How many US traffic cameras are actually broken
Of the 6,216 cameras we checked, 691 were serving a placeholder instead of a road — about 11%. And 0 of those were still being published as working.
Data measured September 2, 2026 · Roadze
These figures are being re-measured
The check behind these numbers gave up on slow cameras too quickly, and that bias ran one way: an agency’s “camera unavailable” card is served from cache in about 270 ms while a real road photo takes about 740 ms, so the working cameras were the ones that dropped out of the count. The share below is therefore too high. A corrected sweep — every camera, not a sample — replaces it within a day.
When a public traffic camera breaks, the state's website rarely says so. The camera stays in the list, the map still shows a dot, and what you actually get is a black image reading "STREAM NOT AVAILABLE" — or the local equivalent.
The gap between what is published and what works is not small. In Georgia, 234 of the 391 cameras we checked were dead — 60% of the sample.
Broken cameras by state
| State | Checked | Broken | % broken |
|---|---|---|---|
| Georgia | 391 | 234 | 60% |
| Massachusetts | 93 | 37 | 40% |
| Idaho | 203 | 71 | 35% |
| Alaska | 145 | 40 | 28% |
| Hawaii | 204 | 38 | 19% |
| Nevada | 206 | 38 | 18% |
| Michigan | 207 | 29 | 14% |
| Arizona | 208 | 27 | 13% |
| New York | 207 | 25 | 12% |
| Virginia | 198 | 18 | 9% |
| Pennsylvania | 156 | 13 | 8% |
| Washington | 208 | 15 | 7% |
| Tennessee | 207 | 13 | 6% |
| Maine | 98 | 6 | 6% |
| Kansas | 140 | 8 | 6% |
| Indiana | 169 | 8 | 5% |
| New Hampshire | 195 | 9 | 5% |
| South Carolina | 203 | 9 | 4% |
| Illinois | 389 | 16 | 4% |
| North Dakota | 203 | 7 | 3% |
| Ohio | 203 | 5 | 2% |
| Vermont | 89 | 2 | 2% |
| Wisconsin | 212 | 4 | 2% |
| Louisiana | 218 | 3 | 1% |
| Alabama | 190 | 2 | 1% |
| Nebraska | 202 | 2 | 1% |
| Connecticut | 208 | 1 | 0% |
| California | 205 | 0 | 0% |
| Colorado | 207 | 0 | 0% |
| Mississippi | 198 | 0 | 0% |
| Montana | 204 | 0 | 0% |
How we measured
We fetch the published image of each camera and compare it with the others. A placeholder is the same file for many cameras at once — two different cameras cannot be looking at exactly the same pixels — so a repeated image hash means a notice board, not a road. Cameras that keep serving an identical image for over 30 minutes across several checks are counted as dead too.
Percentages are over the cameras we actually checked in each state, not over its whole network — we sample up to 200 per state. Cameras that did not answer at all are excluded rather than counted as broken: a timeout is not proof of failure, and counting it as one would inflate every number on this page.
The check runs again every day, so a camera that comes back is un-flagged automatically. These figures move.
What this means if you just want to see the road
On Roadze the cameras we know are broken are pushed to the bottom, so the ones you see first are the ones that actually show something.