How experimental simulation and deep transformation work.

Text-watermark evidence is statistical, not visible markup

A generation-time text watermark can influence choices among plausible next tokens using a keyed, context-dependent rule. A compatible detector accumulates evidence over a passage; it does not locate a “watermarked word,” hidden Unicode string, or metadata field inside pasted prose. The general red-green-list construction and related schemes illustrate why a matching detector needs knowledge that is not visible in the text alone. Kirchenbauer et al.

Ordinary AI-authorship detection is different: it uses observable patterns in text to estimate similarity to a training distribution. It cannot recreate a provider's private key or production detector. Google's SynthID-Text publication describes a distinct watermark system, not a public specification for Claude. SynthID-Text publication

What the current score measures

The active local detector is claude-wm-sim-v1. It creates an experimental 0–100 simulation from multiple observable sequence signals: character/token n-gram distributions, local token transitions, distribution variation, and consistency across overlapping text windows. The displayed score is an unweighted synthesis of those components and is labelled Simulation score, never a provider probability.

Suitable English prose receives a simulation. Short, code-heavy, quote-heavy, unsupported-language, or otherwise low-evidence passages return Inconclusive rather than a misleading score. Reference confidence is currently limited: the corpus/versioning tooling remains in the repository for a future validated release, but no generic human-versus-AI dataset is relabelled as Claude watermark evidence.

Prepared, but not invented

Anthropic's transparency documentation says it has worked with industry and academia on watermarking developments and is preparing for applicable requirements. It does not provide this app with an official Claude watermark endpoint, request schema, credential type, or detection key. Anthropic transparency documentation

The code has separate detector adapters. The active ExperimentalClaudeDetector powers today's simulation. An inactive AnthropicOfficialWatermarkDetector is reserved for future documented integration; no placeholder endpoint or fabricated result is used.

Recompose from meaning, not from synonyms

Transformation uses one existing rewrite-provider request. The writer is instructed to first form an internal content blueprint covering claims, support, facts, names, values, links, terms, relationships, tone, audience, and intent. It then composes fresh prose from that blueprint rather than editing each source sentence in place.

Users may preserve the original tone or select Natural, Professional, Casual, Concise, or Technical. The transformation is not guided by simulation score, and it is never repeated until a score crosses a target.

Quality checks happen before the final comparison

Before a rewrite is returned, the server protects and restores numbers, dates, currencies, URLs, email addresses, code, and citations. It checks extractable factual values and named terms, exact-phrase overlap, wording change, sentence/paragraph restructuring, and length change. A shallow rewrite or low local preservation estimate receives one quality-focused recomposition attempt. This correction is based on writing quality, never on the detector score.

When transformation succeeds, the application performs one independent local simulation on the transformed text and displays before/after signals alongside rewrite quality, wording independence, structural change, and fact-preservation status. The comparison describes the experimental simulator's response; it does not establish that an official provider watermark was removed.

Extensive editing can affect text-watermark detection systems generally; Google makes this point in its SynthID documentation. Google SynthID documentation

Focused processing, clear boundaries