When Your CV Is a Real CV, Not a Resume in Disguise
Almost all advice about making a CV machine-readable is written for a two-page resume that happens to be called a CV. If that is your document, the advice transfers and you can stop worrying. If you have an actual long-form CV — a document whose purpose is completeness, running to many pages with publications, grants, teaching, or a full chronology — the guidance points the wrong way, because completeness and clean text extraction pull against each other. The resolution is usually two documents rather than one compromise.
First, work out which document you have
The word covers three different things, and the answer changes what you should do.
A resume called a CV. Common wherever British or European usage prevails: two pages, reverse-chronological, tailored per application. Everything written for resumes applies without translation, including the structural basics in what actually makes a document readable by software.
A long-form academic or research CV. Its job is to be exhaustive. Publications, conference papers, grants held, students supervised, review work, teaching load, funding amounts. Nobody reads it front to back; people search it for the sections they care about, and length is not a defect.
A CV following a national convention you are applying into. Some markets expect a photograph, a date of birth, nationality, or marital status as a matter of routine; others treat their presence as odd or actively unwelcome. Which is which is a fact about the market you are applying into, and you should follow the local convention rather than a generic template — the point here is only that a drafting tool will silently default to some convention without telling you which.
Only the second and third cases need anything beyond the standard advice.
Where a long CV and text extraction genuinely conflict
The features that make a scholarly CV usable by a human are the ones most likely to survive badly when the file is converted to plain text.
Tables. Grants, teaching and course lists are naturally tabular, and tables extract unpredictably. A three-column table of years, courses and enrolments can come out as a run of numbers with no relationship to the words. If the content is genuinely tabular, expect the extracted version to be unusable and do not rely on it carrying meaning.
Two-column layouts. A layout with a narrow left rail reads correctly to your eye and can extract in either column order, interleaving lines from both. This is the failure you are least likely to notice, because the document looks fine.
Numbered publication lists with hanging indents and reference formatting. These usually extract as text, which is fine, but they can dominate the extracted document — the searchable version of your CV becomes mostly other people’s names.
Headers, footers and page furniture. Running heads repeating your name on twelve pages, page numbers, section markers. All of it lands in the extracted text.
Diacritics and non-Latin characters. Names, institutions and publication titles carry them legitimately. Encoding problems here are real and they affect the one field nobody should get wrong — your own name. Export the file and look at your name in the exported version, every time.
The two-document answer
For a long-form CV, stop trying to make one file serve both purposes and produce two.
The complete CV, formatted for a human reader, as long as it needs to be, with whatever tables and layout make it legible. Send this where a full CV is asked for and where a person will open it.
A short application document, two pages, one column, ordinary section headings, no tables, plain text throughout, tailored to the specific role. Send this into application portals and anywhere a resume is what is really wanted, including cases where the form calls it a CV.
They contain the same facts and are not in tension, because the second is a selection from the first rather than a different account. Keeping them consistent matters: dates, titles and institution names should be identical across both, since those are the fields most likely to be compared against somebody else’s records — the reason they need to match is set out in resume dates that have to match.
If you only ever produce one document, produce the short one for applications and keep the complete one as your own record.
The check that tells you the truth about your file
You cannot know how any particular system will read your document, and nobody can tell you, because that is configured per employer. What you can do is see the worst plausible version of your own file.
Export it, open the export, and select the whole thing into a plain text editor. Read what appears. You are looking for four things: whether your name and contact details are intact, whether the reading order still makes sense, whether the sections are still identifiable, and whether anything has turned into gibberish. Fix whatever is broken in the source and export again.
That five-minute check replaces every confident claim about what software prefers. It is the same reasoning behind why no template can make a document machine-friendly on its own: the property belongs to the file you actually send, and you can inspect it.
Two things specific to academic applications
Do not compress a long CV by removing the specifics. The temptation, when a portal wants two pages, is to shorten each entry. The better move is to remove whole sections that are irrelevant to this application and keep the surviving entries fully specific. Vague entries help nobody and a truncated publication list is more useful than a summarised one.
Beware of the summary a tool writes for the top of an academic CV. Asked to summarise a long research record, a drafting tool will produce a paragraph that promotes you — implying independence you may not yet have, leadership of work you contributed to, or a field position you have not claimed. Long CVs are unusually vulnerable to this because there is so much material to compress and the compression happens in your absence; the way generated summaries quietly promote you is worth reading before you accept one.
The general point holds for both documents. A tool can tidy prose and it can tell you what is missing. What belongs in a complete academic record, and what a specific committee in a specific country expects to see, is knowledge held by people in your field — and worth asking one of them, since they will know the local convention that no template encodes.