Data Sources & Methodology

Where LienWatch Data Comes From

A plain-language description of every data source we use, how often we refresh it, and what the AI layer actually does with it.


Live county delinquent property tax lists (Tax Liens & Tax Deeds)

We run scheduled crawls (cron jobs) of three Florida county tax collector sites — Hillsborough (hillstaxfl.gov), Miami-Dade (mdctaxcollector.gov), and Orange (octaxcol.com) — plus a broader national scraper layer spanning sheriff-sale pages, county tax-deed portals, and state-level deed registries across 14+ states to surface upcoming auctions.

Refresh cadence:

Source list and per-source configuration are documented internally in our docs/scraper-architecture.md. Some HTML county pages change layout without notice — when a source breaks, that county's results fall back to the seed dataset or return zero rows until the next successful scrape.

AIEVR nationwide valuation reports (every US address)

AIEVR generates on-demand property valuation reports for any US address. The pipeline layers four data sources together:

LienWatch does not operate a direct government feed for AIEVR valuation data. The AI narrative and valuations are best-effort web research anchored on listing-side sources, comparable sales selected by the AI within the requested address's market area, and public records surfaced through Google's grounding layer. Where Zillow returns a stub and Gemini's grounding times out, the report renders as “partial” with a cause-specific message rather than a fabricated value.

Auction calendar registry (30 tax-lien enabling-legislation states)

Where a state treasury office publishes an annual or bi-annual auction calendar but does not publish machine-readable data, our staff maintains a curated registry of auction timing (see scraper/tax-lien/sources/registry.js). This is human-curated calendar data, not real-time scraping.

How property scoring works

Every listed property is graded A through F by a rules-based scorer that weighs estimated ROI, LTV ratio, interest rate (for liens), and active-deals count. The scorer (scraper/scorer.js) is deterministic — the same inputs always produce the same grade.

Grades are advisory, not a recommendation; always do your own due diligence before bidding.

Property photos and street view

Property photos come from a tier waterfall — Realtor → Redfin → Zillow → County Assessor → Mapbox static tile → OpenStreetMap → Google Street View Static. When every tier fails, the property card shows a placeholder map tile so the address and listing metadata stay legible without a fabricated image.

We do not own or upload photos ourselves — we render what listing-side and map providers publish.

What we don't do

Not appraisals. AIEVR valuations are AI research, not licensed appraisals. They are not a substitute for a state-licensed appraisal when one is required for financing, legal, or tax purposes.

Honors robots.txt. We do not scrape any site that has explicitly asked us not to. Our national-deals crawler follows robots.txt directives and we pause a source the moment a publisher signals a stop.

No data selling. We do not sell user data. Account information stays in our billing system; usage logs stay with us.