An AI-resilient text rendering technique for reading and skimming documents is introduced, called Grammar-Preserving Text Saliency Modulation (GP-TSM). This method uses recursive sentence compression to identify successive levels of detail beyond the core meaning of a passage, which are de-emphasized by rendering words in successively lighter but still legible gray text. The lighter the text color, the earlier it was cut in the backend's recursive sentence compression process. The darkest subset of text can be read as grammatical sentences that preserve as much of the semantic value of the original document as possible, and every successive level of lighter text can be added to these darkest sentences—adding detail without modifying their grammaticality. A lab study (n=18) showed that participants preferred GP-TSM over pre-existing word-level text rendering methods and were able to answer GRE reading comprehension questions more efficiently. The study evaluated GP-TSM in two user studies. A preliminary study with 18 participants tested a semi-automated version of GP-TSM, while a summative study measured its impact when fully automated. The results showed that GP-TSM improved reading comprehension and the reading experience, with participants finding it more efficient and easier to use than traditional text rendering methods. The method was found to be effective in helping readers bypass certain words during skimming while maintaining an uninterrupted reading flow. The study also found that interactive granularity control provided value for readers, allowing them to adjust the level of detail they wanted to see. The design of GP-TSM is based on the principle of preserving the integrity of the original content while making key information more salient. The system uses opacity modulation to highlight or de-emphasize text, with lighter text indicating less important information. The method was found to be effective in supporting reading and skimming, with participants able to navigate the full complexity of a text, shifting focus seamlessly between different levels of semantic coverage. The study also found that GP-TSM was resilient to AI errors, enabling readers to notice, judge, and recover from automated decisions they disagreed with. The study compared GP-TSM with other text rendering methods, including font opacity modulated by unigram frequency (WF-TSM). The results showed that GP-TSM was strongly preferred over WF-TSM, with participants finding it more effective in helping them complete non-trivial reading comprehension tasks. The study also found that GP-TSM was effective in supporting reading at multiple levels of detail, with participants able to decide how much detail they wanted to read and, in case they wanted a closer read, enable them to do so without requiring any extra action on their part. The method was found to be effective in supporting skimming without interrupting flow, with participants able to skim text while minimizing the impact on their natural reading flow. The study also found that GP-TSM was effective in supporting reading in different contexts, with participants able to use it in various settings, much like other text rendering methodsAn AI-resilient text rendering technique for reading and skimming documents is introduced, called Grammar-Preserving Text Saliency Modulation (GP-TSM). This method uses recursive sentence compression to identify successive levels of detail beyond the core meaning of a passage, which are de-emphasized by rendering words in successively lighter but still legible gray text. The lighter the text color, the earlier it was cut in the backend's recursive sentence compression process. The darkest subset of text can be read as grammatical sentences that preserve as much of the semantic value of the original document as possible, and every successive level of lighter text can be added to these darkest sentences—adding detail without modifying their grammaticality. A lab study (n=18) showed that participants preferred GP-TSM over pre-existing word-level text rendering methods and were able to answer GRE reading comprehension questions more efficiently. The study evaluated GP-TSM in two user studies. A preliminary study with 18 participants tested a semi-automated version of GP-TSM, while a summative study measured its impact when fully automated. The results showed that GP-TSM improved reading comprehension and the reading experience, with participants finding it more efficient and easier to use than traditional text rendering methods. The method was found to be effective in helping readers bypass certain words during skimming while maintaining an uninterrupted reading flow. The study also found that interactive granularity control provided value for readers, allowing them to adjust the level of detail they wanted to see. The design of GP-TSM is based on the principle of preserving the integrity of the original content while making key information more salient. The system uses opacity modulation to highlight or de-emphasize text, with lighter text indicating less important information. The method was found to be effective in supporting reading and skimming, with participants able to navigate the full complexity of a text, shifting focus seamlessly between different levels of semantic coverage. The study also found that GP-TSM was resilient to AI errors, enabling readers to notice, judge, and recover from automated decisions they disagreed with. The study compared GP-TSM with other text rendering methods, including font opacity modulated by unigram frequency (WF-TSM). The results showed that GP-TSM was strongly preferred over WF-TSM, with participants finding it more effective in helping them complete non-trivial reading comprehension tasks. The study also found that GP-TSM was effective in supporting reading at multiple levels of detail, with participants able to decide how much detail they wanted to read and, in case they wanted a closer read, enable them to do so without requiring any extra action on their part. The method was found to be effective in supporting skimming without interrupting flow, with participants able to skim text while minimizing the impact on their natural reading flow. The study also found that GP-TSM was effective in supporting reading in different contexts, with participants able to use it in various settings, much like other text rendering methods