How the CV keyword checker works
The checker reads the job description first. It looks for known multi-word skills (“stakeholder management”, “Power BI”, “health and safety”), single skills and tools (“SQL”, “Kubernetes”, “Xero”), acronyms and certifications (“PRINCE2”, “CIPD”, “ACCA”), product names written with capitals mid-sentence and phrases the advert repeats. Generic advert words such as excellent, team, role and experience are ignored, and words in sentences containing “must”, “essential” or “required” get more weight. The result is an editable list of up to 25 terms, because only you know which ones are real requirements.
Each term is then looked up in the CV. Both texts are split into words with the same rules: letters and numbers only, so a full stop, comma or bullet never changes a word; light stemming, so “reporting”, “reports” and “reported” meet; and a few joined spellings, so “e-commerce” finds “ecommerce” and “Microsoft Excel” finds “Excel”. Every match comes with the sentence it was found in, so you can see whether the CV says “built Power BI dashboards” or only “exposure to Power BI”.
Negated mentions are not matches
A word appearing in a CV is not the same as a candidate having that skill. “No Python experience”, “never used Azure”, “unfamiliar with Snowflake” and “Python: none” all contain the keyword, and a checker that counts them as matches is wrong in the most misleading direction. This tool looks for negation cues (no, not, never, without, none, lack of, lacking, unfamiliar with, cannot and contractions such as haven’t) in the same clause before a term, within a few words, and for short negative phrases directly after it. The scope stops at the end of the sentence, at a semicolon or bracket, at words like but or however, and after a completed negative phrase followed by a comma or “and”, so “No Python experience and strong SQL” still counts SQL. Phrases that only look negative, such as “not only” or “no longer”, are ignored.
When a term is mentioned both positively and negatively, it is listed as mentioned, with a note. That usually means context worth reading, for example “no commercial Python experience; Python for personal projects”.
Why there is no match percentage
Percentages invite comparisons they cannot support. The same CV can score 67% or 100% depending on which terms were extracted, how phrases were split and whether a single full stop sat next to a word. We show three plain lists instead: mentioned, mentioned negatively and not found, each with its evidence. That is enough to decide what to ask a candidate and what to highlight in a pitch, without pretending to measure suitability.
How recruiters use it
- Before a screening call: the “not found” list becomes your question list. A missing term often means the candidate used different words, not that they lack the skill.
- Before a client submission: check that the claims in your pitch appear in the CV, and that nothing you are about to highlight is actually a negative mention.
- When rewriting an advert: run a few strong CVs against your job description to see whether the advert asks for things in words real candidates use.
Limits worth knowing
The checker does not understand meaning. It cannot tell three years of Python from a weekend course, cannot judge recency, does not know that “Postgres” is “PostgreSQL” and cannot read a skills section rendered as an image. It also cannot assess years of experience, seniority or soft skills reliably. Treat it as a fast way to find evidence in a long CV, then read the CV.
From keywords to a submission your client trusts
Once you have the right candidate, CVPitch formats the CV in your agency’s branded template, shows the source passage beside every fact and flags anything it could not support, so the shortlist you send says only what the candidate’s CV says. See plans and pricing or start a free trial.