AI for UK Conveyancers: Reports on Title and the Checks You Keep

A UK conveyancing firm can use AI to draft a report on title from the transaction pack and to build an issue list, as long as the fee earner still reads the register, plan and searches, confirms every flagged point against the document, and signs the report. AI saves reading and typing time. It cannot take over the professional judgement or the accountability.

The benefit is largest on the repeat work: summarising a long lease, listing the entries on a charges register, drafting the standard paragraphs of a report for a lender or a client. The risk is largest where the tool sounds sure and is wrong, so the checks you keep deserve as much planning as the drafting.

Follow me on Instagram@sagnikteaches

What the Law Society's guidance says, and what it leaves to you

The Law Society's guide, Generative AI: the essentials (dated 30 June 2026 on the page, which says it is being updated again), is general guidance for solicitors, aimed particularly at small and medium-sized firms. It is not specific to conveyancing, but several of its points bear on it directly.

Connect on LinkedInSagnik Bhattacharya

• Professional obligations under the SRA Code of Conduct, the SRA Standards and Regulations and the SRA Principles apply whether AI was used by the solicitor or by anyone under their supervision.
• The guide expects firms to define the purpose and use cases of a tool, follow relevant guidance, and align with their wider policies on IT, AI, confidentiality and data governance.
• Firms should regularly review a vendor's data management, security and standards, and establish whether the tool is a closed system within the firm's boundaries or also operates as a training model for third parties.
• It describes hallucination, where outputs are presented as accurate but are incorrect, and refers to fabricated case citations reaching the courts. The same logic applies to a fabricated clause number in a report.

On 21 October 2026 the Law Society added a longer practice note, Using artificial intelligence in legal practice, which is more direct on points a conveyancing team meets every week:

Subscribe on YouTube@sagnikteaches

• The use of AI does not diminish or transfer your professional responsibilities; you remain accountable for all work and advice, whether or not AI helped prepare it.
• Using AI inside your own secure systems, where data is not used to train a model, is not likely of itself to need informed consent, and transparency is the likely approach. Using client information in a way that could breach confidentiality, including to train AI, is likely to need specific consent.
• Pasting confidential client information into a tool without adequate safeguards should be treated as a potential breach of confidentiality, with both regulatory and data-protection consequences considered.
• AI use can change the firm's professional indemnity risk profile, so raise it with your broker or insurer, particularly at renewal.
• Identity and electronic verification processes should be reviewed for resilience to AI-made deepfakes, especially where you never meet the client in person.

That last point matters in a sector where payment and identity fraud are constant, and the tutorial on conveyancing client onboarding, ID and forms covers where AI helps and where checks stay manual at that stage.

Licensed conveyancers answer to the Council for Licensed Conveyancers rather than the SRA. The CLC's AI and Technology Guidance is read with its Code of Conduct and is built around eleven AI and Technology Principles. One of the harms it names is a failure to ensure human review of AI-generated reports on title, leaving clients inadequately advised about title problems, which is the same line this workflow draws. Both Law Society documents are built on SRA rules, so a firm working under another UK jurisdiction's conveyancing practice should read its own professional body's material alongside them; UK Finance publishes a lenders' handbook for each legal jurisdiction in the UK.

None of these documents gives a task-by-task list for a conveyancing file; the fee earner and the firm decide that. Read them, set a short policy, and use the steps below to decide which jobs to try.

What goes into a report on title, and which parts AI can draft

A report on title (or certificate of title, depending on the client and the lender) tells the reader what the firm has found about a property's legal title and what follows from it. For a residential purchase with a mortgage, lenders' requirements usually sit in the lender's instructions or in the UK Finance Mortgage Lenders' Handbook, and the firm's own template governs the format for clients. Always follow the current lender instructions and the firm's template; nothing below overrides either.

The pattern is that AI handles the "what does it say" layer and you keep the "what does it mean" layer. Where the second depends on the first being right, you check the first.

Prepare the pack before you prompt

A report is only as good as the pack. Before you feed anything to a tool, make a clean, ordered set of the documents you are going to rely on: official copies of the register and title plan, the contract and any special conditions, the seller's property information form and fittings and contents form, search results, the lease for a leasehold, management pack material, and replies to enquiries. Give each a short reference. This helps the human reader as much as the tool, and it lets the report cite "Doc 7, page 2" so anyone can verify it.

