Jev Content Guard

Filters fraud, ads, AI slop, spam, clickbait, plagiarism & toxicity, and highlights keywords on page using TypeSafe Jev AI.

🚀 Jev Content Guard filters out fraud, advertising, AI slop, spam, clickbait, "info-gypsy" schemes, and toxicity on any web page, and extracts & highlights core key words from selections or whole pages — powered by the TypeSafe Jev AI content analysis API. ✨ Features Automatic content scanning — the extension analyzes page text in the background as you browse (including dynamically added content via MutationObserver). Selectable scan scope — choose in the popup between Entire page (automatic scanning) and Selection only (nothing is scanned until you ask for it). Excluded pages / domains — configure a list of pages, domains, or wildcard patterns (e.g. example.com, github.com/my-repo/*) that will be excluded from automatic detection and scanning. The list is shown only for the Entire page scope (it applies to automatic scanning) and is preserved when switching to Selection only. Compact card popup — every block sits in its own rounded card (API token · thresholds · scan scope · exclusions). The scope and exclusion labels carry hover tooltips, a successful save flashes a green ✓ on the Save settings button, and the footer strip reminds you how to extract key words (validation errors appear there in red). Context menu actions: Check selection: Select any fragment of a page, right-click and choose “Jev: check selection” to analyze content categories and pin matching badges next to it. Extract main words from selection: Select any fragment, right-click and choose “Jev: keywords in selection” to ask Jev which words have the most meaning and core significance. Extract main words from page: Right-click anywhere on the page and choose “Jev: keywords on page” to extract key conceptual words across the main content of the entire page. Key Word Extraction & Highlighting: Extraction without generation: Since Jev evaluates parallel structured questions rather than generating text, words are tokenized client-side, filtered against multilingual stop words (English, Russian, Spanish, French, German, Chinese), and evaluated in parallel via Jev noul questions. Strict threshold (≥ 60%): Only words judged to carry the core meaning of the text qualify. Jev semantic deduplication: Synonyms and morphological variants are compared pairwise via Jev in the context of the text; redundant terms are deduplicated automatically. Original order & casing preserved: Words stay in the exact casing and sequence in which they appear in the original text. In-text highlighting & interactive navigation: Words above the threshold are highlighted in the document with <mark>. Clicking any keyword chip in the floating card smoothly scrolls and centers the viewport on that word, cycling through multiple occurrences. Sticky viewport placement: The floating keywords card stays fixed in view during scrolling. Automatic cleanup: Closing the card, clicking any link on the page, or closing/navigating away from the page immediately unmounts the card and restores text nodes back to normal.