What an Automated Title Search Misses
Automation reads the recorded index fast and consistently. It cannot see unrecorded interests, offline records, or resolve name ambiguity.
Quick answer
An automated title search retrieves and matches what is recorded, indexed, and machine-readable. It misses unrecorded interests such as possession rights, short leases, and pre-filing mechanic’s lien windows; records held only on paper or microfilm; and anything hidden by name ambiguity. No matching threshold removes both false positives and false negatives at once.
What is an automated title search?
The pipeline is mechanical. Resolve an address to a parcel identifier. Query the recording office index and the municipal datasets against that identifier and against every party name in the chain. Match related documents to each other, mortgage to satisfaction, lien to release. Report what survives the matching.
The boundary that matters is not automation against human abstraction. A human working the same indexes hits several of the same walls, more slowly. The boundary is between what the public record holds in machine-retrievable form and everything else that can still bind a buyer.
Everything else falls into four buckets: interests never recorded at all, records that exist but are not queryable, records that are queryable but cannot be attributed to your party with confidence, and facts that live in the physical world.
Definition
Automated title search — An automated title search is a software-driven retrieval of recorded instruments and public datasets for one parcel, matched and assembled into a report with no human abstractor pulling each record by hand. It reads indexes and structured fields. It does not read land, people, or paper.
What does automation retrieve, and what does it not?
Sorting by failure mode beats sorting by document type, because the fix differs. A never-recorded interest needs an inspection. An offline record needs a clerk visit. An unattributable record needs a person to read it.
| Item | In the public record? | Why automation misses it |
|---|---|---|
| Deeds, mortgages, assignments, satisfactions | Yes, indexed by parcel | Not missed. The core competence, and software does it faster and more consistently than a person. |
| Docketed money judgments | Yes, indexed by debtor name | Retrieved, but attaching one to your owner is a judgment call, not a lookup. CPLR 5018(c) dockets under the surname of the judgment debtor, with no parcel reference. |
| Rights of a party in possession | No | Nothing is filed. Occupancy is a physical fact, established by looking or by an estoppel certificate. |
| Leases of three years or less | No, and not required to be | RPL § 290(3) excludes a lease of three years or less from the definition of a conveyance, so it never enters the index. |
| Mechanic’s lien for finished work not yet filed | Not yet | Lien Law § 10 leaves the filing window open for months after the work ends. The record is accurate today and wrong next month. |
| Encroachments, and easements by prescription or necessity | No | Created by conduct or by physical fact, then declared by a court. A metes-and-bounds description does not say where the garage stands. |
| Instruments predating the system’s coverage floor | Yes, on paper or microfilm | Held offline at the recording office. No endpoint reaches them, and no query warns you they exist. |
| Terms inside the four corners of a document | Yes, inside the image | The searchable layer is typed index metadata. Covenants, easement descriptions and conditions live in the scanned body of the instrument. |
Which interests bind a property without ever being recorded?
Real Property Law § 291 makes recording the system of priority: an unrecorded conveyance is void against a later purchaser in good faith and for value whose own instrument is first duly recorded. The exceptions are the whole problem. An interest can survive while appearing in no index anywhere.
- Parties in possession. New York has held since Phelan v. Brady, 119 N.Y. 587 (1890) that actual possession by someone other than the record owner puts a purchaser on notice and creates a duty to inquire into that occupant’s rights. The occupant may be a tenant, a contract vendee, or a grantee under an unrecorded deed.
- Short leases. RPL § 290(3) excludes a lease for a term not exceeding three years from the definition of a conveyance, so it need not be recorded to be good. A month-to-month tenant at half of market rent is invisible in the record and extremely visible in your first year of ownership.
- Mechanic’s lien windows. Under Lien Law § 10 a notice of lien may be filed during the work and within eight months after completion of the contract or final performance, four months for a single-family dwelling. A freshly renovated property can be lien-free in the record today and encumbered next month for work already finished.
- Ripening adverse possession. RPAPL 501 requires a claim of right, meaning a reasonable basis for the belief that the property belongs to the possessor, and CPLR 212(a) sets a ten-year period for an action to recover real property. Nothing is filed until someone sues, and both the elements and the period differ by state.
- Prescriptive and implied easements. A driveway used by the neighbour for thirty years, a utility line, a landlocked parcel’s right of way by necessity. Established by conduct and declared by a court, usually after the dispute has cost someone money.
- Unrecorded mortgage assignments. An assignment is effective between assignor and assignee without being recorded, so the mortgagee named in the record is frequently not the party holding the note today. The lien is real and correctly reported; the payoff contact is a phone call, not a query.
Why do records that definitely exist still not show up?
Every online recording system starts somewhere, and the start date is a hard floor. In New York City, ACRIS covers Manhattan, the Bronx, Brooklyn and Queens from 1966 forward; earlier instruments sit in bound ledgers and on microfilm, retrieved in person or by order. Staten Island land records are held by the Richmond County Clerk on a separate system. Whether that matters depends on whether the defect falls inside or outside the window, which nobody knows until someone looks.
The second problem catches more people. The index is not the document. ACRIS returns typed metadata: document type, party names, dates, borough, block and lot. The instrument itself is a scanned image. A restrictive covenant, the metes and bounds of an easement, a right of first refusal, or a reverter lives in the body of that image. There is no field to filter on, and optical character recognition over decades of typewritten and handwritten conveyances is not reliable enough to treat as a search index.
Third, one jurisdiction can run two independent record systems that most people conflate. New York City is the clean example: deeds and mortgages go to the City Register through ACRIS, while notices of pendency and judgment dockets go to the County Clerk. A search that stops at the first office returns a clean record on a property in active foreclosure.
Fourth, lag. There is an interval between execution, recording, and the moment an instrument becomes queryable, and a further interval before derived datasets refresh. The length varies by office and by week, and no field in any response announces it. This is why every search report, automated or manual, carries a currency date, and why lenders order a date-down immediately before closing.
Why can’t name matching be tuned to catch everything?
Judgments, federal tax liens, UCC filings, bankruptcy petitions and most court records are indexed by party name, not by parcel. CPLR 5018(c) directs the clerk to docket a judgment under the surname of the judgment debtor first named, with cross-references under every other debtor’s surname. Nothing in those systems ties a filing to a block and lot.
So the software must decide whether ROBERT J SMITH in the judgment docket is the ROBERT SMITH on the deed. Entities are worse. 123 MAIN STREET LLC, 123 MAIN ST. LLC and 123 MAIN STREET, L.L.C. may be one company or three, and one principal may run forty single-purpose entities with near-identical names, some merged, some renamed, some dissolved.
What can be done is to pick a bias deliberately and say so. HeritageDeed biases toward recall on name-indexed searches, then suppresses the false-positive classes it can actually prove: condo master-lot scope, adversary proceedings inside a bankruptcy, same-name entities in the wrong jurisdiction. What survives both passes is flagged as unresolved rather than dropped quietly. An unresolved flag is a request for a human decision, not a defect in the report.
How do name-matching strategies actually fail?
These are the available approaches, in ascending order of cost. Each buys precision or recall at the expense of the other. No row buys both.
| Approach | What it does | Failure mode |
|---|---|---|
| Exact string match | Requires the indexed name to equal the deed name character for character | Misses SMITH, ROBERT J against ROBERT SMITH and every entity punctuation variant. High precision, poor recall. |
| Normalised match | Strips punctuation, case and corporate suffixes, standardises STREET and ST | Still misses middle names, maiden and married names, transliterations, and post-merger entity names. |
| Fuzzy or edit-distance match | Scores near-identical strings and accepts anything above a threshold | The threshold is a policy choice, not a fact. Common surnames flood the report; genuinely misspelled filings still fall below it. |
| Match plus corroborating signal | Requires a second point of agreement: address on the filing, entity officer, prior parcel history | The best precision available without a person, but it discards true hits on filings that carry no corroborating field. |
| Human review of each candidate | A person reads the filing and decides whether it is the same party | Correct, and the reason a full manual lien search costs what it costs. It does not scale to a hundred-parcel screen. |
What only a survey or a site visit can tell you
A deed describes a parcel in words, or by reference to a filed map. It does not tell you that the neighbour’s garage crosses the line by two feet, that the fence has sat three feet inside the boundary for twenty-five years, that the rear addition was built over a recorded utility easement, or that a right of way existing on paper is blocked by a wall.
For commercial work the instrument is the ALTA/NSPS Land Title Survey. ALTA and the National Society of Professional Surveyors revise the Minimum Standard Detail Requirements on roughly a five-year cycle; the 2026 revision took effect on 23 February 2026 and replaced the 2021 standards for surveys commenced on or after that date. The Table A optional items are negotiated between buyer, lender, insurer and surveyor, so survey scope is a decision, not a fixed list.
Physical inspection covers the residue that neither the record nor the survey reaches: who is actually living in the building, whether a second dwelling unit exists that no certificate of occupancy authorises, whether the property is vacant and stripped. Automation can pull the flood layer and the violation history. It cannot tell you the roof is gone.
What is automation reliably good at?
The accounting has to run in both directions to be worth anything. Inside the recorded, indexed, machine-readable record, software beats manual abstraction on axes that matter to anyone working across more than one property.
- Cost and speed per parcel. Minutes and tens of dollars instead of days and hundreds. That changes which decisions are worth researching at all, which is a different thing from doing the same research faster.
- Consistency. The same query runs identically on the first parcel and the four-hundredth. Human thoroughness varies with the hour of the day and the size of the file.
- Exhaustive variant sweeps. Testing every zero-padding form, address spelling and entity name variant is tedious for a person and effectively free for a machine. This is where automation finds records a manual search misses.
- Cross-source assembly. Recording index, tax and lien-sale status, water and sewer charges, violations, court dockets, flood and environmental layers, joined against one parcel identifier in a single pass.
