Why we don't suggest more keywords

AI review shows you a short, defensible set of keywords rather than every term a posting mentions. That's a deliberate cap, and people occasionally ask why they aren't getting more. Here's the case.

1. Stuffing reads as word salad to the person deciding

A resume packed with terms stops sounding like a person. Recruiters are openly skeptical of broad skill lists precisely because a list doesn't show how a skill was used — and using it is the thing being hired.

Jobscan, which sells ATS keyword optimization and therefore has every reason to argue the other way, defines keyword stuffing as overloading a resume until it is no longer a true representation of your abilities, and warns that blatant stuffing leads to quick rejection once a recruiter notices. They also publish match-score targets well short of the maximum, on the grounds that returns diminish quickly and readability drops.

When the keyword-optimization industry tells you to stop adding keywords, that's worth hearing.

2. The returns fall off fast

Covering the must-haves matters. Covering the twenty-third nice-to-have does almost nothing, and it costs you readability — which is the thing that determines whether a human keeps reading.

This is why our chips prioritize must-haves and stop well short of listing everything. The goal is coverage of the skills the posting actually uses, not a matching exercise against its whole vocabulary.

3. Manipulation is now an active detection target

This is the part that has changed recently, and it changes the risk calculation.

Hiding keywords in white text or tiny fonts is now classified in the security literature as an attack, not a technique. A December 2025 academic benchmark tested invisible keywords and fabricated experience against AI resume screeners, found some manipulations exceeded 80% success, and proposed detection defenses that are already being built.

Employers are responding. Greenhouse's 2025 research found 41% of job seekers admit to using hidden text to bypass AI filters, and 52% of those who haven't are considering it. On the other side, 22% of hiring managers report having caught hidden instructions, and 91% of recruiters have spotted candidate deception of some kind.

Prevalence in the wild is lower than self-report suggests but rising: Greenhouse, processing on the order of 300 million resumes a year, found white text in about 1% of resumes in the first half of 2025. ManpowerGroup reports detecting hidden text in roughly 100,000 resumes a year, around 10% of what it scans with AI. Different populations and methods — but both mean the same thing. It's detectable at scale, and it's being detected.

4. Hidden text isn't hidden

Even without detection tooling, white-on-white text is still text in the file. It's fully readable to the parser, and it appears in plain black the instant someone selects all and pastes your resume into an email, a notes field, or a preview pane.

Recruiters describe finding it as an immediate, emotional rejection — at that point they've stopped evaluating a candidate and started evaluating someone who tried to trick them. Reporting has documented candidates who credit hidden prompts with more interviews, and in the same coverage, a recruiter who rejected someone specifically for doing it.

The asymmetry is the whole story: a modest possible upside against a rejection plus a permanent credibility loss.

Requirements aren't keywords

A posting that asks for a Bachelor's degree, five years in the role, or a fast-paced environment is stating a hiring bar, not a skill to weave into a bullet. Those sit under What this role requires, with a check if your resume already states them. They do not count in the keyword coverage line, and they never get a suggestion — you either meet them or you don't.

What we do instead

Show a short set, prioritized toward skills and tools the posting actually uses.
Flag must-haves so you know which gaps matter.
Suggest weaving a term into a bullet where you genuinely used it — not adding it to a list.
Keep hiring bars — a degree, a license, a years-of-experience line, a workplace condition — on What this role requires, not mixed in with keywords you can write.
Report coverage honestly, including when it's short. 7 of 9 stays 7 of 9.

The constructive rule

One covered must-have, sitting in a real bullet with a real result, beats a longer list every time.

Weak: adding cross-functional collaboration to your skills line.

Better: Ran the weekly release sync across engineering, support and sales; cut post-launch escalations by a third.

Same territory, completely different credibility.

What we won't tell you

We don't publish how many terms we'll surface, how we decide which ones are must-haves, or how coverage is determined. Publishing that would turn a coverage read into a target to farm — which is precisely the behavior this whole design exists to discourage.