The listing says two units. The assessor says one.

California (five county parcel registries; Los Angeles and San Francisco city permit feeds)

• California assesses under Proposition 13, so the value on a county's assessment roll is what the present owner paid indexed at no more than 2% a year, not an estimate of what the property is worth today. It resets to the price paid when the property changes hands, so what a buyer would owe is set by their own purchase price rather than by anything the roll holds against the present owner.
• California publishes no statewide property record: each of its 58 county assessors publishes on its own terms. This report reads the county sources named in its own sources list; where a county publishes only a search page for people, the parcel is not identified here at all and the county assessor's site is where to look next.
• Rent comparables are gathered from national listing portals, not from a California public record, and are asking rents rather than leases.
• In the California counties this report can read, it reads two things: which parcel the address is and, in the counties whose assessor publishes values, what the assessment roll holds against it — two of the five publish a parcel number and no valuation at all, so a missing assessment here can mean a county that publishes none rather than a search that came back empty. Alameda, Los Angeles and San Francisco publish one transfer or recording-date field on a county roll snapshot, without a price or earlier chain; a roll snapshot may omit transfers recorded after it closed. San Diego and Sacramento publish no transfer field. City construction records are read only for Los Angeles and San Francisco; their absence elsewhere below is not evidence that no approval was required. State landlord rules are read for every covered California address. A local layer is added only for the municipalities named in this adapter's curated table; elsewhere, the absence of local rules means this service has not curated that city, not that the city imposes none. The local housing department and county recorder are where to settle those remaining gaps before an offer.

Connecticut (OPM statewide parcel and CAMA layer)

• Connecticut OPM supplies parcel identity and assessment rows in all 169 municipalities. Collection years differ: Lebanon retains 2023 CAMA data, while the other 168 towns carry 2026; geographic coverage is not a claim that every assessment is current.
• The source's derived address fields are not trusted: resolution uses validated situs fields, town, postcode when supplied, and refuses conflicting or duplicate records.
• OPM FY2027 municipal real-property rates are retained separately from district rows and flat charges; a buyer-price figure is withheld when the report year is newer, the CAMA collection year is missing or does not match the retained grand list, a district is unresolved, or the parcel does not prove ordinary 70% residential assessment.
• OPM sales are annual grand-list reports rather than a live deed index or complete title chain; individual municipality-years may be absent after revaluation, and non-use or unknown status never becomes a confirmed arm's-length comparable.
• Current statewide landlord rules apply in all 169 towns; local fair-rent adoption is attached only for 39 proven commissions, and the first detailed licensing or occupancy-certificate layers cover Hartford, New Haven, Bridgeport, and Stamford.
• Hartford municipal building permits are checked across the official 2011-2014, 2015-2019, and 2020-current feeds; the other 168 towns, including discovery-only Stamford, are not claimed.
• HUD FY2026 Fair Market Rents cover all 169 legal towns through HUD's town-level New England geography as a broad voucher-program cross-check; they are not live asking-rent comparables.
• Owner and mailing fields in the public CAMA source are neither read nor retained by Listingprobe.
• Rent comparables continue to come from national listing portals and are asking rents rather than signed leases.

Florida (statewide DOR cadastral assessment roll)

• Florida parcel and assessment facts come only from the official DOR 2025 statewide cadastral FeatureServer and retain that assessment year while the checked-in 2026 roll files await a query-layer rollover.
• Address resolution is exact, locality-qualified, bounded, and deterministic; a missing or conflicting locality, duplicate account, truncated page, malformed row, or ambiguous unit is refused.
• Miami-Dade and St. Johns condominium rows that depend on DOR's separate related tables remain unavailable until that join is independently verified; the adapter never borrows a base polygon for a unit.
• Owner, fiduciary, and mailing columns are neither requested nor parsed. The statewide source does not publish bedrooms or bathrooms, so none are inferred.
• Florida tax context requires an exact county + TAX_AUTH_C adopted chain. It keeps school and non-school bases separate, excludes non-ad-valorem charges, and withholds a buyer bill because asking or sale price is not the appraiser's future just-value determination.
• Sale context is limited to up to two DOR-published NAL segments from roll 2025 and is never described as complete sale, deed, or title history; the 2026 preliminary SDF release is tracked separately during rollover.
• Permit automation is limited to exact City of Miami folios beginning 01 and unincorporated Miami-Dade folios beginning 30. County and city feeds are never borrowed across an authority boundary; unsupported and unresolved paths remain explicit.
• Miami-Dade County's automated GIS permit window is only the last three years, while the City of Miami feed begins in 2014. Empty results are limited-window no-matches, not proof that no permit exists; certificate coverage remains unconfirmed.
• The current 2026 Florida Residential Landlord and Tenant Act applies statewide. Local law is added only for five exact DOR PointMatch jurisdictions; a postal city, unknown place, stale source, or failed exact-address lookup cannot select a local rule.
• HUD FY2026 Fair Market Rents cover all 67 counties as a gross-rent voucher-program reference; they are not live asking-rent comparables or subject-property estimates.

Maryland (statewide SDAT assessor accounts)

• Maryland SDAT and the Department of Planning publish one public-domain monthly table covering all 23 counties and Baltimore City; each report cites the portal snapshot read.
• The statewide table intentionally hides protected owner names, and Listingprobe neither reconstructs nor supplements them.
• Maryland tax context uses the current statewide SDAT fiscal-year rate table and exact county/town/class mapping; current-owner Homestead Credit fields are disclosed but never transferred to a buyer projection.
• Maryland exposes up to three SDAT-published transfer segments for the exact resolved account; No Data, non-arm's-length, multiple-account, and malformed segments never become comparable-sale claims.
• Maryland construction records are automated only for Baltimore City, Montgomery County, and Prince George's County; the official 180-authority denominator remains visible and unsupported incorporated places fail closed.
• Permit final dates are not treated as certificates of occupancy, and an empty supported-source query means no matching record was returned rather than no permit exists.
• Maryland local landlord-law layers are curated only for Montgomery County, Prince George's County, Takoma Park, and Baltimore City; other places receive the current statewide layer without guessed local claims.
• HUD FY2026 Fair Market Rents cover all 24 county-equivalents as a broad voucher-program cross-check; they are not live asking-rent comparables or subject-property estimates.
• Rent comparables come from national listing portals rather than a Maryland public record and are asking rents rather than signed leases.

Massachusetts (MassGIS statewide assessor parcels)

• MassGIS standardizes the parcel and valuation extract each city and town supplies, and the fiscal year can differ by municipality; the source beside each value states the year actually read.
• MassGIS' annual FY2023 LA3 report supplies assessor-verified sales for 324 municipalities; it is a past-fiscal-year view rather than live registry/title history, so recent closings may be absent.
• Municipal construction records are read only for Boston (2015 onward), Cambridge (applications from November 2017), and Worcester (2015 onward); every other municipality is a known non-attempt.
• None of those three municipal files establishes certificate-of-occupancy coverage; a missing certificate remains an unverified item.
• The forward tax estimate uses DLS FY2027 class rates only when the MassGIS assessment is from the same fiscal year and its use code is unambiguously Class One residential.
• Local landlord rules are included only for the five municipalities with a primary-sourced curated layer; every other city or town is labelled Massachusetts statewide-only.
• Rent comparables come from national listing portals rather than a Massachusetts public record and are asking rents rather than signed leases.
• HUD FY2026 Fair Market Rents cover all 351 municipalities as a broad voucher-program cross-check; they are not live asking-rent comparables.

New Jersey (NJParcels assessor + DCA permit database)

• Rent comparables are gathered from national listing portals, not from a New Jersey public record, and are estimates rather than leases.
• Equalization ratios and effective tax rates are generated from the Division of Taxation's published tables and checked against the certified PDFs, and carry a citation rather than a live retrieval.
• Equalization ratios and effective tax rates are on file for all 564 New Jersey municipalities, and curated municipal ordinances for 20 municipalities in Atlantic, Camden, Essex, Hudson, Mercer, Middlesex, Passaic and Union Counties; elsewhere in New Jersey a report carries the statewide rules and no researched local ordinance.

New York City (GeoSearch + PLUTO + DOF assessment roll + ACRIS + DOB permits)

• New York City does not reassess a property when it is sold, so the tax figures in this report are what the seller pays now rather than a projection off the purchase price. What can move the charge is the assessment phase-in cap, not the sale.
• Rent comparables are gathered from national listing portals, not from a New York public record, and are asking rents rather than leases.
• New York required no occupancy certificate for buildings erected before 1938, and the city's published file of them only opens in 2012, so for an older building this report draws no conclusion from the absence of one either way. Order it from the Department of Buildings for the parcel if how many homes the building legally contains matters to the deal.
• The annual property tax figure is arithmetic over two Department of Finance publications — its billable assessed value and its published class rate — because New York issues no per-parcel statement as open data, and the class rate available may be a fiscal year or two older than the assessment roll. Confirm the figure against the seller's own statement before underwriting on it.

