Subscribe to our newsletter

Blog

AI-powered notary due diligence: methods, deliverables, and limitations

August 20, 2026

AI-powered notary due diligence means having a tool read through a voluminous file before human review. From titles and leases to previous deeds and appendices, the AI extracts, cross-references, and flags information.

It does not authenticate anything. It does not replace the land registry. It bears no liability. Silex helps the firm query documents, cross-reference them with legal requirements, and structure this initial review.

In short: what the AI should produce

Effective use goes beyond simply summarizing the file. It aims for a practical deliverable: an index of received, missing, or illegible documents; a timeline of transfers; inconsistencies between documents; and a memo of points to verify before the appointment.

Every alert must link back to the original document. This memo is not a legal opinion. It is a tool for the notary and the clerk.

What is it?

In notarial practice, due diligence is the verification work performed before an act. It involves checking titles, encumbrances, conditions, and the consistency of the documents.

The AI acts as a first reader. It prepares. It does not decide.

See also AI for notaries and AI and notarial practice.

Where AI really saves time

The problem isn't reading a single document. It's cross-referencing the entire file.

A preliminary agreement states one surface area. A diagnostic report states another. An old deed mentions an easement. A more recent deed omits it. These discrepancies are easily missed during a quick read.

That is where AI is most useful: comparing the same data across different documents. Dates. Identities. Amounts. Property descriptions. Outstanding conditions. Then, if necessary, cross-referencing them with legal sources.

What to ask the tool

Request a structured output, not a long text. Include the following sections: seller, buyer, co-ownership, landlord, lender. The timeline: date, type of deed, subject, and mentioned charges. Sensitive clauses: easements, reservations, inalienability, and conditions precedent.

Discrepancies: what changes from one document to another. Gaps: missing documents, missing dates, unattached annexes. Questions: what to ask the parties or the registries. Also, demand access to the original document. Without a source page, the alert is not actionable. See AI legal analysis.

Typical inconsistencies to flag

A property description that differs from one deed to another. A surface area or lot numbers that do not match. A contradictory price, deadline, or condition. Divergent identities or marital statuses. An easement mentioned and then omitted. A lease whose duration or purpose raises questions. An expired, illegible, or unsigned document.

Every alert must link back to a document. Without a source page, the alert is not actionable.

What AI must never replace

In Switzerland, notarial practice is largely governed by cantonal law. Authentication remains a human task. Legal advice remains a human task.

AI does not replace verification with the land registry, nor does it replace official extracts and registrations. It does not interpret the intent of the parties. It does not draft or read the deed on behalf of the notary.

A generated summary is not a validated title of ownership. A public chatbot is not a suitable environment for documents containing personal data. See ChatGPT vs Silex and Claude vs Silex.

6-step method

First, index: received, missing, and illegible documents. Next, classify: titles, preliminary contracts, leases, annexes, and correspondence. Then, extract: parties, dates, amounts, and sensitive clauses. Then, cross-reference: compare the same data across different documents.

Then verify: reopen each alert in the source, then check against the registries. Finally, document: who reviewed it, what is confirmed, and what remains open.

AI accelerates indexing, classification, extraction, and cross-referencing. Starting August 26, 2026, Silo will allow you to query these documents within Silex and trace them back to the source. A file analysis agent, available soon, will perform an initial pass over the entire corpus. Verification and documentation remain human tasks.

Check before appointments or signing

Does every alert have a source document? Have discrepancies in area, price, or identity been resolved? Have easements and encumbrances been cross-referenced with the registry?

Are the remaining conditions precedent identified? Are there clear follow-ups for missing documents? Has the memo been reviewed by the acting notary?

Confidentiality

Notarial files contain sensitive data. Identities, assets, finances: everything must remain protected.

Before using a tool: where is the data stored? Who has access to it? Is it used to train a model? Is it isolated between organizations?

See Silex security and AI and data protection. Source: FDPIC – AI and data protection.

Who is this for?

For notaries: prepare without delegating decision-making. See AI for notaries. For clerks: initial review, points to check, summaries. See AI in notarial practice.

For lawyers: contract review related to the transaction. See AI for lawyers. For corporate legal teams: the same document logic applies. See AI for business.

How to choose a tool

Does it actually read multiple documents at once? Does every alert link back to a source page? Can you cross-reference the file with Swiss law?

Is the data hosted in Switzerland, excluded from training, and isolated between firms? Does the workflow require human validation? See the Silex product and the solutions.

The role of Silex

Silex helps prepare clearer due diligence. Query a file. Highlight discrepancies. Generate a summary. Then verify. Three new features, all within the same environment.

Silo, coming August 26, 2026

Silo queries your firm's documents directly from Silex—titles, leases, templates, and annexes—and connects them with legal sources.

Every answer is sourced, allowing you to trace it back to the original document. Files imported into Silo Storage are hosted in Switzerland by Infomaniak on a sovereign cloud. They are not used to train models and are never shared with other organizations.

File analysis, coming soon

The agent will process an entire file at once, including documents, correspondence, pleadings, PDFs, audio recordings, along with a description of the transaction and your instructions.

Target deliverables: summaries, key points, missing items, areas of concern, and draft documents. This is a first pass; the land registry and the final deed remain outside the scope of AI.

New plans, starting August 26, 2026

Our plans include Silo and the new AI agents. Storage and credits are tailored to each offer. See our pricing.

The goal is not to automate the act itself, but to ensure a more reliable initial review. To test it out, book a demo or contact the team. Additional resources: product, FAQ and legal updates.

FAQ: AI-powered notary due diligence

What is AI-powered notary due diligence?

It is the use of AI to extract, cross-reference, and flag alerts within a file, prior to human verification.

What deliverables should you expect?

A document index, a transmission timeline, identified inconsistencies, and a memo of questions. Not a final legal opinion.

Can AI replace the land registry?

No. It prepares them. Official extracts and authentication remain human tasks.

What is the main risk?

Relying on a summary without reopening the original document or cross-referencing the registers.

Can we use a public chatbot?

No, not when personal data is involved. You need a professional tool, isolated between organizations.

How can I query an entire file in Silex?

With Silo, starting August 26, 2026, you can query your documents and cross-reference them with the law. You can trace everything back to the original document. A file analysis agent, available soon, will be able to target an entire corpus at once.

Is Silex suitable for a notary office?

Yes, for research, analysis, and document summarization, with human validation. See also Legal AI.