Live weather on any website.

Three providers dominate embeddable weather. OpenWeatherMap offers the most generous free tier (60 calls per minute, current and 5-day forecast) and is the default for most free widgets, but accuracy outside the US lags behind specialist regional providers. WeatherAPI Pro tier starts around $4 per month and beats OpenWeatherMap on international forecast accuracy, severe weather alerts, and historical data. NOAA (National Oceanic and Atmospheric Administration) is the gold standard for US locations: public-domain data, no API key required, sub-1-hour alert latency, and the underlying source for nearly every commercial US weather product including weather.com and Apple Weather. Poper routes each request to the most accurate source for that region: NOAA for US ZIPs, WeatherAPI for international, OpenWeatherMap as a fallback. Most single-source widgets break when their one provider has an outage. Multi-source widgets keep working.

Two ways to detect a visitor's location for the right local forecast. IP-based geolocation (via MaxMind GeoIP2 or similar) works silently with no permission prompt and gets accuracy of about 50 kilometers in cities, 100-200 km in rural areas. HTML5 Geolocation API (navigator.geolocation) uses GPS and Wi-Fi triangulation for sub-100m accuracy, but requires an explicit permission prompt that 60-80% of visitors decline. For weather, IP-based is the right default: 50km accuracy is more than enough to show the right city forecast, and the lack of friction matters because most visitors never see the widget if it requires a click. HTML5 geolocation makes sense only when you genuinely need precise coordinates (storm tracking, hyperlocal nowcasting). GDPR-wise, IP-based geolocation is permissible under legitimate interest as long as the IP is not stored beyond the request and is not combined with other identifiers to fingerprint the user.

Numerical weather prediction models (GFS, ECMWF, NAM) have a measurable accuracy decay curve. Day 1 forecasts are about 90-95% accurate for temperature and 85% for precipitation. Day 3 drops to 80%, day 5 to 70%, day 7 to about 65%, and day 10 to 50% which is barely better than climatology (the long-term average). Some widgets advertise 14-day or 16-day forecasts but the back half of those is statistically guesswork. Poper caps the user-facing forecast at 7 days and surfaces only the high-confidence next-24h hourly outlook in detail. If you need longer-range planning data (vacation booking 30 days out, event scheduling), use seasonal climate normals instead of a daily forecast. The widget gracefully labels lower-confidence days so visitors are not misled.

Weather widgets are a hidden Core Web Vitals risk: third-party API calls on every page view, image-heavy weather icons, synchronous JavaScript, and large unscoped CSS bundles. The worst offenders move Largest Contentful Paint by 1-2 seconds and push Cumulative Layout Shift from 0.05 into the poor band, which costs you Google ranking. Poper's widget loads asynchronously below the fold by default, fetches forecasts from a global CDN edge cache (visitors never hit OpenWeatherMap or WeatherAPI directly, which also keeps your API costs at zero), uses scoped CSS that cannot bleed, inlines weather icons as SVG sprites instead of separate image requests, and clocks in under 35 KB gzipped. After embedding, run Lighthouse Mobile against your page; you should see no measurable change to Performance, Accessibility, or Best Practices scores.

Unlike reviews (Review, AggregateRating) or events (Event, Place), Schema.org does not yet have a widely-adopted WeatherForecast or CurrentConditions type that Google parses for rich results. Google does pull weather directly from its own data partners (mostly weather.com) for queries like 'weather in Paris', but it does not promote site-embedded weather modules to a SERP card. What you can do for SEO with a weather widget is more indirect: embedding a hyperlocal forecast on a landing page lifts time-on-page and pages-per-session, which Google reads as quality signals. Travel and event pages with embedded weather also rank better for long-tail queries like 'weather for Coachella 2026' because the page is now topically relevant to that query. Poper auto-injects descriptive alt text on every weather icon and a hidden plain-text current conditions block so the page text is fully crawlable, which is the closest practical equivalent to weather schema until Schema.org ships a stable type.