New York State (ORPTS statewide assessment roll)

• New York's statewide assessment roll publishes the property class, the assessed value and the state's own full market value, but no living area, no year built and no count of dwellings — so nothing in the public record here contradicts or confirms what the listing says the building contains.
• The assessed value on the roll is a fraction of market value set by each municipality, and this report divides the two figures the roll publishes to state that fraction rather than quoting the rate the state publishes separately.
• Recorded transfers are read only for Rochester, where the official parcel layer publishes one current snapshot rather than a complete history; the statewide ten-year portal requires an interactive security challenge and is left for a person to consult.
• The forward property-charge estimate uses ORPTS's prior-cycle full-value county, city or town, and school components only when the parcel's exact assessing-unit and school codes match and the roll carries no H/N/S class-tax marker. It is withheld for class-tax and village or other subsection parcels because the state table publishes only a blend or omits an additional local levy; special-district charges and seller exemptions still require the local collector's current statement.
• For an unambiguous Nassau print key, the official LandRecord page adds the current and prior school, county, town, special-district and direct-assessment rows. The report keeps the actual charge separate from seller exemption savings that may not transfer; unsupported parcel-number shapes remain unavailable.
• Construction records come from separate city services only in Buffalo, Rochester and Syracuse. Every other municipality is unsupported rather than treated as having no filings; Syracuse itself warns that its published data is not updating correctly.
• Every report quotes the statewide HSTPA rules. Current HCR Good Cause and ETPA opt-ins and researched registration regimes are added only when the municipality can be pinned; an unsupported municipality is not treated as rule-free.
• Rent comparables for a New York property outside the city come from national listing portals rather than from any public record, and are asking rents rather than signed leases.

North Carolina (NC OneMap parcel and assessment spine)

• NC OneMap physically aggregates all 100 county producers plus the Eastern Band of Cherokee Indians; the reviewed 2026-09-06 snapshot has 68 counties within the 120-day transform-observation contract. The 32 overdue county observations remain explicit, time-bounded upstream gaps and are refused rather than presented as current.
• OneMap parcel-value semantics vary by producer. Listingprobe retains the source's verbatim value type and never turns every county's value into a universal market-value claim.
• Unit-qualified rows must match the requested unit exactly. A unit-less address over separately assessed units and a requested unit that only finds a parent lot are both refused.
• Owner, mailing, legal-description, source-document and subsurface fields in OneMap are never queried or retained.
• NC OneMap's last-sale date is shown only as date provenance: the layer publishes no consideration or transfer qualification, so it never becomes a comparable sale and statewide sale-history capability remains unavailable.
• Current Chapter 42 landlord duties, remedies, deposit rules, notice safeguards, and G.S. 42-14.1 local-law preemption apply statewide. Local ordinance coverage is 0 of 550 official municipalities; a postal city, ETJ, military, tribal, ambiguous, or withheld boundary row never invents local law.
• North Carolina permit history is available only for official boundary-proven City of Raleigh parcels, from the city feed reviewed from 2000-02-22 through 2026-09-11. The other 549 municipalities and every county/unincorporated path remain unsupported.
• HUD FY2026 Fair Market Rent is available as an informational 40th-percentile gross-rent reference for all 100 exact parcel-established counties. It is neither a live listing comparable nor evidence of achievable rent, and metro labels never resolve geography.
• NCDOR's statewide rate is a county levy only. Listingprobe labels price arithmetic as an estimated county levy and never substitutes it for the assessor's parcel bill or a complete buyer tax total.

Wisconsin (V12 parcel, assessment and DOR tax context)

• Wisconsin V12 parcel and 2025 assessment coverage reaches all 72 counties, subject to exact situs, locality, unit, source-manifest, and state-qualified identity validation.
• Total assessed, land, improvement, managed-forest-land, and estimated fair-market values retain their distinct source meanings; missing values are not filled and contradictory arithmetic is refused.
• Owner and mailing fields are never queried or retained.
• Wisconsin DOR offers a redacted RETR sales search and historical downloads in My Tax Account, but the browser portal has no documented unauthenticated parcel API, pagination limit, or refresh-consistency contract. Parcel-level sales history therefore remains unavailable; grantor, grantee, contact, and financing data are never retrieved or retained.
• Current Wisconsin landlord duties, remedies, and local-law preemption apply statewide. The generated registry contains 1933 local authorities, but local ordinance coverage remains 0; unmatched, stale, tribal, or otherwise ambiguous geography never invents a local rule.
• Permit coverage is local only: 1 of 1933 authority paths. Green Bay's source contains issued short-term-rental permits only, renews annually, and never confirms absence of construction permits or certificates of occupancy.
• HUD FY2026 Fair Market Rent is available as an informational 40th-percentile gross-rent reference for all 72 exact parcel-established counties. It is neither a live listing comparable nor evidence of achievable rent, and metro labels never resolve geography.
• Wisconsin V12 supplies the parcel's reported net/gross tax, while DOR's municipal effective full-value rate supplies a separately labelled buyer-price estimate. The municipal rate is an average and can differ across school or special-purpose district boundaries.