Guide

What an AI Due-Diligence Check Actually Covers — and What It Doesn't

What an AI-run due-diligence check actually pulls, the named gaps in coverage and legal rules, and what to verify yourself before closing.

See how Titleman sources every number

A due-diligence run pulls the property, ownership, zoning and risk data our connectors reach — deep in the US, partial in Spain, Germany and France, thin or absent elsewhere — plus any documents you upload. It does not check title, confirm legal compliance, verify zoning entitlement, or apply local law. Use it as a first pass, not a closing file.

What "due diligence" means here, plainly

Our output is a written checklist and report in the room, built from the same data connectors used across the platform. It is not an appraisal, not a valuation, not a legal opinion, and not a title opinion — we do not produce any of those, and nothing below should be read as one. Where a figure is named, its source and date are in the table below; where we don't have a measured figure, we say so instead of guessing.

One honesty note before the checklist: the dedicated due-diligence job type is lightly used in production — two runs since 1 January 2026, two in the last 30 days. What follows describes the checks it draws on from the shared data pipeline, not a separately battle-tested feature.

What a run actually pulls and checks

  • Parcel and ownership records. Deep for the U.S.; a cadastral read (address, year built, floor area, use) for Spain, confirmed on Madrid and Barcelona; parcel data (ALKIS) for 14 of Germany's 16 federal states; cadastral geometry and point lookups in France.
  • Recorded deeds and transaction history. Present for 17 U.S. states. No country outside the U.S. has a transaction database behind our checks.
  • Zoning designation. A mapped zoning layer in 21 U.S. cities, plus development-plan data for Hamburg.
  • Federal risk, environmental and demographic layers. For U.S. coordinates only — flood, seismic, contamination-site proximity, employment density, HUD rent benchmarks.
  • Asset / property type. The run attempts to classify what kind of asset the subject is before proceeding.
  • Subject readiness. A completeness gate runs before analysis — most commonly, missing size or asset type stops it.
  • Comparable transactions and deals, where our comps capability can reach them — a starting list to work from, not a finished, verified set.
  • Whatever you upload. T-12s, rent rolls, offering materials, anything attached becomes part of what the run reads.
  • A written report structure, with sections built where data exists. Some sections routinely come back empty — see below.

What it does not check — named, not hedged

  • Title. No lien search, no chain-of-title check, no encumbrance search. This is not a title opinion and we do not produce one.
  • Legal compliance or entitlement. A zoning designation is a data pull from a public layer, not a determination that a use is permitted. That determination sits with the local planning department, not us.
  • Local jurisdiction rules — tenure, tax treatment, measurement convention. These are not applied to any run, in any country, today. The jurisdiction data never reaches the model regardless of the address typed.
  • On-site condition. No inspection, no walk-through, no Phase 1 environmental site assessment.
  • Rent roll and lease terms. Not derivable from public sources. Without your own documents, that section is empty.
  • Sale price, for Texas subjects. Texas does not disclose transaction prices, and no vendor we use fills that gap.
  • Full U.S. coverage. Nine states carry no parcel data at all, including South Carolina. Several large states are mapped by a handful of counties only — four in Georgia, two in Arizona, eleven of California's fifty-eight.
  • Any government register outside the U.S., Spain, Germany and France. Elsewhere the run works from open sources and whatever you supply — no cadastre, no land registry, no transaction register behind it. This includes markets we see meaningful demand from, such as the UAE and Russia/CIS, where no connector exists at all.
  • Advertised or listing-site data. The run works from registry, transaction and open data, not from listing feeds.
  • Litigation, permits, code violations or open enforcement actions. Not sourced by any connector we run.
  • Financing, existing debt or encumbrances. Not sourced unless you supply it.
  • A reliable "you're out of coverage" warning. The coverage self-check is wired into one product (not due diligence), and across every panel it has ever produced it has not once said "covered."
  • Multi-subject integrity. When a job covers more than one property in a single canvas, colliding facts can be silently dropped — first value written wins.

What to still do yourself before closing

  • Order an actual title search or commitment from a title company. Nothing here substitutes for one.
  • Confirm zoning and permitted use directly with the local planning or building department — not from a data layer alone.
  • Get local counsel for tenure, tax and any jurisdiction-specific legal question. No local rules are applied on any run, anywhere, today.
  • Commission a physical inspection, and a Phase 1 environmental assessment if the deal warrants one.
  • Have a licensed appraiser verify value if you need an appraisal. Our output is an opinion of value at most, never an appraisal.
  • Supply your own rent roll and lease abstracts rather than assuming the system located them.
  • If your subject is outside the United States, Spain, Germany or France, treat every figure returned as open-source research, not registry-verified fact, and check current coverage for your specific market before relying on anything in the report.

See how Titleman sources every number

Bring us a deal you already closed

We run it through Titleman and show your team the finished work next to their own.