U.S. roadside attractions · source audit and routing feasibility · Keenable SELECT

Only 10 of 51 Roadside America state pages yielded parseable attractions, so no all-stop route can honestly be drawn yet

On 2026-08-28 the map pages for all 50 states plus Washington, DC on Roadside America — one concrete operational definition of “roadside attraction” — were fetched and parsed. 51 jurisdiction pages were scanned; only 10 produced extracted rows, for 831 rows in total. Most pages were dynamically rendered, so many addresses arrived incomplete or merged, and no reliable latitude/longitude dataset resulted.

This is a partial source audit, not an exhaustive inventory or an optimized itinerary. There is no authoritative national registry of roadside attractions, and the source warns that attraction status, hours, and prices change without notice. A shortest visiting path was not computed — and an exact route is undefined anyway without a start/end point, a rule for Alaska and Hawaii, current road-network travel times, and a choice between an open path and a round trip.

What each state page actually gave up

Rows extracted per jurisdiction page, 2026-08-28 · hover a bar for address and city coverage

rows were extracted from the pagepage was scanned but nothing parseable came back

What the chart cannot say alone

New York (482 rows) and Massachusetts (296) alone account for 778 of the 831 extracted rows — the “inventory” is really two states plus fragments. roadsideamerica.com · roadsideamerica.com

Seven states — Arkansas, Hawaii, New Hampshire, Nevada, Rhode Island, Utah and Vermont — each collapsed into a single row whose “address” is the whole page’s text merged together, a signature of dynamically rendered maps the fetcher could not unpack. roadsideamerica.com

Street-address fields are nearly full (827 of 831 rows carry one) but the city column is empty in every single state — 0 rows with a separate city anywhere — so no row can be geocoded with confidence. roadsideamerica.com

Delaware is the only small state that parsed cleanly, 46 rows with 46 addresses — proof the pipeline works when the page exposes plain markup. roadsideamerica.com

Method and limitations

Source: Roadside America state map pages, used as one operational definition of “roadside attraction”, fetched 2026-08-28.
51 jurisdiction pages scanned (50 states + DC); 10 yielded parseable entries; 41 yielded nothing usable.
831 raw rows extracted; blanks are preserved as blanks, nothing was filled in.
No coordinates were recovered, so no distance matrix and no path were computed; any “shortest route” claim from this data would be invented.
Attraction status, hours and prices change without notice per the source itself.

Next step: how a real route gets built

Select a licensed or downloadable national POI dataset that ships with coordinates.
Deduplicate it against itself and against these rows.
Define the start/end point, the open-path vs round-trip choice, and mainland/ferry/flight rules for Alaska and Hawaii.
Build a road-time matrix from current road-network travel times.
Solve an approximate asymmetric traveling-salesperson path and publish the ordered coordinates on an interactive map.

The 831 rows, as extracted

Search by state, attraction name or address · blanks stay blank

State	Attraction	Address as extracted	Source

Data: 831 raw rows extracted from Roadside America state map pages (roadsideamerica.com/map/<state>) fetched 2026-08-28, plus a per-state extraction summary (rows extracted, rows with an address, rows with a city, per jurisdiction). 51 jurisdiction pages scanned; 10 produced rows. A bar measures rows extracted from that page, not attractions verified to exist. For space, the seven single-row states’ merged page-text “addresses” are truncated to their first line in the table and marked as merged; the table shows the raw rows delivered with this audit, and per-state totals come from the extraction summary. No coordinates, no route: this report deliberately claims neither.