The 4-country hard-block list and why it's so short
BLOCKED_REGIONS = ["CN","KP","IR","RU"] — shorter than Meta's 5-7 list and shorter than Spotify's 7. Reasons: (1) Pinterest is small business-wise (450M MAU vs Facebook's 3B), not at Meta's scale where every sensitive market needs separate negotiation; (2) the content form is images + shopping discovery — no Facebook Live, no Reddit-style political debate, no TikTok short-video controversy, so political sensitivity is low; (3) ad revenue depends on a handful of high-value markets (US / CA / UK / DE / BR), so emerging-market bans don't move the top line much. On the 4 bans: China has GFW-blocked pinterest.com since March 2014, with the image CDN i.pinimg.com poisoned alongside; North Korea is fully offline; Iran banned Pinterest after the September 2022 Mahsa Amini protests; Russia's Roskomnadzor banned it in 2022 citing Ukraine-related content.
CDN architecture — i.pinimg.com + Fastly + Cloudflare trio
Pinterest's image delivery stack is relatively unique among high-traffic sites: (1) the main CDN is its own i.pinimg.com with origin on AWS and Fastly as global edge cache — Fastly is known for low latency + VCL programmability, and Pinterest has been a top Fastly account since 2015-2016; (2) Cloudflare serves as backup CDN to handle Fastly outages — during the June 8 2021 Fastly global outage, Pinterest degraded to Cloudflare and kept serving; (3) originals are cached at 6 resolutions (originals / 736x / 564x / 474x / 236x / 200x75), so one image may have 6 subresource URLs; (4) video (Idea Pins) uses Cloudflare Stream + Mux hybrid with HLS adaptive bitrate. AF3 implication: low image latency doesn't mean Pinterest's core features (login, search, recommendations) work. AF3 probes pinterest.com (main), api.pinterest.com (API gateway), i.pinimg.com (CDN), and accounts.pinterest.com (auth) independently.
Pinterest Lens — mobile-only image search
Pinterest Lens, launched in 2017, is a 'search by image' feature — point your phone camera at an item (clothing, food, furniture) and Lens returns visually similar Pins and buyable links. Tech stack: Pinterest's proprietary visual embeddings (historically ResNet-family, migrated to ViT + CLIP-class models in 2023), indexed across 300M+ images. Key constraint: Lens is **mobile-app only — there is no web version**. Pinterest hasn't officially disclosed why, but it's generally attributed to the app's direct camera access + local preprocessing saving backend compute. User impact: web users can't use Lens — you need the iOS app (requires a non-China Apple ID; the China App Store pulled Pinterest in 2014) or the Android APK (Google Play version). Chinese users wanting the full Lens experience need: (1) an international Apple ID; (2) a VPN reachable to pinterest.com; (3) note that Lens queries go through api.pinterest.com not i.pinimg.com, so CDN speed doesn't affect Lens accuracy.
Anti-VPN posture — why Pinterest scores 1★
AF3's 1★ tier includes Pinterest / Reddit / Telegram / most AI tools — commonalities: no IP-type discrimination, no ASN blacklist, lenient on datacenter IPs. Pinterest IP scoring: residential 30, quality_vpn 30, normal_vpn 28, datacenter 27 — essentially flat. Rationale: Pinterest's business is ad-based and wants maximum eyeballs; VPN users still see ads, so blocking them is net-negative. Compared to Netflix 5★ (datacenter 0), Instagram 4★ (datacenter 8), Pinterest's leniency gives users maximum freedom. This also means mainland China users can use Pinterest on any stable VPN (mainstream VPN services, self-hosted nodes, corporate VPN exits — all work) without hunting residential IPs like Netflix requires. Conversely, if you can't reach Pinterest from a non-blocked country, 99% it's local ISP DNS hijacking or a corporate firewall — not Pinterest defenses.
Pinterest Predicts annual trend report — actually taken seriously
Every early December Pinterest publishes Pinterest Predicts — data-mined trend forecasts for fashion, home, food, weddings, and beauty for the coming year, based on 10B+ searches, 100M+ board saves, and regional/demographic slicing. Industry credibility: Pinterest claims > 80% hit rate since 2020 and is widely cited by Vogue, Cosmopolitan, Condé Nast titles. Real hit rates are disputed (many 'hits' are vague and unfalsifiable), but for designers, brands, and e-commerce buyers it has practical value — Pinterest users have higher purchase intent than Instagram users, 'Save' signals 'plan to buy later,' and the data leads sales by 3-6 months. 2024 confirmed hits: Cherry Red as the dominant lipstick/apparel color, Mob Wife Aesthetic, Blueberry Milk Nails. For AF3 users: the report is publicly accessible at predicts.pinterest.com with no login or regional lock — any VPN node works.
2018 IPO and the shape of the business — ad platform or e-commerce
Pinterest went public on NYSE (ticker PINS — note not Nasdaq as the brief said, correcting here) on April 18 2019, priced at $19 with a 28% day-one pop. Financials show 97% of revenue comes from ads (Promoted Pins) and only ~3% from commissions + Verified Merchant Program fees. This pins the positioning: **Pinterest is fundamentally an ad company dressed in e-commerce-discovery clothing.** Versus Instagram Shopping and TikTok Shop: those are closed-loop (see → click → in-app checkout → shipping); Pinterest is see → click → jump to the brand's own site, a longer conversion funnel. This is the main reason Pinterest's multiple has trailed Meta/Snap. Impact on AF3 users is minimal — ads or e-commerce, the VPN-Pin experience is the same. In 2025 Pinterest launched a deep Amazon partnership — Amazon ads + products appearing directly in Pin feeds, with advertisers becoming Amazon merchants — potentially the biggest revenue inflection for the next 3 years.