Google Maps Traffic API — Live Accidents, Closures & Incidents

Live road incidents from Google Maps for any map area you define with a bounding box: accidents, construction, closures and other slowdowns, each geo-coded and returned as flat JSON.

What you get

Operation	Returns	Billed as	Notes
trafficAlerts	Traffic Alerts by Area	1 area query	every live traffic incident in a map area

One area query returns as many incidents as the area currently has. Live test runs came back with 4 to 30 incidents each: 30 across New York at detail level 13, 11 across London, 4 in downtown Los Angeles. You are billed for the query, not for the incidents in it, so a busy city and a quiet suburb cost the same.

Coverage

Global. Verified live against the United States (New York, Los Angeles, Chicago), the United Kingdom (London) and Brazil (São Paulo). Coverage density follows Google Maps itself, so it is dense in metros and sparse on rural roads.

Use cases

Fleet & Dispatch Routing

Poll the corridors your vehicles run and reroute around accidents and closures before drivers reach them. Pair it with a Waze feed to catch accidents neither source reports on its own.

Incident Monitoring Dashboards

Filter to accidents and closures only and drive a live map for dispatchers or a newsroom. Every row carries a Google Maps deep link so an operator can open the scene in one click.

Traffic & Mobility Analytics

Run the same bounding box on a schedule to build a time series of incident density by area, and compare construction and closure load across neighbourhoods or cities.

Insurance & Risk Signals

Track accident density in defined zones over time as an underwriting and claims-triage signal, with each row geo-coded to a coordinate pair.

Navigation Product Enrichment

Layer Google Maps incident points onto a routing or mapping product as a second opinion alongside whatever primary incident source it already consumes.

FAQ

Does this return the same incidents as a Waze feed?

Largely not, which is the point. In a Los Angeles test both sources reported four accidents at the same moment and none of them matched: the nearest Waze counterpart to each Google accident was between 3 and 16 km away. If you need to catch as many incidents as possible, run both.

Can I get the time an incident was reported?

No. Google returns only an incident type and a coordinate pair. Each row carries a scrapedAt stamp recording when this actor fetched it, which is not the same as when the incident began. If you need report times, descriptions and reliability scores, use a Waze-sourced feed.

Why did my city-sized bounding box get split into several queries?

Google caps the area per query by detail level: 500 km² at level 13, and less at higher levels. Anything larger is rejected outright upstream, so this actor splits the area into tiles, queries each, and merges the results. Each tile counts as one query for billing, and the run log states the tile count before any charge.

How do I keep costs predictable when polling continuously?

Set Max area queries per run to the tile count you expect. The run stops there and the report states how much of the area it covered, so a mistyped bounding box cannot quietly multiply your bill.

Can I track the same incident across repeated runs?

Approximately. Google publishes no incident ID, so each row carries a derived incidentId fingerprinted from its type and position. It held stable across about 90% of rows in back-to-back polls; Google occasionally nudges a coordinate by a few metres, which changes the fingerprint. Treat it as a best-effort correlation key, not a durable identifier.

Output

Every row is flat JSON, exportable as CSV, JSON, Excel or via the API. A run report is saved to the key-value store with the numbers, any failed items and what they cost.

Google Maps is a trademark of Google LLC. This actor is not affiliated with, endorsed by, or sponsored by Google.