Think about data too. A pack contains names, addresses, bank details for the completion funds and sometimes identity documents. The report does not need most of them. Keep identity documents and client-money details out of any AI tool, mask what you can, and use a firm-approved business account rather than a personal one. The tutorial on whether solicitors can use ChatGPT without breaching confidentiality goes through the account and settings questions in more detail.

Step 1: Register entries into plain English with a trace

Extraction prompts work best when they require the tool to quote the source line and point to its location. That makes every output checkable in seconds.

Below is the text of the charges register and property register
from official copies. For each numbered entry, give me:
- entry number (as printed)
- the entry's wording, quoted exactly
- a one-sentence plain-English description
- the type: charge / restriction / notice / covenant / easement / other
- any date, amount or named party mentioned
Do not interpret legal effect. Do not omit any entry. After the
list, state how many entries there are in each register.

TEXT: [paste]

The count at the end is a cheap integrity check. If the register has 11 entries and the tool lists 10, one was dropped, and you find it before the report goes out.

A realistic sample output (illustrative): the tool lists entry 4 as "a restriction requiring a certificate from a named conveyancer before disposal", quotes it, and describes it as a type of restriction. Your job is to read the restriction in the register and decide what it requires in this transaction: the certificate, a consent, or nothing because the condition has already been satisfied. The tool did the clerical work and you did the lawyering.

Step 2: Lease extraction into a table

Leaseholds are where the largest time savings sit, because a lease can run to forty or fifty pages of dense drafting. Ask for a structured table with page references, and keep the instruction to quote.

From the lease text below, extract into a table:
term (years, start date, unexpired term as of [date]); ground rent
and review provisions; service charge provisions and any cap;
repair and decoration obligations; restrictions on alienation,
subletting and pets; any forfeiture clause; consent requirements
and fees; and anything unusual. For each, give the clause number
and quote the operative words. If a topic is not covered, write
NOT FOUND, never assume.

Illustrative result: the table says ground rent "doubles every 25 years" and quotes clause 3.2. You read clause 3.2 and find that the doubling applies only on the first two reviews and then converts to a fixed increase. The tool compressed the clause into something shorter and wrong. That is the key failure mode for lease extraction: a confident summary of a clause with conditions. The remedy is the quote requirement, and your reading of every clause the advice depends on.

Check unexpired term, ground rent and any review mechanism against the lender's current requirements, and read the whole lease for onerous covenants. Do not stop at the table.

Step 3: Cross-document consistency checks

One valuable use is a consistency check across documents: does the seller's property information form say no disputes, while the replies to enquiries mention a dispute with a neighbour? Does the contract's description of the parking match the title plan? Does the fittings form list items the seller's agent's details exclude? These are the kinds of mismatches a tired fee earner misses on page 40 of a pack.

I am giving you extracts from three documents about the same
property. List every point on which they differ or appear to
conflict: for each, the topic, what each document says (quoted),
and the document reference. Do not resolve conflicts. If you find
none for a topic, say so.

A realistic mistake: the tool flags an apparent conflict about the boundary because one document says "the garden fence" and another "the rear boundary". Not a conflict. The discipline of listing, quoting and referencing makes it quick to dismiss the false positives and chase the real ones. Real ones (an unregistered right of way mentioned in a reply to enquiries, a planning condition that the property information form omits) become entries on the issue list.

Step 4: Build the issue list, then rank it yourself

Ask the tool to compile the points from steps 1 to 3 into an issue list, with the source of each. Then rank them yourself. A useful scale is: must resolve before exchange; advise client and record instructions; note for the file only. Ranking depends on the client's circumstances, the lender's requirements and the risk, and that is your judgement.

A filled-in example (illustrative; the three rows show the range):

Do not let the tool assign the rank. It will rank by how alarming the wording is, which is not the same as legal or commercial importance.

Step 5: Draft the narrative from your confirmed facts

Only now do you ask for the report prose. Give the tool the confirmed issue list and your decisions, and the firm's template sections. The prompt supplies the facts; the tool supplies the sentences.

Using ONLY the confirmed facts and decisions below, draft the
"Title and issues" section of a report on title in the firm's
usual tone: clear, short paragraphs, no legal conclusions beyond
those I have written. Where I have written "ADVISE", use neutral
wording that recommends the client takes advice or gives
instructions. Do not add issues. Do not soften or strengthen
anything I have decided.

