Cleaning text from clutter and traces of AI
Insert the text and get an analysis: where is the water, where is the office, where are the revolutions by which machine generation is recognized. With line-by-line edits and an explanation for each.
Machine text is produced not by a dictionary, but by a rhythm: sentences of the same length, paragraphs of the same length, and a final summary that retells what was said. The tool shows the edits line by line: quote, what the defect is and what to replace it with. Separately lists places that work and that should not be touched during general correspondence.
AI answers are free, but limited by the overall site limit. If the limit is reached, come back later: repeated clicks will not speed up the response.
Machine text is recognized by its rhythm, not its words.
It is commonly believed that the generated text is revealed by individual words. In fact, what gives away more is the rhythm: sentences of the same length, paragraphs of the same volume, symmetrical structures in each of them and the obligatory final paragraph, which retells everything said above. A person doesn’t write like that because a person gets tired, gets carried away and changes the pace.
The second marker is statements without content. “Quality product”, “individual approach”, “team of professionals” pass any literacy test and do not report anything. Everyone writes them, so the reader scrolls through them, but the search engine does not see them as different from thousands of the same pages.
Therefore, the tool shows edits line by line rather than giving you different text. The substituted text does not teach anything and usually loses what was alive in the original. There is a ready-made option, but it comes last and is marked as one of the possible ones.
A separate section is what to leave as is. When there are a lot of edits, there is a great temptation to rewrite everything, including good places. They were usually the reason the text worked at all.
What the tool intentionally doesn't do
Does not insert numbers, deadlines, benchmarks and names of companies that did not exist in your input. If the output requires a number that you did not provide, in the answer will be marked “to clarify” rather than a plausible number. This limitation is worth in the system prompt and works against the main property of language models: It’s difficult to complete what’s missing.
The request text is sent to the AI provider, and the response is returned to the browser. The site does not save these texts in the database. To limit expenses separately the number of requests and the anonymized connection identifier are taken into account. If the free limit is reached, the tool will offer to come back later.