- Document-to-document matching at scale. Pairing every mortgage with its satisfaction, assignment chain or release across decades of filings, which is the only way a mortgage count becomes an encumbrance count.
- Structured, diffable output. A JSON record can be re-run next week and compared field by field. That is how a portfolio gets monitored, and no abstractor’s PDF supports it.
How should a professional use an automated search?
Screen wide, verify narrow
Run automation across the entire candidate list, because that is what it is priced for. Escalate to a human abstractor, counsel, or an underwriter on the two or three parcels you actually intend to close.
Read the flags before the findings
Unresolved name matches, scope warnings and coverage notes are the report describing its own limits. A complicated parcel that comes back with no flags at all deserves less trust than one carrying three.
Send someone to look at the property
Occupancy, encroachment, condition and undisclosed tenancies appear in no index. Half an hour on site closes the largest gap in any record-based search, and it is the cheapest step here.
Order a survey wherever the boundary carries money
New construction, an addition, a shared driveway, a fence in the wrong place, any commercial acquisition. A survey is the only instrument that reconciles the described parcel with the occupied one.
Check the currency date and re-run before you commit
Every report speaks as of a moment, and the interval between that moment and your closing sits outside it. That interval is what a date-down search and, on a financed purchase, the lender’s policy exist to handle.
Get a policy if a lender is involved
HeritageDeed reports on public records at $49, $79 and $129 and delivers structured JSON alongside the PDF. HeritageDeed does not sell, issue or underwrite title insurance, and a search report is not an attorney’s opinion of title. A financed purchase requires a policy from a licensed underwriter, or an attorney title opinion letter where the investor permits one.
Key takeaways
Key takeaways
- An automated title search is bounded by what is recorded, indexed, and machine-readable. Three separate failure modes live outside that boundary: never recorded, recorded but offline, and recorded but not attributable.
- Possession rights, leases of three years or less, open mechanic’s lien windows, prescriptive easements, ripening adverse possession claims and unrecorded mortgage assignments can all bind a property while appearing in no index.
- Judgments and tax liens are indexed by party name, not by parcel. No matching threshold removes false positives and false negatives at once, so the honest approach is to choose a bias, state it, and flag what stays unresolved.
- The index is not the document, coverage floors are hard, and recording and court records often sit in separate offices. An accurate report can still be incomplete, which is why every report carries a currency date.
- Automation is genuinely better than manual abstraction at cost per parcel, consistency across many parcels, exhaustive variant sweeps, cross-source assembly, and structured output that can be re-run and diffed.
- A search is not an opinion of title and not insurance. Use automation to decide what to pursue; bring counsel, a surveyor and an underwriter to what you decide to close.
Frequently asked questions
- Can an automated title search replace a full title search?
- No, and not for the reason most people assume. It covers the recorded, indexed, machine-readable record thoroughly and cheaply. It cannot see unrecorded interests, cannot read instruments held only on microfilm, and cannot resolve name ambiguity without a person. Use it to screen, then escalate on the parcels you intend to close.
- What is the most common thing an automated title search misses?
- Name-indexed filings. Judgments, federal tax liens and UCC filings are docketed under a party name with no parcel reference, so software must decide whether the debtor is your owner. Entity variants and common surnames make that decision probabilistic, and every threshold trades a missed lien against a false one.
- Why do some deeds not appear in an online record search?
- Coverage floors. Every online recording system has a start date, and instruments recorded before it exist only as paper or microfilm at the office. In New York City, ACRIS reaches back to 1966 for four boroughs, and Staten Island records sit with the Richmond County Clerk on a separate system.
- Can a title search find an unrecorded lease?
- No. Under RPL § 290(3), a lease for a term not exceeding three years is not a conveyance and need not be recorded, so it never enters the index. Occupancy is discovered by inspecting the property or by requesting estoppel certificates from tenants before closing.
- Will a search show a mechanic’s lien on recent construction?
- Only if one has already been filed. NY Lien Law § 10 permits filing during the work and within eight months after completion of the contract, or four months for a single-family dwelling. A recently renovated property can therefore be clean in the record today and encumbered later for work already complete.
- Does an automated search detect encroachments or boundary problems?
- No. Encroachments are physical facts, not recorded ones. A deed describes the parcel; only a survey reconciles that description with what is actually built and occupied. For commercial transactions the standard product is an ALTA/NSPS Land Title Survey, whose 2026 requirements took effect on 23 February 2026.
- Is a HeritageDeed report an opinion of title?
- No. HeritageDeed reports what public records show for a parcel as of a stated date, delivered as a PDF and as structured JSON at $49, $79 and $129. It is not a legal opinion, and HeritageDeed does not sell, issue or underwrite title insurance. A financed purchase requires a policy from a licensed underwriter.
Sources
Primary records and official documentation cited in this article.
HeritageDeed provides public-record search reports only. Reports do not constitute title insurance, an attorney opinion of title, or a title insurance commitment.