CONFIRMED FACTS AND DECISIONS: [paste]

The instruction not to soften or strengthen is there because a tool left alone will smooth awkward findings into reassuring wording. "There is a restriction that must be dealt with before completion" becomes "a standard restriction applies". You read the draft against your list and put it right.

The checks a fee earner keeps, in full

Whatever a tool did, these stay with a person, and they are the part of the routine that protects the client and the firm.

1. Read the official copies yourself. Every entry, including notices, restrictions and the title plan. Compare plan and physical description.
2. Read the operative lease clauses that the advice relies on, not just the table. Check the unexpired term and the ground rent against the lender's requirements.
3. Read the search results and decide which findings need further enquiries or advice.
4. Verify every factual statement in the report against a document in the pack, and make sure the document reference is correct.
5. Check the lender's current instructions and any special conditions before the report goes out. These change, and no tool can know which version applies to your instruction.
6. Confirm identity, source of funds and other compliance checks using your normal systems; these are not AI tasks.
7. Take the client's instructions on each issue, and record them.
8. Sign it yourself. If you would not be willing to explain to a client or a regulator how a statement got into the report, take it out.

A small-firm example: one purchase, start to finish

Say a four-fee-earner firm acts for a buyer of a leasehold flat with a mortgage. The paralegal assembles a pack of 14 documents and names each one. The fee earner runs the register extraction (11 entries, count matches), then the lease extraction. The lease table gives 15 headings; she reads clauses 3, 6, 8 and 11 in full because the advice depends on them, and corrects the ground rent summary. A consistency check across the property information form, the replies and the contract flags two differences: one false positive about a fence, and one real mismatch about who is responsible for the roof and common parts.

She builds an issue list with five entries, ranks them herself, calls the client to take instructions on the roof point, and then asks the tool to draft the narrative from her notes. The draft changes "must be resolved" to "should be considered" in one place, which she reverses. She checks every factual statement against a document reference, reads the lender's instructions once more, signs the report and sends it. A second fee earner reads the finished report with the pack. The firm's own timing might show the drafting stage shrinking while the checking stage does not; that is the right shape.

Where it goes wrong

• The dropped entry. A register had a notice at the foot of a page. Show-up: the count in the tool's output is lower than the register's. Fix: always ask for the count, and count the register yourself.
• The over-neat clause summary. A conditional clause reduced to a single plain sentence. Show-up: the quoted wording contradicts the summary. Fix: quote requirement, read the clause.
• The invented reference. A clause or form number that does not exist. Show-up: clicking the reference leads nowhere. Fix: every reference must resolve to a page in the pack.
• The stale form. A report that cites a superseded form or lender requirement. Show-up: the lender's checker rejects the report. Fix: confirm the current version from the source before use, not from the tool.
• Client data in the wrong place. A personal account used under time pressure. Show-up: the usage log or a colleague's remark. Fix: an approved account, a short policy, and a rule that you report the slip the same day.

Two more jobs worth trying: enquiries and the client summary

Beyond the report itself, two smaller tasks fit the same pattern of "tool drafts, fee earner decides".

Drafting additional enquiries from the issue list

Once you have ranked the issues, the enquiries to the seller's conveyancer are mostly standard phrasing around a specific point. Give the tool your issue and your wanted outcome and ask for a numbered enquiry in the firm's tone.

Draft one enquiry for each issue below, addressed to the
seller's conveyancer. State the point, quote the document and
page, and say exactly what we need (a document, a confirmation,
an indemnity quote). One short paragraph each. No assertions of
law.

ISSUES: 1. Register entry 4 restriction: need the certificate.
2. Roof repair responsibility: lease cl. 8.1 says the lessee repairs
the structure; the management pack says the freeholder does.
Need a written explanation of the current practice.

A typical result is neat and polite and sometimes adds a request you did not ask for ("please also confirm there are no disputes with neighbours"). Strike anything you did not choose to ask, because every extra enquiry costs time on the other side and can invite a reply you then have to deal with.

A plain-English summary for the client

Buyers rarely read a full report. A one-page cover note in plain English, covering what they are buying, what the main issues are, what you recommend and what you need from them, improves understanding and cuts the phone calls. Feed the tool the finished, signed-off report (with names masked) and ask for a summary at a stated reading level, with the instruction not to add or remove any issue. Compare it against your list. The check is that the summary mentions every "must resolve" and "advise client" item and nothing the report does not say.

