Hacker News AI Detector

Hacker News AI Detector is a web tool that scores every story on Hacker News's front page for AI-written content, using the Pangram AI Detection API to label posts as Human, Mixed, or AI-Generated. The page mirrors the live HN top 30, showing each story's AI probability percentage and enabling users to sort, filter, and compare rankings.

What is Hacker News AI Detector?

It's a real-time dashboard that analyzes the text of each Hacker News front-page story using Pangram v3.3, an AI detection model with 99.98% stated accuracy. It takes the story URLs from HN, feeds the article text to Pangram, and outputs a probability score (0–100% AI) plus a label (Human, Mixed, AI, or Check Failed if analysis fails). The tool runs entirely in the browser (no installation) and is maintained by salahadawi.com, as shown by the site's branding.

Key Features

• Real-time HN scoring — Analyzes the current top 30 stories, updating with each page load; today it scored 88 stories across the front page and new submissions.
• Filterable story list — Filter by AI probability range (e.g., only Human stories) or by label (Human, Mixed, AI, Check Failed). Reset filters with one click.
• Trend charts — Shows 14-day trend, 7-day average, and 30-day average of AI percentage; today 15% of stories were AI-generated, up 2pp week-over-week.
• Comparison view — Side-by-side ranking of the top 12 stories by HN upvotes vs. AI probability, with connecting lines to show discrepancy.
• Per-story details — Each story card includes HN rank, upvote count, comment count, time posted, and a specific AI% (e.g., 98% AI for the #1 story).
• Browse all analyzed articles — A separate page lists every story ever scored (6,586 all-time as of the data shown), with the same scoring labels.

Who is it for?

• Hacker News readers who want to quickly identify which front-page stories might be AI-generated slop, so they can skip low-quality content. For instance, they can filter to see only Human-labeled stories.
• Content moderators and editors who monitor HN for trends in AI-generated submissions, using the trend data and per-story scores to spot suspicious patterns.
• Data journalists researching the prevalence of AI content on social news sites — they can use the browse-all page to export or analyze historical scores.

What can you do with Hacker News AI Detector?

• Identify AI-generated stories before reading them: Scan the front page and see each story's AI probability at a glance; the #1 story today scored 98% AI, so you can skip it.
• Compare HN's popularity ranking with AI likelihood: In the comparison view, see how the most upvoted stories differ from the most-AI-suspected ones — revealing whether algorithmic or human judgment favors AI content.
• Track the AI trend over time: Use the 14-day trend chart and daily percentage to see if AI-generated submissions are increasing (up 2pp week-over-week).

How does Hacker News AI Detector work?

The site crawls Hacker News every few minutes to fetch the current top 30 stories. For each story, it retrieves the article text (extracted from the linked page), sends it to Pangram's AI detection API, and stores the score. Results are displayed in a ranked list with color-coded labels and a percentage. Users can interact with the dashboard to filter, sort, or compare stories without any account or payment.

FAQ

How accurate is the AI detection?

Pangram v3.3 claims 99.98% accuracy, though the page notes some stories fail the check (e.g., if the article text cannot be retrieved). The tool does not specify false positive/negative rates.

What does the "Check failed" label mean?

It means Pangram could not analyze the story, often because the linked page was inaccessible, behind a paywall, or didn't contain extractable text. Those stories are marked unscored and excluded from the AI percentage aggregates.

Is Hacker News AI Detector free?

Yes, the web interface is free to use — no sign-up or payment required. It appears to be a personal project maintained by salahadawi.com.

Can I see older data?

Yes, click "Browse every analyzed article →" to see a paginated archive of all 6,586 stories scored historically, each with its AI score and label.