Affiliate disclosure check: find affiliate links and ad labels near the top of each article

What it does

Checks in bulk whether articles make their advertising nature clear, following Japan's stealth-marketing rule that took effect in October 2023.

• Detects affiliate links: Amazon Associates (tagged links, amzn.to), Rakuten, A8.net, ValueCommerce, Moshimo, afb, AccessTrade, JANet, felmat, Rentracks, CJ, Awin, Impact, rel="sponsored" and more
• Checks whether disclosure wording (ad, PR, promotion, affiliate, sponsored, in Japanese and English) appears near the top of the article (first 300 visible characters by default, configurable)
• Classifies each page: no disclosure, only ambiguous wording (e.g. "product provided"), disclosure not near the top, disclosed near the top, or no affiliate links
• Flags contradictory wording ("personal opinion", "not sponsored") next to an ad label, disclosures only at the end, and links found only in site header or footer
• Outputs a table, JSON and CSV

How to use

python scripts/affiliate_disclosure_check.py ./content --json report.json --csv report.csv
python scripts/affiliate_disclosure_check.py --url-list urls.txt --top-chars 400

URLs are fetched with one GET each, with a delay between requests (1 second by default). Add custom short-link domains with --extra-domain.

Requirements

• Python 3 (standard library only)
• Network access when checking URLs

Caveats

• Not legal advice. It checks wording and position only; font size, color, text inside images and JavaScript-inserted labels are not detected, so the agent reviews flagged pages by hand.
• The rule regulates the advertiser (the business supplying the product); the report never calls a blogger's page illegal.
• "Near the top" is a practical convention, not a character count set by law.