How LLM Brought Back a Google Maps Feature (Claude Skill Included)

In 2018, Google Maps introduced 'Your Match' – a percentage indicating how likely a user would like a restaurant, based on their past reviews. The author relied on this feature until it was quietly removed in 2023. To recreate it, they used a user-based collaborative filtering approach: find people whose ratings correlate with yours, then recommend places they liked that you haven't tried. However, scraping Google Maps directly proved difficult due to virtualized lists and ignored synthetic scroll events, so they switched to two paid APIs: DataForSEO (for reviews of seed places) and Apify's contributor-scraping actor (for reviewers' histories). The pipeline collected ~730 reviewers and ~32,500 reviews, producing ~4,000 candidate places. The scoring formula weighs supporters by affinity (overlap with your seed likes) and dampens for activity (review count) to avoid bias from prolific but indiscriminate reviewers. Costs were significant: Apify charged $0.0015 per review row (total $48.67), while DataForSEO was cheaper per review but lacked contributor-level extraction. The final skill and script are available in a repository, with a warning that the method is not a real recommender system and doesn't correct for popularity bias.