# Claude, Rhine and GPT Zero: inspect the writing behind the score

Read the original and transformed passages together before deciding whether a rewrite is useful.

[HTML](https://lukejbyrne.com/library/guides/gz9lkwbeari/)
Content language: en
Author: Luke Byrne
Video published: 2026-03-18
Guide published: 2026-11-02

[Watch the video](https://www.youtube.com/watch?v=GZ9lKwBearI)
Video language: English

## Keep an unchanged copy of each sample

The March test runs Claude-generated text and a random LinkedIn post through an AI humanizer, then scans results with GPT Zero. The title names Rhine. The description does not include a direct humanizer product link or reproduce the complete sample passages.

Begin with the Claude writing test at 0:00 and the humanizer at 0:32. Preserve the original so the rewrite can be checked for more than its detector result. The task is to assess what happened to the writing: information, meaning and natural phrasing.

1. Keep the original sample and its initial detector score together.
2. Save the humanized version as a separate text.
3. Compare the two before making further changes.

## Treat the detector result as one observation

The first-result chapter at 1:33 reports a 100% human score. That is the reported output of the detector for one sample, not proof of who wrote it or a reliable authorship verdict. A change in score does not show that the content became more accurate or useful.

The second sample begins at 2:03, with its humanization at 3:09. Comparing another input helps reveal whether the transformation preserves the point of a passage or merely changes its wording.

1. Inspect the score attached to the exact saved version.
2. Read both versions for changed facts, missing qualifications and awkward transitions.
3. Avoid treating a favourable score as permission to skip the actual reading.

## Edit for meaning and coherence

At 3:27 the video reads the transformed LinkedIn output and flags strange results. The description emphasises that humanization can lose meaning and coherence, even when a detection scan looks favourable. That tradeoff is the useful editorial finding to inspect.

Keep the parts that improve clarity and restore any meaning the transformation damaged. Read the final paragraph continuously: does it make the same point, with concrete wording and understandable connections? The retained scripting resource is https://ailuke.short.gy/GZ9lKwBearI_scripti. This article does not claim that any detector can establish authorship from a score.

## The takeaway

Judge a rewrite by fidelity and readability. The detector score describes a scan result, not the quality or authorship of the text.

## Workflow resources

- [ailuke.short.gy](https://ailuke.short.gy/GZ9lKwBearI_skool)
- [ailuke.short.gy](https://ailuke.short.gy/GZ9lKwBearI_scripti)

## Chapters

- [0:00 — Testing Claude's writing with GPT Zero](https://www.youtube.com/watch?v=GZ9lKwBearI&t=0s)
- [0:32 — Trying the AI humanizer tool](https://www.youtube.com/watch?v=GZ9lKwBearI&t=32s)
- [1:33 — First result: 100% human score](https://www.youtube.com/watch?v=GZ9lKwBearI&t=93s)
- [2:03 — Testing a random LinkedIn post](https://www.youtube.com/watch?v=GZ9lKwBearI&t=123s)
- [3:09 — Humanizing the LinkedIn post](https://www.youtube.com/watch?v=GZ9lKwBearI&t=189s)
- [3:27 — Reading the humanized output (it gets weird)](https://www.youtube.com/watch?v=GZ9lKwBearI&t=207s)
- [4:29 — Final thoughts and takeaways](https://www.youtube.com/watch?v=GZ9lKwBearI&t=269s)

## Sources and limitations

Written companion prepared 2 November 2026 from this exact published video description and its chapter sequence. The full recording was not independently watched or transcribed for this article, and the products and reported results were not retested. Checklists are editorial adaptations of the documented workflow. Model availability, interfaces, prices and offers may have changed since publication; use the retained original resources for current requirements.

[Original video](https://www.youtube.com/watch?v=GZ9lKwBearI)