AI-generated correspondence adds a new layer of complexity to transactions

Conveyancers are starting to notice something in their inboxes. The enquiries arriving on files are longer, more numerous and, on closer reading, oddly generic. Sentences are polished but the substance is thin. Points are raised that have no bearing on the property in question. Increasingly, the explanation is that the correspondence was drafted, at least in part, by an AI tool.

Volume without value

The immediate practical problem is quantity. A tool asked to produce enquiries on a report or a pack will happily generate twenty where an experienced practitioner would raise three. Each one still has to be read, assessed, answered or pushed back on. That is time on a fixed fee, and it is time that does not move the transaction any closer to exchange.

It also changes the character of the exchange between firms. Traditional enquiries carry an implicit signal: someone has read the title, looked at the searches and identified a genuine gap. When enquiries are machine-generated in bulk, that signal weakens. The receiving firm can no longer assume that every question reflects a considered concern, which makes triage harder rather than easier. Responding at length to a point that was never really a point wastes effort on both sides.

Verification becomes the real question

The more serious issue is accuracy. AI tools are fluent, and fluency is persuasive. Correspondence may cite a covenant that does not exist, misdescribe a planning position, reference the wrong edition of a protocol, or confidently assert a requirement drawn from nowhere. If a fee earner takes an inbound letter at face value and acts on it, the error propagates into the file.

That puts the burden back where it has always been: verification against source material. If a letter refers to a restriction, check the register. If it raises a search result, read the result. If it asserts a legal position, confirm it. None of this is new practice, but the volume and confidence of AI-drafted correspondence makes shortcuts more tempting and more dangerous. The cost of assuming has gone up.

File management and the audit trail

There is a record-keeping dimension too. If a firm uses AI internally to draft enquiries or client letters, the file should show what was generated, what was reviewed and by whom, and what was sent. Supervision obligations do not change because a drafting tool was involved. A partner reviewing a file in two years' time, or an insurer reviewing it after a claim, will want to see human judgement applied at the right points.

The same applies to inbound material. Where correspondence is suspected of being machine-generated and materially inaccurate, it is worth noting on the file how the point was checked and resolved. That protects the fee earner if the issue resurfaces later.

Practical steps firms are taking

Several approaches are emerging. Some firms are tightening their enquiry policy so that only enquiries relevant to the specific property and title are raised, and saying so openly in their standard correspondence. Others are training junior staff to spot the markers of generated text: repetition, generic phrasing, questions already answered in the pack, and requests for information that does not exist for that property type.

Pushing back politely but firmly also helps. A short reply noting that a question is already addressed at a specific page of the contract pack, or that it does not arise on this title, is faster than a full substantive answer and sets expectations for the rest of the transaction.

Underpinning all of it is the quality of the underlying data. If a fee earner has clear, well-presented search results and title information in front of them, dismissing a spurious enquiry takes seconds. If the information is scattered across formats and systems, every questionable point becomes a research task. The defence against noisy correspondence is not more correspondence; it is faster access to reliable source material.

AI is not going to leave conveyancing, and used well it can genuinely reduce drafting time. The risk is asymmetric though: the firm generating the text saves minutes, and the firm receiving it loses hours. Until practice settles, the discipline that protects a file is the old one, applied more consistently.

See how Searchpoint gives you the search data you need to answer enquiries with confidence.