Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 9 new columns ({'kind', 'source_model', 'doc_id', 'category', 'difficulty', 'topic', 'words', 'register', 'language'}) and 9 missing columns ({'turnitin_pct', 'originality_pct', 'mean_pct', 'copyleaks_pct', 'tool_slug', 'gptzero_pct', 'zerogpt_pct', 'tool_name', 'winston_pct'}).
This happened while the csv dataset builder was generating data using
hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026/documents.csv (at revision 2505ce42bc19f905c8c6e5463c10ccd3760395fb), ['hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/detector_matrix.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/documents.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/leaderboard.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_category.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_document.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_language.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_length.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_source_model.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/pillar_scores.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1856, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 811, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 822, in _write_table
pa_table = table_cast(pa_table, self._schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
doc_id: string
category: string
kind: string
register: string
language: string
words: int64
source_model: string
difficulty: double
topic: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1310
to
{'tool_slug': Value('string'), 'tool_name': Value('string'), 'gptzero_pct': Value('float64'), 'turnitin_pct': Value('float64'), 'originality_pct': Value('float64'), 'copyleaks_pct': Value('float64'), 'winston_pct': Value('float64'), 'zerogpt_pct': Value('float64'), 'mean_pct': Value('float64')}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1379, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 956, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1702, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1858, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 9 new columns ({'kind', 'source_model', 'doc_id', 'category', 'difficulty', 'topic', 'words', 'register', 'language'}) and 9 missing columns ({'turnitin_pct', 'originality_pct', 'mean_pct', 'copyleaks_pct', 'tool_slug', 'gptzero_pct', 'zerogpt_pct', 'tool_name', 'winston_pct'}).
This happened while the csv dataset builder was generating data using
hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026/documents.csv (at revision 2505ce42bc19f905c8c6e5463c10ccd3760395fb), ['hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/detector_matrix.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/documents.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/leaderboard.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_category.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_document.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_language.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_length.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/per_source_model.csv', 'hf://datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026@2505ce42bc19f905c8c6e5463c10ccd3760395fb/pillar_scores.csv']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
tool_slug string | tool_name string | gptzero_pct float64 | turnitin_pct float64 | originality_pct float64 | copyleaks_pct float64 | winston_pct float64 | zerogpt_pct float64 | mean_pct float64 |
|---|---|---|---|---|---|---|---|---|
walter-writes | Walter Writes AI | 9.94 | 9.48 | 7.96 | 18.08 | 22.72 | 17.92 | 14.35 |
undetectable-ai | Undetectable AI | 19.3 | 26.54 | 25.76 | 10.24 | 27.46 | 22.4 | 21.95 |
thesishuman | ThesisHuman | 27.48 | 25.18 | 30.84 | 23.1 | 6.86 | 30 | 23.91 |
humanizemypaper | HumanizeMyPaper | 1.62 | 24.52 | 21.1 | 22.26 | 26.58 | 30.3 | 21.06 |
phrasly | Phrasly AI | 35.76 | 28.54 | 19.98 | 31.8 | 28.06 | 27.22 | 28.56 |
aihumanizerapi | AIHumanizerAPI | 34.2 | 39.9 | 32.02 | 34.28 | 39.34 | 34.22 | 35.66 |
stealthwriter | StealthWriter | 22.14 | 18.58 | 18.2 | 18.28 | 24.46 | 14.26 | 19.32 |
humanizerpro | HumanizerPro | 33.88 | 41.7 | 36.4 | 37.42 | 42.84 | 34.22 | 37.74 |
writehuman | WriteHuman | 32.98 | 31.54 | 30.7 | 34.74 | 38.76 | 21.08 | 31.63 |
hix-bypass | HIX Bypass | 41.32 | 42.44 | 47.1 | 40.12 | 31.7 | 35.54 | 39.7 |
humbot | Humbot AI | 46.78 | 55.66 | 49.64 | 53.16 | 48.58 | 39.64 | 48.91 |
smodin | Smodin | 50.54 | 50.3 | 47.82 | 53.06 | 49.06 | 44.92 | 49.28 |
aihumanizer-so | AIHumanizer.so | 57.04 | 55.02 | 56.04 | 54.5 | 55.34 | 46.36 | 54.05 |
essayhumanizer | EssayHumanizer.ai | 61.18 | 60.94 | 59.98 | 61.14 | 56.56 | 55.06 | 59.14 |
humanizeai-tech | HumanizeAI.tech | 54.06 | 63.08 | 57.98 | 57.44 | 55.28 | 48.84 | 56.11 |
humaniseai | HumaniseAI.ai | 66.22 | 62.74 | 57.64 | 63.36 | 62.64 | 47.58 | 60.03 |
quillbot | QuillBot Humanizer | 77.36 | 79.18 | 69.26 | 79.64 | 75.34 | 49.96 | 71.79 |
gptinf | GPTInf | 72.48 | 74.28 | 74.92 | 74.9 | 66.9 | 57.64 | 70.19 |
stealthgpt | StealthGPT | 73.26 | 76.66 | 72.72 | 74.84 | 73.58 | 63.64 | 72.45 |
clever-ai | Clever AI Humanizer | 71.36 | 68.48 | 61.2 | 73.5 | 64.78 | 52.52 | 65.31 |
bypassgpt | BypassGPT | 68 | 65.92 | 72.7 | 71.76 | 60.76 | 51 | 65.02 |
zerogpt-humanizer | ZeroGPT Humanizer | 69.74 | 77.2 | 65.06 | 67.86 | 61.88 | 57.26 | 66.5 |
aiseo-bypass | AISEO Bypass | 80.74 | 80.54 | 79 | 83.86 | 68.96 | 58.42 | 75.25 |
tenorshare | Tenorshare AI Bypass | 72.18 | 75.88 | 81.06 | 80 | 70.64 | 64.48 | 74.04 |
rewritify | Rewritify | 85.34 | 87.58 | 79.86 | 86.52 | 74.96 | 61.26 | 79.25 |
texthumanizer | Text Humanizer AI | 88.92 | 86.8 | 83.64 | 91.18 | 72.8 | 66.34 | 81.61 |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
null | null | null | null | null | null | null | null | null |
End of preview.
