Coverage and data health

requires agreementRealAuction robots.txt is `Disallow: /`. Florida tax deed lists are public records available directly from each county Clerk of Court — ingest those instead. Use RealAuction only under a data agreement or the subscriber's own authenticated session.

requires agreementSame RealAuction platform and robots policy as realtaxdeed.com.

requires agreementliveauctions.govease.com/robots.txt is `User-agent: * / Disallow: /` — verified 2026-08-05. GovEase is the widest-reach platform in the researched set (59 counties across AL, CA, IA, IN, MO, MS, OK, PA, TN, TX, WA). The data itself is public — no login — but this entry stays `requires_agreement` so nothing in the ingest path can crawl it by default. The `govease` adapter takes the browser-capture route instead: pages are opened in the operator's own browser and read from data/govease-capture/, and the adapter imports node:fs rather than politeFetch, so there is no code path from it to this host. See docs/govease-capture.md.

requires agreementTHE OPERATOR HOLDS AN AGREEMENT WITH THIS HOST, confirmed 2026-09-11, and `lib/ingest/adapters/bid4assets.ts` is the only caller that passes `underAgreement`. The decision stays `requires_agreement` rather than becoming `allow` because that is what it is: permitted to us, on the strength of a specific agreement, and not to anyone else reading this file. Same stance as RealAuction for everybody without one. 40 counties across CA, FL, LA, NV, PA, WA. On the original basis, and read this before concluding the host is open: robots.txt now returns 404 rather than the 403 recorded on 2026-08-05, so there is no robots file and the obvious reading is that crawling is permitted. It is not. The edge control moved rather than lifted — /storefront/taxsales answers 200 to a browser user agent and 403 to ours, verified 2026-09-11, so the block is on who is asking. Getting past it would mean sending a user agent that is not ours. An agreement is a reason to be given access, not a reason to take it by looking like someone else, which is why the adapter reads browser captures and never fetches.

requires agreementSchneider Geospatial's qPublic/Beacon — the assessor portal for hundreds of counties across GA, AL, SC, FL, MS and TN, and the single largest free source of parcel values that is not a GIS layer. Two independent reasons not to crawl it, verified 2026-08-14. Its robots.txt allows `*` but names ClaudeBot and a dozen other automated agents with `Disallow: /`, and carries `Content-Signal: search=yes, ai-train=no, use=reference` — a posture that permits indexing and refuses automated collection. Separately, an ordinary request to an Application.aspx page returns HTTP 403, so there is an active edge control as well as a stated one. Either alone would settle it; together they leave no reading where a crawler is welcome. The data is a public record at source, so the route is the county assessor directly, or Schneider's own terms.

requires agreementSan Juan County NM. Its assessor's document library was the one free bulk tax roll found for either state — a 2026 tax roll published as CSV — and the host returns HTTP 403 to an ordinary request, including for /robots.txt, verified 2026-08-14. The crawl posture cannot be read without defeating the control, so the same stance as bid4assets and DataScout applies. The file is a public record: ask the assessor for it, or fetch it from a browser session, rather than working around the block.

requires agreementServes an Azure WAF JavaScript challenge on every request including /robots.txt, which returns HTTP 403 — verified 2026-08-10. The crawl posture cannot be read without executing the challenge, i.e. without defeating an active edge control, so the same stance as bid4assets applies. DataScout resells county assessor data that is a public record at source; where a specific county's parcel data is needed, go to that county's assessor instead. A subscription here comes with terms, and those terms are the route in.

requires agreementBehind DataDome bot detection — /robots.txt returns HTTP 403 with a captcha-delivery challenge, verified 2026-08-10. Crawl posture unreadable without defeating the control. Propwire is a free front end over a commercial property database; its value here would be comps and AVM, which is the job ATTOM already does under an API key and a contract. Use an account and a documented API, not the web app.

requires agreementThe statutory source of record for Georgia tax sales — O.C.G.A. § 9-13-140 requires the ad to run in the county legal organ, and the Georgia Press Association aggregates all 159 counties here with the full parcel list in the body. robots.txt reads `Allow: /`, which is misleading: every notice detail page is behind a click-through agreement stating the reader 'may not use the content of this site in any database, compilation, archive or cache' and 'may not engage in ... screen scraping, database scraping, or spidering', plus a Cloudflare challenge — verified 2026-08-12. Ingesting a notice is precisely the prohibited use, so robots is not the governing document. This is the single highest-leverage Georgia source and the correct path to it is a licence from the GPA, not a crawler.