AI resume rewriters have a credibility problem. Most will invent experience if you let them. MeritSlate's rewriter has one absolute rule:
Never invent employers, titles, dates, degrees, or metrics.
Here's how to stay on the right side of that line.
Where the line is
Tailoring (fine)
- Reordering bullets so the most relevant ones go first
- Swapping synonyms when the JD uses a different word for the same skill
- Quantifying outcomes you actually achieved
- Moving skills into the summary when they're buried at the bottom
Lying (not fine)
- Inflating team size or budget
- Backdating an internal transfer so it looks like two companies
- Adding a tool you only saw a demo of
- Writing metrics you never measured
The test for every edit: could you defend the line, as written, in the interview? Tailoring changes emphasis and vocabulary; lying changes facts. The same rule governs quantifying your bullets — honest estimates pass, invented precision doesn't.
Two kinds of claim, two different rules
Every line on a resume is one of two things, and the distinction tells you how much freedom you have.
Verifiable facts — employer names, job titles, employment dates, degrees, certifications, licences, security clearances. These have a right answer that someone else holds a record of. You have zero editorial latitude here. Not "rounding up", not "simplifying", not "what it should have been called".
Descriptions of your work — which bullets you include, what order they're in, which vocabulary you use, what you choose to emphasize. This is where tailoring lives, and the latitude is genuinely wide. Two honest resumes for the same person applying to two different jobs can share almost no bullet text.
What actually gets verified
People misjudge the risk in both directions — panicking about bullet wording while casually adjusting an end date. A standard background check typically confirms employment dates, titles and degrees, because those come from records an employer or registrar holds. Nobody calls your old manager to ask whether the migration you led was really 40% faster.
That asymmetry is exactly why the facts column is non-negotiable. A discrepancy in a date or a title is findable, it's usually found after you've been hired, and it is treated as dishonesty rather than as a rounding error — regardless of how good you are at the job.
Employment gaps
Don't disguise a gap by stretching the dates on either side of it. Gaps are common, unremarkable, and easy to explain in a sentence. Falsified dates are none of those things, and they turn a non-issue into a fireable one.
Titles that don't match the market
If your title was internal jargon, translate it in a parenthetical rather than replacing it — the full pattern is in the keywords guide. You keep the verifiable fact and gain the searchable term.
"We" versus "I"
Claiming a team's work as solo is the most common honest-person mistake. The fix isn't to hedge everything into invisibility; it's to name your specific contribution. "Led the API workstream of a 6-person replatform" is stronger than both "replatformed the system" and "was part of a team that replatformed the system".
Skills you have but never used at work
A real category with a legitimate answer. If you learned a tool through a serious side project, a course with real output, or open-source contributions, you can list it — in a section that says what it is. A Projects section, or a skills list that distinguishes professional from project experience, is honest and useful. What's not honest is dropping it into your work history so it reads as paid experience.
Auditing what an AI wrote for you
If you use any AI tool to rewrite bullets — ours or anyone's — read the output against this checklist before it goes anywhere:
- 1
Every number: did I measure this?
Language models are fluent at inventing plausible percentages. This is the single most common fabrication.
- 2
Every tool and technology: did I actually use it?
Rewriters pull vocabulary from the job description. Some of it may describe work you never did.
- 3
Every scope claim: team size, budget, user count
'Cross-functional team of 12' is a fact, and it's an easy one for a model to hallucinate.
- 4
Every verb: did I do this, or did I watch it happen?
'Led', 'owned' and 'architected' are claims about your role, not decoration.
What to do when you're under-qualified
You still have two levers:
- 1
Lead with transferable experience
If the JD asks for Python but you used R heavily, highlight the statistical rigor — it's the underlying skill the hiring manager wants.
- 2
Use a portfolio link
One well-built side project beats three inflated bullets — it's evidence nobody has to take on faith.
And then apply anyway. Job postings routinely list more requirements than any real candidate has; the gap list existing is not a reason to disqualify yourself. It's a reason to be precise about what you do bring.
Want to see this in action? Run a Merit Score — you'll see where you're genuinely short vs just missing keywords you already have.
Frequently asked questions
- Is tailoring a resume for each job dishonest?
- No — tailoring changes emphasis, not facts. Reordering bullets, surfacing a relevant project, and using the posting's vocabulary for work you genuinely did are all fair. The line is crossed when you add a skill you haven't used or a result that didn't happen.
- Do I really need to tailor every single application?
- Tailor properly for roles you actually want, and accept lower odds on the rest. A focused ten-application week with tailored resumes reliably outperforms fifty generic sends, because generic resumes lose on exactly the requirement-coverage axes that decide ranking.
Written by
MeritSlate Team
The team building MeritSlate's deterministic scoring engine — nine evidence-anchored axes, a nine-vendor ATS parse simulation, and every point backed by receipts.
