[Places][2026-07-22.0] About 40% of websites values for Türkiye do not respond; a reachability signal would help

Summary

Of the Türkiye places that carry a websites value, roughly two in five do not answer an HTTP request. There is nothing in the record that lets a consumer tell a listed website from a working one, so any adoption or digital-presence figure built on the field overstates by about a third. Adding a last-checked timestamp or a reachability flag to the websites entries, or dropping values that have failed for several releases, would fix that at the source.

Release and scope

• Release: 2026-07-22.0, theme places, type place, Türkiye (bbox 25.5..45.0, 35.7..42.3, addresses[1].country = 'TR', names.primary IS NOT NULL, operating_status not closed)
• 1,786,700 records in scope, 684,496 with a non-empty websites value (38.3%)

What we measured

We requested the first websites value of every one of those 684,496 records once, in September 2026, from a host in Türkiye:

• GET on the URL as listed (https:// prefixed if no scheme), redirects followed (max 5), 12 s timeout, identifiable user agent, headers read only, one request per site, no retries.
• Counted as reachable: a 2xx response.
• Counted as unreachable: connection or DNS failure, timeout, or a non-2xx status.

Result: 413,777 reachable (60.5%), 270,719 unreachable (39.5%). So the share of Turkish places with a working website is 23.2%, not the 38.3% the field suggests.

Caveats: a single attempt on a single day will misclassify some sites that were briefly down or that block non-browser clients, so the true unreachable share is somewhat lower than 39.5%. It is not close to zero: the pattern by province and sector is stable and the bulk are parked domains, expired hosting and DNS that no longer resolves.

Breakdown

Per-province and per-sector counts (listed vs reachable) are published as CSV/JSON under CC BY 4.0, with the SQL that produced them:

• https://pinlyx.com/research/turkey-business-digital-report
• DOI: https://doi.org/10.5281/zenodo.22217770

We can share the per-record list (GERS id, URL as listed, failure class) if it is useful for the Places pipeline; it is ~270k rows.

Suggestion

Any of these would let consumers stop re-crawling:

1. a last_verified (or similar) timestamp per websites entry,
2. a boolean or status alongside the URL when the pipeline already checks liveness,
3. or removal of URLs that have failed across consecutive releases.

Related: #570 (the addresses[].region field for the same extract).

Start by locating the Places pipeline and schema definition for the websites field, then inspect how release 2026-07-22.0 produces it. Compare the proposed reachability representations against the reported Turkish cases and determine which the project can support; done means consumers can distinguish listed from reachable websites without re-crawling.