, include both once — "line management (people management)" reads naturally and satisfies either matcher.
How many keywords is enough?
There is no fixed number, and pursuing one leads to keyword stuffing, which modern systems detect and human reviewers find obvious. A more useful test: take the ten most repeated substantive terms in the job description and check that the ones you genuinely have appear somewhere natural in your CV. That is generally sufficient.
Stuffing fails in a specific way worth understanding. A block of thirty comma-separated technologies with no supporting bullets reads to a hiring manager as someone who has heard of thirty things — which is a weaker position than clear depth in six.
Tailoring without rewriting everything
Tailoring per application is genuinely effective and does not require starting over. In practice it means adjusting three things: the summary's opening line to match the target role, the ordering of your skills so the relevant ones appear first, and two or three bullets reworded toward the posting's emphasis. That takes about ten minutes and captures most of the benefit.
Find your gaps automatically
Comparing your CV against a posting by hand is slow and easy to get wrong. Paste any job description into CVEdge's Job Match and it identifies which keywords are missing, which are already covered, and which would raise your match score most — so you can tailor from evidence rather than guesswork.