Reading Your Own Resume as a Skeptic
The last pass on a resume should not be a proofread. It should be a factual audit: go through the page line by line asking how you know each claim is true and who else could confirm it. Anything with no answer gets narrowed or cut. This takes twenty minutes, catches almost everything an AI draft introduced, and doubles as interview preparation.
It is a different pass from checking spelling and formatting, and it has to be done separately, because the two use different attention. Reading for typos makes you read words; reading as a sceptic makes you read claims.
Set up for it
Three conditions make the pass work:
Distance. Not immediately after writing. A day is ideal, a few hours is enough. Fresh from drafting, every sentence still feels like a memory rather than a claim.
A different format. Print it, export a PDF, or view it at a different size. Familiar layout makes eyes skip; unfamiliar layout makes them read.
Your own record open beside it. The private document holding your real dates, titles and plain-language descriptions — the facts file assembled before drafting began. Auditing a resume against nothing is just re-reading it.
The four questions
For every line, in order:
1. How do I know this? The answer should come immediately: a document, a person, a memory of a specific event. If you have to reconstruct a justification, flag the line.
2. Who else would say the same? Name a person. If nobody comes to mind, the claim is unwitnessed, which is fine for some claims and dangerous for ownership claims — those are exactly what a reference conversation ends up testing.
3. What would I be asked about this? Write the obvious follow-up. If you cannot answer your own follow-up, the line is a liability.
4. Did I write this sentence? Or did a tool write it and you approved it? Approved sentences deserve extra suspicion, because they arrived fluent and were never tested against memory.
The fixed-field sweep
Separately from the line-by-line, do a fast pass over the fields that either match an external record or do not: employer names, dates, titles, qualifications, credential status, and every number. Read each one against your source document rather than against your own sense of what is right.
This is quick because there are not many of them, and it catches the highest-severity errors. Dates in particular should be read from a payslip or letter, since they are the element most reliably confirmed by someone else’s records, and a reformatting step by any tool can silently normalise them.
The patterns to look for
Some specific things worth counting rather than judging:
- How many bullets make you the sole agent of a result? If nearly all of them, the page is misattributing collective work, and the distinction between your contribution and the team’s result needs restoring in a few places.
- How many numbers can you source? Any you cannot, delete now. Not soften — delete.
- Does the summary agree with the dates below it? Generated summaries routinely claim a seniority the timeline contradicts.
- Which skills appear only in the skills list? Those are unevidenced by construction, and the weakest of them should come off.
- How many bullets could describe almost anyone? A vague bullet is not a safe bullet; it is an uninformative one that also hides overstatement.
What to do with the flagged lines
Three outcomes, and choosing quickly matters more than choosing perfectly:
Narrow it. Keep the subject, drop the elements that outrun the evidence. Usually the best result.
Replace it. Same work, different claim — one you can support. Responsibility and artefact claims survive when outcome claims do not.
Cut it. When the unverifiable part was the only content.
Resist a fourth option that presents itself as caution: hedging the sentence into vagueness. “Helped contribute to improvements” is not a narrowed claim, it is an unfalsifiable one, and it takes up space without informing anyone.
The output is a prep sheet
By the end you have, for every line on the page, a short answer to how you know it and what you would be asked. That is not a by-product — it is most of the work of preparing for the conversation, since the questions your resume invites are largely predictable from the page itself.
Keep the notes. They are useful the second time, and they record which claims you deliberately narrowed and why, which stops a stronger version of the same sentence reappearing in a later draft.
Do it every time you tailor
Tailoring a resume for a specific posting reintroduces risk, because tailoring means adding emphasis and sometimes adding claims. A skill added because the posting asked for it, a verb strengthened to match the seniority in the advert — these are the changes that need auditing most and get audited least, since the tailoring pass feels like a small edit to an already-checked document.
A short version of the audit is enough here: read only what changed, and apply the four questions to it. The discipline is the same one that makes the whole approach work, which is that a tool can help with wording but the claims stay yours to verify, exactly as the split between fixed facts and editable prose implies.