Background
Battery-Electric Vehicles (BEVs or EVs for short) are here on our planet and proliferating rapidly. Like your cell phone and laptop computer, EVs use rechargeable batteries, only much, much larger. People want to travel in their EVs and because distances are more than can be done on a single charge, charging stations are needed on the road networks. Since there is not yet a charging station on every corner like with gas/petrol stations, EV drivers need to know where charging stations are so they can plan their trip. Charging station operators (aka charge point operators or CPOs) currently maintain their own maps of the locations and details of their stations, and many EVs connect to various CPOs and display stations on an in-car map. Advanced EVs will also calculate a trip plan that includes charging stops and estimated stop durations, and update the estimates on the fly.
How is OSM Doing with Mapping of EV Charging Stations?
EV charging infrastructure data on OSM is currently not sufficiently trustworthy to be used by EV drivers, at least in the United States. I can say that because I’ve been driving an EV in the US for over 4 years now. There are tag inconsistencies and data inaccuracies of various kinds including wrong or incomplete station metadata and wrong GPS coordinates. This might be fixable if enough people care about it.
What Happened on 22 June 2025?
Last summer on 2026-06-22 a bulk import of Tesla Supercharger station data in the United States was done. It attempted to import data from supercharge.info, which tracks Tesla Supercharger stations worldwide and uses OSM as a basemap (but does not write to OSM). It was planned and executed with insufficient review, so there were many errors of various kinds: wrong GPS Lat/Lon, duplicated stations, injection of previously-deleted tags and data that didn’t even come from supercharge.info, overwrites of good tag data, L2 Tesla Wall chargers tagged as Tesla Superchargers, etc. Read the saga at Bulk import of Tesla Superchargers in the United States.
The importer (OSM user @GA_Kevin) did some partial fixes with help from me and others, but he ultimately became upset at the criticism of all the errors and left the forum. It was recommended that no further imports be allowed by @GA_Kevin until this import is fixed up. Here we are a year later and many fixes have been done, but many errors remain. After all, it’s a continent-wide chunk of data with over 3200 stations now open in the US alone (and new ones added almost every week).
What Has Happened Since Then?
Since then @GA_Kevin has done more automated edits, not discussed here as required by the Automated Edits Code of Conduct and not under the @GA_Kevin_Import username. A recent changeset added a ref:afdc tag to 1508 EV stations and not just Tesla Superchargers. When asked about it on the changeset discussion, first by me and then by at least 2 members of the DWG, he claims it’s not an automated edit.
The ref:afdc tag by itself it not as concerning as the discovery of an automated tool by @GA_Kevin that is regularly adding charging station data to OSM. It imports/scrapes data from several websites, all of which have data quality issues discussed at length in the Bulk Import thread above. Then it either directly uploads it to OSM or waits until someone hits the “Upload Now” button. What is that tool? US Charging Board 1.0.
What is US Charging Board 1.0?
The project is on github. It runs hourly. Last summer @GA_Kevin stated that ±100 m (± 328 ft) position accuracy was good enough for him. That met serious opposition. Now the positional error seems to be getting worse. Consider the new Tesla Supercharger in Kissimmee, Florida. On TMC we only had 1 incomplete street-level photo. Supercharge.info had it mapped at this location:
As soon as the station went live, supercharge.info marked it live, and an OSM node shortly appeared at that exact Lat/Lon, placed by US Charging Board 1.0:
The source imagery is listed as Esri World Data but that’s impossible because Esri World Imagery is not yet up-to-date enough to show the things that are getting mapped. Even the Osceola County Orthoimagery 2026 was not quite new enough:
From street-level imagery in the thread on TMC, I was able to see that the station was about 400 feet to the south, closer to the lake. What I didn’t know were 2 things: 1) were there 1 or 2 pull-through “trailer-friendly” stalls and 2) how close to the lake did the other stalls extend? Both questions were answered by the Tesla smartphone app, which showed a tiny aerial image, most probably from Google Maps, which of course I did not copy. With that I could map the station as a 2-part multipolygon relation and merge in the bogus node to save the history. The distance from the bogus node to the centroid of the 2-part relation is about 390 ft:
What’s the latest?
More recently a new 28-stall Tesla Supercharger station in Escondido, CA went live, but before that I saw from street-level photos that the pin on supercharge.info was about 950 ft south on the wrong side of the mall.
I waited to see if US Charging Board 1.0 would drop an OSM node there and, sure enough, it did. Before I could write up all this, someone on TMC noticed the error. A supercharge.info editor saw it, then saw the bogus node by @GA_Kevin and flagged me. I have left the node at the bogus location so people can see [edit 2026-09-04 this has been corrected now].
What is US Charging Board Doing?
US Charging Board 1.0 is doing 1 of 2 things:
- automatically grabbing data from supercharge.info and inserting a node on OSM
- automatically grabbing data from supercharge.info, then waiting until @GA_Kevin hits the button to “Upload Now”
It doesn’t matter if it’s 1) or 2) because the result is the same: erroneous data is being added to OSM without review. In this most recent case it’s erroneous location data with a 950-ft+ error. This is because the GPS Lat/Lon coordinates were copied verbatim from a 3rd party without review. In both the previously-mentioned cases the 3rd party was supercharge.info which, as explained above, does not actually write to OSM but uses it as a basemap. Because OSM is only a basemap, they do not have to be so careful about position, so very often (maybe 40% of the time) the location coordinates are wrong (and as seen here can be very wrong). Later, as people people visit on the ground and take street-level photos, we can work out the correct position. Sometimes that never happens and we have to wait until aerial imagery appears for us to use in OSM. If anyone is interested, I can make a list of the times this has happened. The surprising thing is that it’s not only out in the middle of the desert. It is often in big cities like LA.
The use of the tool itself is not the issue. If it helps us add charging station data to OSM, that would be a good thing. The issue is that the tool is being misused as a regular, schedule importer of often-wrong data from 3rd parties that is then added to OSM without sufficient review. The very same issues of data review have been discussed with ATP data importation.
What Should We Do Now?
The use of erroneous data without review is what should be immediately stopped, and in this case an automated tool is accelerating the addition of erroneous data, which then makes it even more work to correct. As mentioned several times, the cleanup of the erroneous data from the 2025-06-22 import is still ongoing and is not complete. If anyone is interested, I can give a list of POIs.
