You can test most of what automated screening does to your CV in about twenty minutes, without any tool. Copy the text out to see what extracts, check your CV against the advert's essential criteria one by one, and have someone unfamiliar with your work read the top third. Those three tests catch the large majority of preventable rejections.
Test One: What Actually Extracts
Open your CV, select all, copy, and paste into a plain text editor. Not a word processor, which preserves formatting, but something that shows raw text. What appears is close to what a parser receives.
What to Look For
- Your name and contact details near the top, present and intact
- Job titles, employers and dates sitting together rather than separated
- Each bullet arriving as one coherent sentence
- Nothing missing entirely, particularly anything from a header or text box
- Nothing out of order, which is the signature of a column layout
If your phone number has disappeared, it was in a header. If your skills are interleaved with your job history, you have a two-column layout. Both are structural and neither can be fixed by rewriting sentences.
Test Two: The Criteria Grid
Open the advert and copy every essential requirement into a list. Then go through your CV and write, beside each one, where the evidence sits. Not whether you have the skill. Where a reader would find it.
Any requirement where you cannot point to a specific line is a likely rejection reason. This is the single most useful check available and almost nobody does it, because it produces uncomfortable answers.
What Counts as Evidence
- A named tool, standard or registration, written out in full
- An activity with a scale attached, such as a caseload or a budget
- A dated position that makes the recency clear
- An outcome, ideally with a number
A claim in a profile paragraph is not evidence. "Strong stakeholder management" at the top of page one does not satisfy a requirement for stakeholder management experience; a bullet describing which stakeholders and what you managed does.
Test Three: The Twenty-Second Read
Give your CV to somebody who does not know your work. Twenty seconds, then take it back and ask three questions: what job do I do, what level am I at, and name one thing I achieved.
If they cannot answer all three, your top third is not working. That is the space where screening decisions are made, and it is where most CVs put a paragraph of adjectives instead.
Test Four: The Date Check
Every role needs a month and a year, formatted the same way throughout. Inconsistent formats make an automated timeline unreliable, and missing months are widely read as concealing a gap even when they are not.
Read your history as a continuous line. Where there is a break of more than a couple of months, add a dated line explaining it: study, caring responsibilities, travel, illness, redundancy. A short explanation is neutral. An unexplained gap is a question, and readers with forty applications resolve questions against the candidate.
Test Five: Read It Aloud
Errors that survive a spellchecker rarely survive being read aloud. It also exposes the sentences you would never say to a colleague, which are usually the ones written to impress rather than to inform.
Four minutes, and it catches more than any grammar tool.
What the Tests Cannot Tell You
None of this predicts a score, and any tool claiming to give you a definitive percentage against a specific employer's system is guessing. Employers configure screening differently, weight criteria differently, and increasingly use semantic reading that behaves differently from keyword matching.
What the tests do reliably establish is whether your document is readable, whether it addresses the advert, and whether a human can place you in twenty seconds. Those three properties are necessary regardless of how any particular employer screens.
Where Tools Genuinely Help
Free checkers are useful for the mechanical layer: confirming text extracts, flagging missing contact details, spotting inconsistent dates. Treat that as a proofreading service rather than as a prediction.
Be sceptical of anything that reports a match percentage against a job description. The number is generated by that tool's own comparison, not by the employer's system, and optimising towards it tends to push people back towards keyword stuffing, which reads badly to humans and is recognised for what it is.
Building a Base Document You Only Test Once
Running every test for every application is unsustainable, and it is unnecessary. Separate the checks that belong to the document from the checks that belong to the application.
Document-Level, Tested Once
- Extraction, layout, contact details in the body
- Consistent date formatting throughout
- Fonts, sizes and single-column structure
- File naming convention and document properties
Application-Level, Every Time
- The criteria grid against that specific advert
- The top third, reordered so the relevant evidence sits first
- Vocabulary alignment where the advert uses a standard industry term
Once the base document passes the mechanical tests, it keeps passing until you change the layout. That reduces per-application work to roughly fifteen minutes of genuine tailoring, which is sustainable across a long search in a way that rebuilding the document each time is not.
Frequently Asked Questions
Can I See My Actual Screening Score?
You can ask. Where an organisation holds a score about you as personal data, a subject access request can require them to provide the personal data they hold, although how much detail you receive varies.
Do Free CV Scanners Reflect Real Systems?
Partly. They test extraction and formatting reasonably well. Their scoring is their own and does not represent any specific employer's configuration.
Should I Aim for a High Keyword Match?
Aim for genuine coverage of the essential criteria in natural language. Deliberately maximising keyword density produces text that reads badly to the human who sees it next.
How Often Should I Retest?
Run the extraction test whenever you change the document's layout, and run the criteria grid for every application, because the criteria change each time.
Is a Portal Preview Worth Reading?
Yes. Many portals show what they extracted before you submit. If your job titles or dates are wrong there, fix the document rather than only correcting the form.
The Twenty-Minute Routine
Extraction test once per document version. Criteria grid once per application. Twenty-second read once, with a real person. Those three, in that order, address the mechanical, the targeted and the human layers, and they cost less time than the applications they save you from wasting.
InsightCircle
