Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsAdding more text made my writing linter’s score fall enough to change a draft from review-required to PASS—even though the original flagged issues remained. The problem was not that the checker stopped finding them: it combined two kinds of measurements in one score, and extra words diluted one of them.
What changed when I tested the same post two ways
In a first-person account published by Frank Chu on October 1, 2026, the same published post received different results depending on whether the checker analyzed the entire file or just the body. The author reports that the whole-file version contained 930 words and scored 2, producing PASS. The body-only version contained 746 words and scored 3, producing REVIEW / REWRITE. Both versions had the same three listed issues. Chu attributes the difference to the word-count denominator and the removal of some dashes in the front matter. Source
As an Amazon Associate I earn from qualifying purchases.
A second example made the dilution clearer: a score of 9 at 78 words fell to 3 at 648 words after the author added text that did not trigger the same objections. The issue count rose from one to three because the added material introduced mild flags. A lower score therefore did not mean fewer findings. Source
Free tools Windows power users keep installed
One-click scans. No signup required.
Why adding words lowered the score
The checker combined two unlike kinds of tests. Word-level checks worked as set-based rules: a flagged phrase contributed once, rather than adding more to the score every time it was repeated. Em-dash and bold-formatting checks, by contrast, were density-based ratios measured against word count.
#1 Best Overall
When these contributions were combined into one score and compared with one threshold, adding words could reduce the density-based contribution. That reduction could offset the set-based terms, making the overall score low enough to pass even when findings stayed the same or increased. This describes the behavior of Chu’s checker, not a universal design used by writing linters. Source
What the checker now analyzes
Chu changed the analysis input to exclude text that was not intended to represent the prose under review. The checker removes an initial HTML comment note, front matter, fenced code, indented code and inline code before calling the analysis function. Notes and metadata can distort measurements; code can also produce false word-level matches—for example, interpreting struct.unpack as English “unpack.” Source
The title and description are evaluated together in their own region rather than being mixed into the body score. Chu’s stated reason is that the title’s short length could fall below the detector’s ten-word minimum if assessed alone. Source
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Why cleaning the input is not a complete fix
Removing notes, metadata and code makes the input better match the text the checker is meant to judge. It does not fix the scoring design: the body still receives one scalar score against one threshold, so actual prose padding can continue to dilute its density-based contribution. Source
Rank #3
Chu’s proposed fuller repair is to score the set-based and density-based rule families independently and let either family fail the post. That preserves the distinction between “did this trigger occur?” and “how frequently does this formatting appear?” instead of allowing one measurement to compensate for the other. Source
What the archive check showed—and did not show
Chu reports applying both approaches to 100 archive posts: whole-file scores were higher in 40 cases, while only one verdict changed. The author characterizes the practical damage as limited, but says the gate had produced a number that was trusted for the wrong reason. These are the author’s reported results, not an independent benchmark or evidence about how common the failure is across other tools. Source
The author also says the experiment took four minutes once the idea of deliberately trying to defeat the checker occurred. The useful lesson is not that every linter can be broken in the same way; it is that a quality gate should be tested against plausible ways of changing its input without improving the thing it is supposed to measure. Source
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →How to pressure-test a writing quality gate
- Keep the findings visible. Compare issue counts and individual findings alongside the aggregate score. A lower score is not proof that the underlying problems disappeared.
- Check each measurement’s denominator. Identify which checks count occurrences and which calculate a rate per word, line or other unit. Extra unaffected text can shift a rate.
- Separate the text regions. Decide whether notes, metadata, code, titles and descriptions belong in the measurement, and test short regions against any minimum sample-size requirement.
- Test compensation between rule families. If a score averages or sums distinct checks, ask whether a favorable change in one family can conceal a failure in another.
- Try a padding test. Add text that does not resolve the original findings, then see whether the score or verdict changes. Treat this as a test of the gate’s logic, not a way to improve the writing.
Chu’s closing question is a useful prompt for any automated writing or code gate: “If you have an automated quality gate on your own writing or code, when did you last try to defeat it on purpose?” Source
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