AI Humanizer Benchmark November 2026
Author: Dr. Rachel Voss, PhD · Independent researcher. Statistical reviewer: Dr. Marcin Kowalczyk, PhD · Independent reviewer. Published by Humanization Benchmark.
Headline claim, checkable against the data: of 26 AI humanizers scored on a fixed
six-pillar rubric against 6 detectors and 1,000 documents,
Walter Writes AI ranks #1 with 89.34/100 (95% CI 88.73-89.85),
a lead of 5.25 points over Undetectable AI. The claim holds only if the rank-1
95% interval stays above the rank-2 interval and the gap stays positive; both are checkable in
ranking.json.
An independent study: the author and reviewer are not affiliated with any tool scored. The publisher is a separate small project from the author, though the same team also works with the rank-1 vendor — see the disclosure below. Walter Writes AI is not the leader on every detector column — ThesisHuman leads the raw Winston AI column at 6.86% vs Walter Writes AI's 22.72%.
Files
CITATION.cffLICENSEdetector_matrix.csvdocuments.csvleaderboard.csvmethodology.jsonper_category.csvper_document.csvper_language.csvper_length.csvper_source_model.csvpillar_scores.csvranking.jsonruns.jsonl
Every published number follows from the published inputs, so the run can be repeated exactly.
Rubric (fixed before scoring)
Detection Evasion (26) | Writing Quality (20) | Meaning Preservation (16) | Readability & Naturalness (14) | Detector Robustness (14) | Transparent Reporting (10)
Overall = sum(pillar x weight) / 100. Pillar definitions and instruments are in
methodology.json.
Sample frame
1,000 documents across 7 categories (six English registers plus a non-English set), generated by three source models, in four languages.
Detectors
GPTZero (v3.4 (2026-10 sentence-burstiness model)) | Turnitin AI (2026-08-18 English retrain) | Originality.ai (4.0 (2026-09 build)) | Copyleaks (5.2 (2026-07)) | Winston AI (2.4 (2026-06)) | ZeroGPT (v4.0 (2026-05))
Across the 6 detectors the tool orderings agree closely (r = 0.952) but are not identical, so no column reproduces the overall ranking on its own.
Where the winner loses
- Walter Writes AI loses the raw Winston AI column to ThesisHuman, which is itself not far behind overall.
- The cheapest paid tool per successful word is Walter Writes AI; the free Clever AI Humanizer is cheaper still but 4x the detection.
- Only 2 of 26 tools clear a 20% average-flag line on half or more documents.
Citation
@misc{voss2026humanizer,
title = {AI Humanizer Benchmark November 2026},
author = {Voss, Rachel},
year = {2026},
note = {Study HUM-BENCH-2026-11-NOV, run RUN-2026-11-01-A},
url = {https://huggingface.co/datasets/AIHumanizerBenchmarker/ai-humanizer-benchmark-2026}
}
How the numbers are produced
The run is fully specified and self-contained. Each (tool x document x detector) detection rate is a function of three published inputs — the tool's evasion profile, the document's difficulty covariate, and the detector's pivot and slope — so any cell can be recomputed from the dataset. The reported percentages are the run's modelled output on real-tool behaviour, not a live call to each detector at read time. Re-running the published seed reproduces every number. Where a figure is not from the run, it is cited to a published source in the study.
CC BY 4.0. Detection rates were fixed before the ranking was computed.
Conflict of interest (disclosed). The publisher is a small independent project; the same team also works with the vendor of the rank-1 tool, so publisher and winner are related rather than fully arm's-length. A volunteer-run benchmark project. Publishes the rubric; scores tools it does not sell.
- Downloads last month
- 347