How accurate is the FCC National Broadband Map?

Date the data before you judge it

Almost every complaint about the map is really a complaint about its vintage, and the vintage is knowable.

Fixed availability is collected through the Broadband Data Collection, which runs on two as-of dates a year: June 30 and December 31. Providers do not file on those dates. They file in a window that opens after them and closes about two months later. The eighth collection is the clean example: the FCC announced it on December 18, 2025, opened the filing system on January 2, 2026 for data as of December 31, 2025, and required that “all availability and subscription data must be submitted no later than March 2, 2026.” The next cycle took the same shape — the window for data as of June 30, 2026 “open[ed] on July 1, 2026, and close[d] on September 1, 2026.”

Now do the arithmetic for today. The September 1, 2026 deadline has only just passed, and a vintage does not appear on the map the moment the window shuts — the Commission has to process, validate and publish the filings first. So the freshest availability claim you can read right now describes either the end of June 2026 or, more likely, December 31, 2025. Either way it is months old, and everything built, sold, retired or capacity-capped since then is missing by design. That is not a bug in the map. It is what a twice-yearly regulatory filing is.

The Commission is explicit that the map carries history as well as the present: under the data archival policy announced in June 2026, “the four most recent vintages of BDC data will remain published and available to view on the NBM (i.e., data from the past two years),” with older vintages available only as downloads. If you are comparing two readings of your address taken months apart, check which vintage each one came from before concluding anything changed.

There is a second, quieter vintage underneath the first. Provider claims are not free-floating points; they are laid onto a location layer, the Broadband Serviceable Location Fabric, which is versioned separately. The FCC’s notice for the current cycle is unambiguous: the updated Fabric “must be used by filers of fixed broadband availability for reporting their availability data as of June 30, 2026,” and licensees were pointed to “the June 2026 Fabric data.” Corrections to that layer land on releases, not on the day you submit them — Fabric challenges had to be in by September 1, 2026 “to be reviewed ahead of the next update of the Fabric in December 2026.” So when you look at a pin, you are looking at one snapshot of who claims service laid on an older snapshot of what counts as a home.

The two directions the map is wrong

Errors are not random noise. They run in two directions with different causes, and knowing which one you have determines what you do about it.

Overstatement — the map says served, nobody will sell you service. The structural cause is written into the reporting standard. The FCC requires providers to report availability for every location where, as of the as-of date, they “have a subscriber or can otherwise provide fixed broadband service within 10 days at a standard installation charge.” A “standard broadband installation” is itself defined as one “with no charges or delays attributable to the extension of the network of the provider,” completable “not later than 10 business days after the date on which the service request is submitted.” The Government Accountability Office read that standard the way a consumer would: FCC “allows a provider to report that it serves a location even if it does not actually serve it, as long as the provider claims it could begin serving that location within 10 business days,” and this “could lead to artificially inflating availability information.” A green pin can legitimately mean “we believe we could hook this up quickly,” not “we have a line to your house.”

Layer on data that providers report about themselves — the FCC validates, verifies and audits those filings, but nobody measures service at your address — and, for fixed wireless, coverage that is modeled rather than surveyed. A fixed wireless provider filing coverage maps “must model its data using a customer premise equipment (CPE) antenna height between four and seven meters,” which is a generous assumption for a house. The Commission is currently asking whether to relax it further, and asked in the same breath: “Is there a danger that terrestrial fixed wireless coverage maps will overstate availability if we allow providers to model coverage assuming antenna heights greater than seven meters?” Overstatement is the direction you should expect.

Understatement — real service exists, the map does not know. This is the new-construction case, and it is a timing artifact. A home finished after the Fabric snapshot has no location record for any provider to report against, so nobody can claim it even if the fiber is in the ground out front. A house completed in, say, August 2026 missed the June 2026 Fabric and the June 30, 2026 availability filing with it, and the earliest filing that can carry it is the one for data as of December 31, 2026. Understatement is quieter than overstatement because nobody complains about a service they did not know to ask for, but it is the direction that costs you an option.

What the independent reviewers found

The FCC’s own pages describe the process. They cannot grade it. That job falls to GAO and to the state broadband offices that have to spend money against the map.

