# How Humanizer Handles Writing Sample Conflicts Against Pattern Recommendations

> Humanizer writing sample conflicts override default AI patterns preserving author's unique style discover how this process works

- Repository: [Siqi Chen/humanizer](https://github.com/blader/humanizer)
- Tags: internals
- Published: 2026-09-12

---

**When a provided writing sample conflicts with Humanizer's established pattern recommendations, the sample's voice characteristics override the default AI-writing patterns in sections §1–§21 to preserve the original author's unique stylistic fingerprint.**

The `blader/humanizer` repository provides a specialized skill designed to humanize AI-generated text while prioritizing authentic voice characteristics. When you supply a writing sample, the system analyzes your specific rhythmic and stylistic markers to ensure the output reflects your authentic style rather than enforcing standardized "human-like" patterns that would normally override such preferences.

## How the Voice Override Mechanism Works

### Priority of User-Provided Samples

When Humanizer receives a writing sample, it immediately analyzes the text's sentence length, word choice, punctuation patterns, openings, and transitions. This analysis creates a stylistic fingerprint that takes precedence over the default recommendations. According to the *Voice* section in [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md) (lines 40–44), **the sample overrides the patterns below**, including specific sections like §6 which governs dash usage.

### Preservation of Stylistic Elements

In practice, this override behavior ensures that patterns normally flagged for removal are retained when they appear in your sample. If your writing frequently uses em-dashes—which Humanizer would typically remove—the system preserves them at approximately the same rate they appear in your sample. The rewrite follows your sample's rhythm and punctuation style rather than enforcing generic "humanization" standards that would strip these elements.

## Technical Implementation in SKILL.md

The override behavior is explicitly codified in the [Voice section of SKILL.md](https://github.com/blader/humanizer/blob/main/SKILL.md#L40-L44). The documentation states:

> "If the user gives a writing sample, read it first and match its sentence length, word choice, punctuation, openings, and transitions. **The sample overrides the patterns below**, including §6: if the sample uses dashes, keep them at about the same rate."

This directive ensures that sections §1 through §21—which contain the default pattern recommendations—are explicitly subordinated to the characteristics extracted from your provided sample. The system defers to your established voice rather than imposing standardized corrections.

## Practical Code Examples

The following examples illustrate how Humanizer preserves conflicting stylistic elements when processing a writing sample:

```python

# Example: Using Humanizer with a conflicting writing sample

from humanizer import Humanizer  # hypothetical import for illustration

sample_text = """
The new update — released last week — improves performance dramatically.
We’ve added several features, and the UI is now sleeker.
"""

# Humanizer will keep the em‑dashes because they appear in the sample,

# even though dashes are normally a pattern to remove (see § 8).

result = Humanizer().rewrite(sample_text)

print(result)  # Output retains the dashes, matching the sample's style

```

```python

# Example: Sample with a strong “not‑X‑but‑Y” contrast

sample_text = """
It’s not just about speed; it’s about reliability.
"""

# Humanizer respects the contrast because it reflects the user’s voice,

# overriding the generic “remove not‑X‑but‑Y” rule.

result = Humanizer().rewrite(sample_text)

print(result)  # The contrast is preserved as in the original sample

```

## Summary

- **Sample precedence**: When provided, a writing sample's stylistic characteristics override Humanizer's default pattern recommendations in sections §1–§21 of the skill definition.
- **Pattern preservation**: Elements like em-dashes that would normally be flagged for removal are retained at similar frequencies when present in the sample, as specified in [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md).
- **Voice matching**: The system analyzes sentence length, word choice, punctuation, openings, and transitions to replicate the original author's rhythm and flow.
- **Explicit override**: The Voice section (lines 40–44) explicitly states that the sample overrides the patterns below, ensuring predictable behavior when conflicts arise.

## Frequently Asked Questions

### What happens when a writing sample contains punctuation that Humanizer normally removes?

When your sample includes punctuation patterns like em-dashes—which Humanizer typically flags for removal under §6—the system preserves these elements at roughly the same rate they appear in your sample. The sample's override directive ensures your specific punctuation style survives the rewriting process rather than being sanitized to match generic preferences.

### Does a writing sample override all Humanizer rules or just specific sections?

The writing sample overrides the patterns described specifically in sections §1 through §21 of the skill definition. According to [`SKILL.md`](https://github.com/blader/humanizer/blob/main/SKILL.md), these sections contain the default AI-writing pattern recommendations that are subordinated to the voice characteristics extracted from your provided sample.

### How does Humanizer analyze a provided writing sample?

The system performs a multi-dimensional analysis of your sample, examining sentence length distribution, vocabulary choices, punctuation frequency and placement, sentence openings, and transition patterns. This analysis creates a comprehensive stylistic profile that guides the rewrite process, ensuring your unique voice characteristics are replicated in the final output.

### Can Humanizer apply generic pattern recommendations if my sample contains conflicting styles?

No, the architecture explicitly prioritizes the provided sample's voice over generic rules. When you supply a writing sample, Humanizer is designed to honor your authentic stylistic choices—including patterns that conflict with standard "humanization" recommendations—ensuring the output sounds like you rather than a generic humanized template.