What a professional sign-off on an AI-assisted draft covers

Signing off an AI-assisted draft is an act of authorship rather than proofreading. When a reviewer puts their name and membership number on a report that goes to a local planning authority, they are saying the methodology is appropriate, the evidence supports the conclusions and the work falls inside their competence. The draft's origin changes none of that. What it changes is where the reviewer's time goes, because the structure and the policy framing arrive already built and the professional judgement does not.

What the reviewer is being asked to confirm

Three things, and they are worth separating.

That the method used is the right one for this site and this question, measured against the relevant standard or guideline. That the site-specific evidence in the report is accurate, sourced and sufficient. And that the conclusions follow from that evidence rather than from a plausible template.

A fourth sits underneath all of them. The reviewer must hold the competence, and where relevant the licence or registration, for the work in question. An acoustician who reviews a BS 4142 assessment is inside their scope. The same acoustician reviewing a bat survey is not, whatever the quality of the drafting.

Professional bodies say the same thing in their conduct rules. Acting only within your competence, and not lending your name to work you did not supervise, are baseline obligations across the institutes, and no software changes that.

Reading a draft against the site evidence

An AI-assisted draft is grounded in policy text, national guidance and the structured data it was given. It has not stood in the garden, opened the manhole or watched the junction at half past eight in the morning.

So the review starts with the site, not with the document. Walk the site or use the survey data you trust, build your own picture, and then read the draft against it. Reading the draft first anchors you to its framing, which is precisely the failure mode to avoid.

Watch for the specific ways a structured draft goes wrong. Policy citations that are correct in general but not the operative policy for this authority. Thresholds quoted at national level where a local plan sets a lower one. Constraint layers described accurately from open data but out of date, because a conservation area boundary was extended last year. And reasonable-sounding sentences about site conditions that no one has verified.

Where the underlying site information was thin in the first place, the problems compound. The note on briefing a specialist when the site data is incomplete covers how to set that expectation before the work starts rather than at review.

Sections that have to be rewritten rather than approved

Some parts of a technical report cannot be inherited, however well drafted.

• The methodology statement. It must describe what you did, on what dates, with what equipment and under what conditions. Nobody else can write that sentence for you.
• Survey results and raw data. These belong to the person who collected them.
• Evaluation and professional judgement. Assigning value to an ecological feature, deciding whether an impact is material, grading condition, setting a significance level. This is the part clients are actually paying for.
• Mitigation commitments. If the report says something will be done, you are recommending it and someone will be conditioned to deliver it.
• Limitations. Write your own, covering what you could not access, what you did not test and what the report should not be used for.

Everything else, including the site description, the policy context, the regulatory background and the structure, is fair game for an assisted draft that you then check.

Missing data flags and what to do with them

PlanScope's drafts insert a marker where a site-specific fact is missing rather than filling the gap with something plausible. That is deliberate, and it is the single most useful feature of the draft for a reviewer.

Resolve each flag, do not delete it. A flag removed without a value becomes an unexamined assumption, and unexamined assumptions are what show up in a consultee objection three months later.

Keep a simple discipline. Every flag gets one of three outcomes: a value with a source, a statement that the information is not available and why that does not prevent a conclusion, or a recommendation for further work. A report that leaves flags unresolved and unremarked is a report that has not been reviewed.

The scope of the declaration

A sign-off declaration should say what was reviewed, which version, on what date, and against what standard. It should also say what the reviewer did not do.

That second half carries most of the risk management. If you did not visit the site, say so. If you relied on a third party topographic survey, name it and date it. If your review covers the acoustic chapter and not the transport chapter, make the boundary explicit rather than leaving a reader to infer that a signed report is signed throughout.

Within PlanScope, a review request moves through defined states: sent, accepted, in review, changes requested, signed off, rejected and expired. Use changes requested rather than signing with caveats buried in an email. A rejected or amended draft with a clear reason is a better record than an approval with a footnote.

Check that your professional indemnity cover extends to review work carried out on documents you did not originate. Some policies treat that differently from authored work, and the time to find out is not after a consultee challenges the methodology.

Records worth keeping

Keep the version you reviewed as well as the final one. Keep your comments in a form that shows what changed and why. Keep site visit notes, photographs with dates, and the correspondence that settled any disputed point.

Six months later, when the case officer asks how a figure was derived, the difference between a defensible position and an uncomfortable one is usually a folder. Competence, evidence and record keeping travel together, which is the same message running through the competence expectations under the building safety regime.

Where review cannot substitute for authorship

Several report types require a licensed or competent professional as a matter of law, and in those cases an AI-assisted draft is a starting point for that professional to work from, never something to be reviewed into compliance.

An Environmental Statement under the Town and Country Planning (Environmental Impact Assessment) Regulations 2017 must be prepared by competent experts. Protected species survey work and any mitigation licence require a suitably licensed ecologist under the Wildlife and Countryside Act 1981 and the Conservation of Habitats and Species Regulations 2017. The statutory biodiversity metric under the Environment Act 2021 expects a competent person. Land contamination reporting to BS 10175 requires a suitably qualified and experienced person. A Fire Statement at Planning Gateway One requires fire engineering competence.

In each of those, the professional is the author. The draft saves them the framing, the policy assembly and the document structure. It does not save them the survey, the licence or the judgement, and a declaration that implies otherwise is worth less than no declaration at all.

Reviewers who want to take on this work can register through the professional directory, and anyone scoping a project can browse the report types and their professional flags to see which documents carry a legal competence requirement.

This is general guidance, not legal advice; check the position with your professional body or your insurer before relying on it. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval. For specialist technical reports, qualified professional input may be required.