The hard number is this. Between November 2022 and November 2023, filers submitted approximately 8 million challenges to the map’s fixed availability data. The FCC accepted — meaning it deemed the evidence sufficient — and forwarded to providers almost 4 million of those, “about half of which providers ultimately conceded, resulting in updates to the map.” Roughly two million conceded errors in a single year is a floor, not a ceiling: it counts only the mistakes someone bothered to contest and could evidence. GAO’s own reading is that “such successful challenges demonstrate that a nontrivial amount of inaccurate data ends up on the National Broadband Map despite FCC’s validations, verifications, audits, and enforcement referrals.”

The process critique is separate from the error count. In its report published April 17, 2025, which carried 14 recommendations, GAO concluded that “the accuracy of the broadband availability data on the map is uncertain” because “FCC has not documented or assessed the sufficiency of its processes for ensuring the information’s accuracy.” The point is not that the FCC does nothing; it is that the quality checks were not documented well enough for anyone outside to evaluate them. And this is where a stale mental model will mislead you: in a follow-up published July 22, 2026, GAO recorded that “FCC has not yet addressed the recommendations as of July 2026, but officials reported taking steps to do so,” while agencies continued relying on the map to target tens of billions of dollars in federal broadband funding and “the accuracy of the map’s availability data is uncertain.”

That funding link is why this matters beyond your own order. Washington State’s broadband office puts the consequence in one line: BEAD funding is based on the state’s unserved population, so “if this map is inaccurate, your location may be mistakenly listed as served. If that happens, state or federal sources are less likely to fund new service or upgrades.”

What a map listing actually means

How to correct the map when it is wrong about you

The first two steps cost you an afternoon and answer the question you actually have; the rest fix the public record.

1. Date the release. Establish which as-of date the data comes from before you argue with it. A listing is a claim about a June 30 or December 31 snapshot that reached the map months later, and a gap is silence about everything since.
2. Ask the provider in writing. Put the address through the provider’s own order flow, and if it stalls, ask support directly: “Can you confirm in writing whether you can install service at this street address, and if not, what the reason is?” Save the reply with its date. This is your real answer, and it is also the evidence the FCC rule requires. If the answer is “available but not orderable,” that gap has its own causes — drops, wiring, capacity and waitlists.
3. Classify the error. Service claimed but unbuyable is an availability problem. A missing, misplaced or mis-addressed home is a location problem against the Fabric. They go to different challenge types with different effects.
4. File the matching challenge. An availability challenge needs “evidence and details of a request for service (or attempted request for service), including the date, method, and content of the request and details of the response from the provider, or evidence showing no availability at the disputed location (e.g., screen shot, emails).” A bare assertion that nobody serves you is not a filing. A Fabric challenge is now a separate rule of its own — 47 CFR 1.7006(g), effective July 24, 2026 — and asks the FCC to add a missing location, correct an address, or reposition a point. Washington State’s broadband office notes that the FCC “is accepting challenges on a rolling basis,” so there is no window to miss for a consumer challenge.
5. Expect a months-long clock, and solve your connectivity separately. Once the provider is alerted it has 60 days to reply, and the location is shown as “in dispute/pending resolution” while that runs. If the provider accepts your allegation it “shall submit a correction for the challenged location in the online portal within 30 days of its portal reply” — 90 days end to end, and that is the fast path. If it denies the allegation, “the provider and the challenger shall have 60 days after the provider submits its reply to attempt to resolve the challenge,” and only if that fails does “the Commission … have 90 days to review the evidence and make a determination.” That is up to 120 days of party process before the FCC’s own clock even starts, so about seven months is the realistic worst case. Nothing about that timeline installs a line at your house, so keep working the ordering problem in parallel.

Caveats, confidence and limits

This page can tell you how the dataset is built and how wrong it has been measured to be. It cannot tell you what is available at your address — nothing published can, because availability is address-specific and only the provider’s own order system, keyed to your exact street address, settles it. If you are working from a ZIP code rather than a street address, start with why ZIP-level results cannot answer an address-level question.

Two more limits worth stating. First, this guide is about the federal fixed-broadband dataset. Cell carrier coverage maps are a different data source with different failure modes, covered in why carrier coverage maps can be wrong, and a carrier’s own marketing map has no challenge process of its own. Second, we describe the map; we do not republish its data. The underlying location layer is licensed, and we verify a dataset’s license before building anything on top of it.