Enhance your X (Twitter) experience with Twat Doc: a powerful userscript featuring customizable content filters to hide spam, foreign languages, they/them lib content (🏳️⚧️, 🏳🌈, 🍉, 💉, etc), and low-quality posts; automuting based on user scores; visual tweaks like relative timestamps, larger buttons, and direct video links to view videos with your brower's native player
2025-09-19 - Added performance.clearXhrPostCacheOnUrlChange to settings and defaulted to true which fully addresses the memory leak, but with a slight side effect that causes all info added to posts to be lost on back navigation.
Ever since Trump's 2016 victory, the landscape of the entire internet has no longer been the same. Twitter was one of the many platforms to respond to this event by introducing TweepCred, the social credit system used to boost/deboost content. Suddenly, small liberal accounts with but a few hundred followers began achieving massive reach, many times getting 100k-300k likes and millions of views on their posts. Not even conservative accounts with 100k+ followers could achieve such reach on their best day.
Fast forward to today and we see that nothing has changed. We've all witnessed the thousands of vile posts on X celebrating the killing of Charlie Kirk. All were from far-left, unverified accounts with only a few hundred followers, and their posts regularly received upwards of 300k likes and millions views on their posts. Meanwhile, large, verified conservative accounts struggled to reach a fraction of those numbers with their posts that lament Kirk's passing.
Here is an unverified account with 539 followers and massive like/view-to-follower ratios because the Dorsey-era algorithms favor liberal content. In this case, the user is grandstanding over Kirk's death:
This one, while not Kirk-related, still illustrates how an account with only 852 followers can achieve engagement numbers rarely seen by the largest conservative accounts:
Enter Twatter Feed Doctor. In addition to providing a suite of features to remove cruft from the feeds, it also seeks to expose the Dorsey-era algorithms that are still in effect today. As you're scrolling through the feeds, you will be able to spot liberal accounts on sight as they are notorious for adding liberal-centric keywords, hashtags, and emojis to their usernames and bios. You will likely notice that they receive far more likes and views than your average account.
If you at least enjoy that one feature, please promote this userscript on X so that we can bring awareness to the biased algorithms. Imagine how much more victorious we could be in future elections if the visibility playing field were finally leveled once and for all.
This userscript enhances the X (formerly Twitter) experience by providing customizable content filters, automuting, and various quality-of-life improvements. Below is a succinct description of each content handler available in the script. These handlers process posts in feeds, allowing removal, modification, or scoring based on specific criteria. Enable or disable them via the settings; they activate on the next feed load without requiring a full page refresh. Note that enabling many handlers may cause performance issues on slower devices.
Note: Content is not removed from feeds unless [code]general > enableDeletes[/code] is enabled. Otherwise, it will display the qualifying information in the post template.
About content removal: X tracks viewed content in For You feeds. Removing posts from API responses before they appear in the feed may cause X to contuously refetch them, potentially blocking new content while X attempts to clear its internal content delivery queues. The [code]experimental > clickNotInterestedIn[/code] option allows such posts through, but auto-clicks "Not interested in this post" in the caret menu, notifying X via API to stop showing them and hopefully improve feed flow. Enable with deletes if feeds appear stagnant or empty.
Flags and scores authors based on indicators like pronouns, specific emojis (e.g., 🏳️⚧️, 🏳🌈, 🍉, 💉), or hashtags (e.g., #ACAB, #BLM) in posts, with lower scores for more matches. Supports automuting if scores hit thresholds. Processor-intensive due to regex usage; optional flag emoji checks (e.g., 🇪🇺, 🇮🇳) are experimental and may yield false positives.
Tracks and hides previously viewed posts stored in browser local storage, with cleanup schedules based on hours to prevent localStorage bloat. Configurable age and view count thresholds for hiding; optional auto-click of 'Not interested' in post menus to influence platform algorithms (use cautiously).
Enhances the user interface and display elements on X for better usability and accessibility. These options modify post templates, timestamps, and media handling without altering core functionality.
Video links added to post template:
Filters out foreign-language posts from feeds, especially in trends and searches. Configurable language confidence thresholds reduce mislabeling errors; options to hide banned languages or Grok-translated posts. Recommend adding language filters like 'lang:en' in [code]search > args[/code] for backend enforcement.
Provides granular control over content types from users, including allowing or hiding reposts, quotes, self-reposts, self-quotes, and unavailable quotes. Ideal for reducing self-promotion or redundant content.
Analyzes engagement metrics like view-to-follower or like-to-follower ratios to identify potentially biased or low-quality content. Configurable like count thresholds; options to show extended metrics or always display ratios for all posts.
Forcibly removes content from muted, blocked, or blocking accounts, including their reposts and quotes, even in feeds where X normally shows them (e.g., profiles, communities).
Cleans the Replies tab in user profiles by removing non-reply content like posts, reposts, and quotes.
Removes all community-related content from feeds like For You or trends. Future plans include whitelisting joined communities.
Removes subscription-locked posts. Untested with subscribed accounts; may affect promoted content from followed subscriptions.
Controls visibility based on account attributes, such as hiding content from users with default profile pictures.
Cleans up feeds flooded with hashtag and mention spam, especially in political trends. Configurable thresholds hide posts exceeding a set number of hashtags or mentions. Applies universally to all feeds.
Detects and notifies about users spamming trends or searches by tracking post counts in feeds. Includes optional scoring multipliers to adjust user quality scores based on post frequency, aiding in muting spammers. Scoring feature is experimental and may need refinement.
Improves search result quality by hiding posts where hashtags, mentions, or search terms dominate the content beyond a configurable percentage threshold. Options to include or exclude hashtags, mentions, and search terms in the calculation for targeted filtering.
Handles removal of posts referencing unavailable content.
Hides content from verified business accounts (yellow/gold star badges).
Removes posts containing $ symbols, targeting stock, crypto, or financial content.
Hides content from accounts marked as parody.
Removes posts from accounts with 'elon' or 'musk' in the name, excluding @elonmusk.
In searches or trends, hides content from accounts whose usernames contain the search query.
In this example where "Kiefer" was trending, a good portion of the feed results were accounts with "kiefer" in the authors' name fields.
Removes posts with links matching domains in the author's bio. Effective for filtering news or self-promoted content, but may hide legitimate links like YouTubers' own videos.
These content handlers are either a work in progress or need refinement before they're ready for public use. Many of these are features I began working on to suit my own purposes and may remove content you wish to see. It's best to keep these disabled.
Alerts on inconsistent capitalization in racial terms (e.g., 'black' capitalized but not 'white'). Currently non-functional and should remain disabled.
Hides content from gimmick video accounts based on username patterns. Criteria hardcoded; disable to avoid unintended filtering.
Removes posts containing hardcoded banned links. User-defined links planned; keep disabled unless matching personal preferences.
Color-codes posts by age for quick staleness assessment. Disable until colors and thresholds become configurable.
Miscellaneous experimental filters for testing new features. Keep disabled as they are not fully developed.
