Research

State of US Internet Speed 2026

Median download, upload, and ping speeds by U.S. state — measured from real, consenting visitors to SpeedCheckTest and updated continuously.
speedchecktest.com/study/state-of-us-internet-speed-2026Ongoing studyLive methodology

This is an ongoing study. Every number comes from speed tests run by visitors who explicitly accepted the consent banner — anonymized, aggregated by state, with no IP addresses stored. We publish a state's data once it reaches a minimum sample size, so this page grows as the dataset grows.

Loading study data…

Median download speed by state

We're still collecting data. The first states appear here automatically once they cross the minimum sample threshold.

Medians shown in Mbps. A state is published once it has at least 30 consented tests in the past 12 months.

Key findings

  • Findings will appear here automatically as state-level data crosses the sample threshold.

Methodology

We publish only what we can substantiate. This section describes exactly how the dataset is built so journalists and readers can judge the numbers on their own merits.

How tests are measured

Each visitor runs SpeedCheckTest in their browser: ping is the minimum of 8 HTTP round-trips, download and upload are measured over short parallel-stream transfers. Results include ping (ms), download (Mbps), and upload (Mbps). The tool is free and works on any device.

How state is determined

State attribution comes from Cloudflare's cf-region header, which maps the visitor's IP to an ISO 3166-2 region code at the network edge. The raw IP address is never read or stored by our server — only the state code and test values are saved.

Consent

A record is kept only when the visitor explicitly clicks "Accept" on the consent banner. Visitors who decline are excluded entirely. You can opt out at any time by declining consent.

Thresholds and percentiles

We report medians (p50) as the headline metric because they resist outliers better than averages. We also show p10/p25/p75/p90 for download, upload, and ping in the downloadable CSV. A state must reach a minimum sample (30 tests in the trailing 12 months) before it appears, so small or noisy samples never drive claims.

Limitations

  • Self-selection — visitors who run a speed test skew toward people actively checking their connection, which can differ from the general population.
  • Plan mix — results reflect whatever plans testers happen to have, not the advertised speed of every ISP.
  • Network conditions — Wi-Fi congestion, device capability, and the chosen test server all influence a single result.
  • Coverage — states appear only after they accumulate enough consented tests; sparsely populated areas take longer.

Frequently asked questions

Why use medians instead of averages?

Medians are resistant to extreme values, so one gigabit fiber connection doesn't distort a state's typical speed. Averages are more sensitive to a handful of very fast or very slow tests.

How many tests does a state need?

At least 30 in the trailing 12 months. We may raise this for states where we want extra confidence.

Is my IP address stored?

No. State attribution happens at the Cloudflare edge and only the region code is saved. Your IP never reaches our database.

Can I download the data?

Yes — once a state is published, the full state-level dataset is available as a CSV from the "Download full dataset" link next to the table.

When will all 50 states be covered?

As traffic grows. We intentionally don't publish thin samples, so a state can remain "collecting" until it clears the threshold.

For journalists

SpeedCheckTest maintains this study to give journalists and researchers an independent, methodologically transparent dataset on real-world U.S. internet speed. We're happy to provide on-the-record commentary, early access to aggregate data, chart-ready assets, and methodology walkthroughs for deadline-driven stories.

Request data or commentary: [email protected] · [email protected]. Full press kit: Press & Media Kit.

We won't claim numbers we can't substantiate. What we offer is a verifiable tool, a published methodology, and data you can check yourself.