Before and after (illustrative): the draft says "The lease contains some unusual terms that should not cause problems". You change it to "The lease makes you responsible for repairing the inside of the flat and contributing to the building's costs. The ground rent can rise at set reviews. We recommend you read section 4 of our report before exchange." The first version reassures without saying anything; the second tells the client what to look at.

Measuring whether it helps

Pick three numbers from your case-management system for the first ten files: minutes from pack complete to first draft report; number of corrections the second reader makes; and number of enquiries raised after the report that the issue list should have caught. If drafting time falls and the other two stay flat or fall, keep going. If corrections rise, the extraction prompts need tightening, the quote requirement is being skipped or the pack is poorly ordered. Review after ten files and again after thirty, and record what you changed.

Writing it into the firm's policy

A one-page policy is enough to start: which tools are approved, for which tasks (register and lease extraction, consistency checks and first-draft narrative are sensible starting points), which are not (identity checks, source-of-funds, anything needing legal judgement), who checks the output, and where the usage note is recorded on the file. Review it twice a year. The 90-day implementation plan for a small law firm shows how to sequence adoption, and the tutorial on cutting admin on each transaction covers the work around the report.

For the wider picture of what AI contract review does well and badly, see AI contract review for small firms; much of that applies to lease review. If you would rather map your own report process with someone, that is what my AI implementation consultation covers, and you can start at the contact page.

Questions conveyancers ask about AI and reports on title

Can AI write the whole report on title?

It can draft the narrative from the pack, but it cannot be the report. The fee earner has read the title and the searches, formed the view and signed the report, and that responsibility cannot be handed to a tool. Use AI for the first draft and issue list; the checks and the conclusion stay with you.

Is client data safe in a conveyancing AI tool?

Only if you have checked the tool. The Law Society's generative AI guidance says firms should review the vendor's data management and security, and establish whether the tool is closed to the firm or also trains models for third parties. Use an approved business account and keep identifiers out where you can.

Should I trust AI to spot title defects?

Treat it as a second reader that sometimes sees what you missed and sometimes invents what is not there. Anything it flags goes back to the document for you to confirm. Anything it does not flag is not cleared, because it can miss a restriction on page three as easily as a person can.

What about AI use and my regulator's duties?

Your duties under the SRA Standards and Regulations apply whether or not AI helped. The Law Society's practice note of 21 October 2026 says AI does not diminish or transfer your professional responsibilities, and its generative AI guide tells you to review outputs for accuracy. Licensed conveyancers should read the CLC's AI and Technology Guidance. Record your firm's AI use in its policies and tell clients where relevant.

Further reads

• How Small Law Firms Use AI to Answer Client Status Questions — Keeping buyers and sellers updated without more calls.
• How Estate Agents Use AI to Prepare Valuation Appointments — The estate agent's side of the same transaction.
• How Small Law Firms Use AI to Draft Letters and Routine Documents — Routine letters and documents drafted for review.
• Which Legal Tasks Should a Small Firm Never Hand to AI? — Tasks a small firm should never hand to AI.
• How to Summarise Bundles and Transcripts With AI Safely — Summarising large bundles without losing detail.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — Check the firm is ready before widening AI use.
• Best AI for Conveyancing: ID Checks, Searches and Client Chasing — A stage-by-stage guide to AI in a conveyancing file: ID and source-of-funds checks, reading search results, and chasing clients, with the human checks kept.
• Conveyancing AI Software: What a Fee Earner Should Test in a Demo — A demo script and question list for small conveyancing firms: integration, data location, audit trail, per-file versus per-seat pricing.
• Conveyancing Fraud: A Call-Back Procedure for Changed Bank Details — Write your firm's call-back steps for changed bank details, with a client script and rules for deepfake and voice-clone attempts.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: the Law Society's Generative AI: the essentials (dated 30 Jun 2026) and practice note Using artificial intelligence in legal practice (21 Oct 2026), the CLC's AI and Technology Guidance, the UK Finance Mortgage Lenders' Handbook site, SRA Code of Conduct references quoted by the Law Society, vendor privacy pages (all read 4 Nov 2026) and standard conveyancing practice; always check current lender handbooks and forms.