Hemingway Editor vs Readability Analysis Tool (RAT)

Hemingway Editor scores writing for readability and highlights sentences and issues that may weaken or complicate a message. Editor Plus adds rewriting for wordy sentences and passive voice, grammar fixes, and controls for tone, length and style. It also provides document review with writing feedback and suggestions, and supports importing and exporting Word, Markdown and HTML files. The Classic desktop app works offline and keeps writing on the computer; it can publish to WordPress or export HTML and Markdown for a blogging platform or CMS. Hemingway Editor is available on web, macOS and Windows. The free plan is available. Individual 5K annual access costs $8.33 a month, billed as 100 USD annually, and includes 5,000 AI sentences per month. Hemingway Editor 3 desktop costs $19.99, paid once and does not include new AI features. Hemingway says document content is discarded after processing and is not used to train AI models. For analysis or rewriting, content is sent through Hemingway to an LLM provider, currently OpenAI, Google or Anthropic, and then discarded. Editor Plus does not yet have SOC 2 or ISO certification.

Readability Analysis Tool (RAT) checks German text for readability anomalies using configurable rules. It accepts .docx files and marks detected issues as comments in a separate copy with a -rat.docx suffix, leaving the original intact. RAT also creates an HTML report with statistics, readability formulas, anomalies, and a quality gate indicating whether text may be too trivial or difficult. Its documented formulas are Flesch and Wiener Sachtextformel. Report details can include reading and speaking time, sentence and word counts, average word and sentence measures, common parts of speech, and configured keyword percentages. Users can adjust anomaly rules and thresholds; the report records anomalies they reject as false positives or mark as incorporated. RAT runs from the command line using a Java JAR, with configuration and output-directory options. The project lists performance of 40 seconds for 10,000 words, or 50 pages. It runs on Linux and Windows, uses Java, UIMA, and DKPro Core, and is licensed under GPLv3.