CV red flags worth checking (and ones to ignore)

A small set of CV red flags reliably deserves a recruiter’s attention: dates that overlap with no explanation, titles that outrun the duties listed beneath them, achievement claims with nothing verifiable behind them, and team work described entirely in vague pronouns. The flags screening folklore obsesses over (employment gaps, short stints, unusual formatting) predict very little on their own. The dividing line is simple: a real flag survives a clarifying question, while noise dissolves the moment context arrives. Treat every flag as a prompt to verify, never as a verdict.

Which CV red flags actually predict problems?

Four, in rough order of how often they pay out.

Unexplained date overlaps. Two full-time roles running in parallel, a degree completed in one country while a full-time job ran in another, a “present” end date at a company that no longer exists. Overlaps often have innocent explanations: contracting, a permitted side engagement, garden leave during a notice period. That is exactly what makes them useful. A candidate with a clean story tells it in one sentence; a candidate without one starts revising the timeline in front of you.

Title inflation relative to duties. A “Head of Engineering” whose bullets describe mid-level feature work, a “Director” with no budget, headcount or named initiative anywhere in sight. The flag is not the big title; titles are nearly worthless across company sizes, and a VP at a five-person startup writes code all day. The flag is the mismatch: led, owned and managed are doing a lot of work when they never attach to a concrete object.

Claims with no artifacts. Real work leaves residue: named products, systems, talks, repositories, certifications with registration numbers, metrics tied to a mechanism. “Improved conversion by 40 percent” can be entirely true. The same claim with no mention of what changed, over what period, or how it was measured is decoration. Ten years of senior experience with zero specifics anywhere is the strongest version of this flag.

Vague pronouns around team work. When every achievement is “we”, you cannot tell what the candidate personally contributed; the CV may describe a team they merely sat near. The inverse flags too: “I” attached to clearly team-scale work (“I migrated the company to the cloud”) suggests either inflation or a working style worth probing. This one rarely needs CV-stage action, because behavioral questions about pressure and conflict expose it within minutes: people who did the work can describe the Tuesday it nearly failed.

Which red flags are mostly noise?

The ones screening habits weight heaviest.

Employment gaps. Layoffs, caregiving, illness, burnout recovery, relocation, visa timing. None of these predict performance, and penalizing gaps quietly filters for people who were never sick, never cared for anyone and timed the market well, qualities with no connection to the job. A gap plus a one-line explanation is a closed question.

Short stints with context. One short stint means nothing: companies fold, reorganizations land badly, honest people leave bad fits early. Even a pattern of short stints earns one question, not a rejection, since contractors and early-startup people accumulate them legitimately. The actual flag is a candidate who, asked about the pattern, cannot or will not give context.

Unconventional formats. A two-column designed CV mostly tells you your parser struggled, not that the candidate is weak; how extraction works and where it breaks is covered in our explainer on CV parsing, and the DOCX versus PDF question has its own answer. Career changers and international applicants also use conventions that read as odd locally. Format is a tooling and culture signal, not a competence signal. Typos are similar: weigh them for an editor, shrug for a backend engineer.

Here is the whole post as one table:

What you see	Signal or noise	The move
Overlapping dates, no explanation	Signal	Ask for the timeline in their words
Title far above the duties described	Signal	Ask for scope: budget, headcount, decisions
Big claims, zero artifacts	Signal	Request one link, name or mechanism
All “we”, no “I” (or the reverse)	Signal	Behavioral questions in the interview
Employment gap with context	Noise	Accept the one-line explanation
A few short stints, explained	Noise	One question, then move on
Unusual format or template	Noise	Fix your parsing, not your shortlist

How do you check a flag without assuming the worst?

Convert it into one neutral question. “Walk me through 2022; I see the role at X and the contract at Y overlap” invites a story. “Explain this discrepancy” invites a defense, and you learn less from a defense. Specific, dated, unhesitating answers close the flag. Vague answers, or details that shift between the CV and the call, upgrade it.

Reserve artifact requests for load-bearing claims, the one or two the hire actually depends on. A link, a named stakeholder, a metric with its mechanism. For technical claims, a short set of screening questions that verify real skills beats any amount of CV archaeology. Reference checks then cover whatever is still open, and only that; references aimed at everything verify nothing.

What you should not do is silently downgrade a candidate on the flag alone. Half of these flags have innocent explanations, and the people most likely to trigger false positives (contractors, immigrants, career changers, parents) are not a group you want your process structurally rejecting.

Can software spot these flags for you?

Partly, and the division of labor matters. Once a CV is parsed into structured data, overlap detection is arithmetic: sort the date ranges, find the collisions. Language models go further and read for title-duty mismatch, pronoun vagueness and claims that never touch a specific; we cover the technique and its failure modes in using LLMs to flag CV inconsistencies.

Two caveats keep the human in the loop. False positives are routine, because the model sees the overlap but not the contracting arrangement that explains it. And under the EU AI Act, hiring systems sit in the high-risk category, where human oversight is a requirement rather than a preference. Both point to the same design: software surfaces the question, a recruiter asks it, and nobody is rejected by arithmetic.

That is the design Recruitifly ships. Fly parses CVs into structured profiles and scores candidates against the job, which is what makes timeline and consistency checks cheap enough to run on everyone, and every action it proposes (a shortlist, an outreach draft, a stage move) waits for a recruiter’s confirmation before anything happens. A flag becomes a question in your screening call, never an automatic no. We are in private beta; if you would like to see your own pipeline with the noise stripped out, talk to us.