AI Text Pattern Analyzer guide
Review measurable writing patterns without pretending that ordinary text can identify Claude, ChatGPT, another model, or a human author.
What this tool does
The AI Text Pattern Analyzer calculates a small set of transparent statistics from English prose. It measures variation in sentence length, repeated three-word sequences, repeated sentence openings, vocabulary variety, and the use of selected transition phrases. When a measurement crosses a disclosed review threshold, the report explains both the observation and several possible reasons for it.
This is not a model-authorship detector. The same pattern may appear in AI-assisted writing, templates, technical documentation, school assignments, accessibility-focused prose, translations, or carefully edited human work. A language model can also be prompted or edited to avoid common patterns. For those reasons, the result never presents an “AI probability” and never claims that a passage came from Claude or any other named service.
How to use it
- Paste at least 150 words of continuous English prose.
- Select Analyze writing patterns.
- Review the measurements before reading the highlighted signals.
- Consider each alternative explanation shown beside a signal.
- Copy the report if you need a record of the observations and limitations.
For a fairer comparison, analyze complete passages written for similar audiences and purposes. Comparing a short email with a formal research section is not meaningful because their expected structure differs. Remove quoted material if it was written by someone else.
Benefits
- Performs all analysis locally in the browser
- Shows the measurements behind every highlighted signal
- Avoids invented authorship percentages
- Requires a more meaningful sample instead of scoring a single sentence
- Provides a copyable report with a decision-use warning
Understanding the report
“Several patterned-writing signals found” means that at least three configured thresholds were crossed. “Limited patterned-writing signals found” means one or two thresholds were crossed. “No distinctive patterns found” means none were crossed. These labels describe only this tool's rules; they are not confidence levels and do not establish whether AI was used.
Sentence variation compares the standard deviation of sentence lengths with their average. Repeated phrases count three-word sequences used more than once. Vocabulary variety divides unique lowercase words by all analyzed words, so it naturally changes with topic and sample length. The transition count uses a short list of explicit connectors rather than a hidden model.
The practical notice is important: never use this report as proof of cheating, authorship, job performance, policy violations, or misconduct. For consequential review, use document history, sources, drafts, direct discussion, and clearly defined institutional procedures.