Score low-quality risk, not “AI guilt”
The extension reads visible feed text in the browser. It scores low-signal patterns such as generic phrasing, engagement bait, vague authority, repeated structures, thin external links, and comment duplication.
Reddit-first Chrome extension concept
Clean up AI slop before it takes over your feed.
AI Slop Blocker is a browser extension concept for people who want Reddit, X, and LinkedIn feeds with less repetitive AI slop, bot-like replies, SEO spam, and engagement bait. It does not label author identity. It scores low-quality risk, explains why a post was folded, and lets you reveal it with one click.
Reddit firstX / Twitter nextLinkedIn scorer later
User research
Help shape the Reddit-first Chrome beta.
Join the private research list and pick the feeds you want cleaned up first. Your email is sent through a server-side endpoint only — not stored in browser localStorage or exposed in static files.
Reddit AI Slop blockerStart where community signal and comment quality matter
Chrome extension conceptFold noisy posts without changing Reddit itself
Local by defaultScore visible posts in your browser before any optional cloud workflow
Conservative controlsBadge low risk, fold medium risk, hide only high risk when enabled
Direct answer
What is AI Slop Blocker?
AI Slop Blocker is a feed-quality blocker for low-quality risk, not an AI-authorship detector. It looks for combinations of weak signals: template openings, generic advice, engagement bait, dense cliches, missing specifics, promotional links, and repeated comment patterns.
The important distinction: a single signal should not trigger action. A dash, a bullet list, or a polished tone is not enough. The blocker should fold only when several low-signal patterns appear together and no strong human signal offsets them.
- Primary use case: make Reddit feeds easier to scan.
- Privacy angle: score visible posts locally by default.
- Trust angle: every folded item shows a score and reasons.
How it works
How AI Slop Blocker decides what to fold.
Protect human signal
Specific numbers, dates, prices, version numbers, code, logs, screenshot details, firsthand experience, sourced claims, clear disagreement, and subreddit-specific context all lower the risk score.
Fold first, reveal always
The MVP should badge low-risk items and fold medium-risk items. Hidden-by-default behavior is reserved for high-risk patterns later, and the recommended first version keeps folded posts one click away.
More precise than “AI detector” claims
AI Slop Blocker should never label a post as “AI.” It labels risk: likely low-signal, repetitive, promotional, or engagement-bait content.
Safer than keyword muting
Keyword mutes miss disguised slop and hide useful AI discussions. Combined signals make false positives easier to handle.
Built per platform
Reddit slop, X slop, and LinkedIn slop do not look the same. Each platform needs its own scorer and thresholds.
What AI Slop Blocker should fold — and what it should keep
AI slop is not simply “content made with AI.” Useful summaries, code explanations, translations, and drafts can still help people. The problem is low-quality feed noise: posts that look polished but add no evidence, no lived detail, no specific answer, and no reason to trust the author.
AI Slop Blocker is being designed for people who open Reddit to find human signal and instead hit recycled growth hooks, vague expert claims, bot-like replies, and thin link drops. The first version focuses on Reddit because users can validate the filter quickly in real communities; later versions can use separate scorers for X and LinkedIn.
Signals that raise risk
- Template openings and repeated listicle structures.
- Generic advice with no example, source, test, price, date, or version.
- Engagement bait such as “thoughts?” wrapped around obvious claims.
- High-density cliches, vague authority, and fake urgency.
- Thin links, affiliate-like promotion, or repeated product recommendations.
- Comments that repeat the same pattern across unrelated threads.
Signals that protect a post
- Concrete numbers, dates, prices, versions, commands, logs, or code.
- Firsthand experience, screenshot descriptions, tests, or failure details.
- A clear disagreement, tradeoff, caveat, or subreddit-specific answer.
- Sources, citations, or named tools that match the claim.
- Authors, subreddits, or topics you whitelist locally.
Why local feedback matters
Each folded item should offer two simple actions: Not slop and Hide more like this. Those choices should be saved locally to tune your own thresholds and rules, not used to build a public shame list or upload your feed by default.
How to judge if it works
The right metric is not “AI detection accuracy.” The right question is whether folded posts are usually low-value. Before launch, AI Slop Blocker should test 300-500 real samples labeled slop, borderline, and not slop, then optimize for high precision first. Missing some slop is acceptable; folding useful posts is the real product failure.
AI Slop Blocker FAQ
Is this an AI detector?
No. It does not promise to prove whether text was generated by AI. It scores low-quality risk from combined signals and describes the reason: for example, Score: 78, Reasons: generic phrasing / engagement bait / low specificity.
Will my Reddit or LinkedIn feed be uploaded?
The planned MVP is local-first. Visible feed text is scored in the browser by default, with no feed upload required.
What platform comes first?
Reddit first, because low-quality risk is easiest to validate where subreddit context, comments, and moderation norms are visible. X and LinkedIn should use separate scorers after the Reddit blocker is useful.