Pinterest says it is expanding its skin tone search tool, which debuted in US in 2018 and uses machine vision to sort beauty-related pins, to five more markets
James Vincent / The Verge :
Context & Ripple Effects
Pinterest's skin tone filter began as a 2018 beta feature that let US users narrow beauty search results by skin tone range — one of the first mainstream applications of machine vision tuned explicitly for inclusivity. The rollout to five more markets extends it along a path Pinterest has been walking since 2015, when it began localizing search in France, Germany, Brazil, and Japan on the strength of nearly half its users being outside the US.
First-order effects
- Beauty-related searches in five additional markets now return results sorted by skin tone range, shifting discovery behavior from unfiltered pin feeds to tone-specific browsing for those users.
- The move lands alongside Pinterest reporting Q2 revenue of $1.18B (up 18% year over year, above estimates), even as lukewarm Q3 guidance pushed shares down more than 8% after hours.
Second-order effects
- Beauty brands and advertisers buying into Pinterest's discovery surface gain a skin-tone-targetable audience beyond the US for the first time, widening the commercial footprint of an inclusivity feature originally framed as a social good.
- With monthly active users up 11% to 640 million, expanded inclusive search becomes an engagement differentiator Pinterest can lean on while its sales guidance disappoints investors.
Third-order effects
- The 2023 body-type algorithm update — trained on 5B+ images to surface diverse bodies — shows the same playbook repeating: build an inclusivity-focused vision system, prove it in one market, then scale it globally. If the pattern holds, fairness-tuned machine vision shifts from differentiator to table stakes in visual discovery, raising the baseline expectation for how search engines classify people-adjacent imagery.
The trend: Consumer platforms are adopting a two-track approach to AI internationalization: launch inclusivity-trained machine vision domestically first, then export it through the same localization machinery built for earlier market expansions.