Reviewing an AI-drafted report: the reviewer's scope

Reviewing an AI-drafted planning report is a narrower job than writing one, and the danger for the reviewer lies in the gap between what the client thinks they have bought and what the reviewer has actually done. Set the boundary in writing before you start. A review confirms that the document is structurally sound, that its policy references are current and correctly applied, and that its site-specific facts are supported by evidence you have seen. It does not convert a desk-produced draft into a survey report.

That distinction is what protects you. It is also what makes the service useful, because a drafted document reviewed properly by a competent professional is a legitimate and efficient way to produce the narrative parts of an application.

What review means and what it does not

Write the scope into the engagement in four lines.

• Structure and completeness against the validation requirement the report is answering.
• Policy citations: correct, current, correctly applied to the facts.
• Factual grounding: every site-specific statement supported by a document, a drawing or a survey you have seen.
• Professional judgement: the conclusions are ones you are prepared to stand behind in your own name.

Then say what is excluded. No site visit unless one is priced in. No survey, monitoring or measurement. No structural, ecological or geotechnical opinion beyond your own discipline. No verification of the client's ownership, instructions or intentions.

PlanScope drafts are produced from structured planning data with AI assistance, and the platform's own standing statement is that reports should be reviewed before submission and that planning approval is not guaranteed. The review you are being asked for is the step that statement refers to, so it needs to be a real piece of work with a record.

Facts that must be independently verified

A drafted report will contain assertions that read plausibly and are only sometimes right. Check these every time.

1. The site address, the red line boundary and the site area, against the application drawings.
2. The description of development, word for word against the application form.
3. The constraint list: flood zone, listed buildings and their grades, conservation area, Green Belt, Tree Preservation Orders, protected sites. Verify each against the primary source rather than accepting the draft's summary.
4. The existing lawful use, and the planning history including any conditions that restrict it.
5. Any measurement: floor areas, heights, distances to boundaries, numbers of units and parking spaces.
6. Local plan policy numbers and the adoption date of the plan, because policies are renumbered when plans are replaced.

Constraint detection tools, including the automatic checks that produce a draft's constraint list, are a screening aid. A screening result showing no designation within a radius is not evidence that none exists, and a reviewer who relies on it without checking the primary dataset has taken on a risk they cannot see.

Site-specific data and the flags left in a draft

A well built drafting pipeline marks missing site data rather than inventing it. PlanScope inserts a to-be-confirmed flag wherever a site-specific fact is absent, which is a deliberate safeguard.

Treat every flag as a task. Either you obtain the fact and insert it, or you tell the client in writing that the fact is needed and that the report cannot be submitted until it is supplied. What you must not do is delete the flag because the sentence reads better without it.

Keep a short table in your review record: the flag, what was needed, who supplied it, the document relied on, and the date. If the report is ever tested, that table is the answer to the question of how the figure got there.

Where uploaded photographs or documents contain third party or confidential material, note that the position on retention and sharing should be checked against the platform's current policies before anything sensitive is uploaded. That is a matter for the client rather than for your review.

Policy citations and edition checking

Policy churn is where drafted documents age fastest. The National Planning Policy Framework has been revised repeatedly and its structure changed with it, moving from paragraph numbering to policy-coded chapters in the current edition. A draft that cites a paragraph number from an older edition is not wrong so much as unusable, and a case officer will notice.

So check three things. That every national policy reference names the edition. That the reference resolves in the edition named. And that local plan policies quoted are from the adopted plan rather than an emerging one, with the status of any emerging policy stated explicitly.

Then read the application of the policy, which is the part that needs professional judgement. A draft can cite the right policy and draw the wrong conclusion from it, and that error is invisible to any automated check.

Recording the review and its limits

Your review record is a professional document in its own right. Six items cover it.

The scope as agreed, including exclusions. The version of the draft reviewed, with a date and a file reference. The list of documents relied on. The verification table for site-specific facts. The changes you made or required, with reasons. And a statement of the limitations of the review.

Sign-off matters procedurally too. Where a review is arranged through PlanScope's professional directory, the request moves through defined statuses from sent and accepted to in review, then to changes requested, signed off, rejected or expired. Using "changes requested" rather than signing off a draft you are not satisfied with is the right use of that mechanism, and it creates the record.

Where the draft must be rewritten rather than edited

Some drafts cannot be repaired by editing, and recognising that early saves everyone money.

Rewrite from scratch where the report is structured around the wrong policy test, where the conclusion contradicts the evidence in the same document, where the site description does not match the site, or where the document is answering a different validation requirement from the one the council has set. Editing a document whose logic is wrong produces a document whose logic is still wrong and whose prose is inconsistent.

Say so plainly and price the rewrite. A client who is told at the outset that a draft needs rebuilding usually accepts it. A client who receives a heavily edited document two weeks late, with an invoice for the extra time, does not.

Declining work that needs survey evidence

Some report types cannot be signed off on a desk review at all, because the law requires a competent or licensed person to have done the underlying work.

For these, a draft is a starting point for the professional rather than a substitute for their work. If you are asked to sign one off without the survey, decline, and say which survey or licence is missing. Our notes on Fire Statement competence at Gateway One and on checking a bat licence holder before you hire set out the competence questions for two of these in more detail, and the distinction between a chartered town planner and a planning agent matters for the narrative reports.

For specialist technical reports, qualified professional input may be required. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval.