AI and Planning

Reading an automated constraint check and its limits

7 min readBy PlanScope Editorial Team

An automated check tells you which published datasets intersect your boundary. Curtilage, tree preservation orders, local records and earlier obligations are not among them.

An automated planning constraint check tells you which published national datasets intersect a boundary you drew. That is a useful thing and a narrow one. It is a query against mapped data at a moment in time, and the answer is only as good as the currency, the resolution and the completeness of the layers being queried.

Used properly it saves a developer several hours per site and catches the designations that decide whether a scheme is viable. Used as a clearance certificate it produces expensive surprises at validation.

Which datasets an automated check uses

PlanScope checks a site against Environment Agency flood data, Historic England listed buildings and conservation areas, and Natural England Sites of Special Scientific Interest layers, and feeds the result into a rule-based engine that suggests which reports a project is likely to need.

Those are the authoritative national sources for the constraints they cover. They are also, between them, a partial picture of what affects a planning application. Knowing which layers were queried is the first thing to establish about any automated check, because the important question is usually what was not looked at.

The same applies to the thresholds the engine applies. Major development at ten or more units, 1,000 square metres of floorspace or half a hectare. Transport Assessment at fifty or more units or 2,500 square metres. Flood Risk Assessment in Flood Zone 2 or 3 or on sites of one hectare or more. Affordable Housing Statement at ten or more units. These are the thresholds the product uses to flag reports, and a council's own validation list and national policy thresholds take precedence over them.

Flood zones and the date of the mapping

The Flood Zones are policy zones showing the probability of flooding from rivers and the sea, ignoring the presence of defences. That last point catches people. A site behind a substantial flood defence still sits in Flood Zone 3 on the map, because the zone describes the undefended position.

The mapping is revised. The national flood risk assessment has been rebuilt more than once and zone boundaries have moved, so a check run eighteen months ago may not match the map today. Record the date of the check.

Two things the flood layer does not carry. Surface water risk, which is mapped separately and is the source that affects more sites in practice. And groundwater or sewer flooding, which are not nationally mapped at all in a form usable at site scale.

A result of Flood Zone 1 therefore means low probability of river and sea flooding, and nothing more.

Heritage designations and curtilage

Listed building data gives the list entry, the grade and a point or polygon from the national heritage list. Conservation area data gives boundaries where the authority has supplied them.

The limitation that causes the most trouble is curtilage. The protection of a listed building extends to structures within its curtilage that were in place at the relevant date, and the extent of curtilage is a legal question decided on the facts rather than a line on a map. A national dataset cannot tell you whether the barn at the end of the garden is curtilage listed. Neither can any automated check.

Setting is the second. Harm to the setting of a heritage asset is a material consideration whether or not the site itself is designated, and setting has no mapped extent. A site 200 metres from a Grade I church with an open view towards it may raise a heritage issue that a proximity search does not flag.

Third, non-designated heritage assets, including locally listed buildings and locally identified archaeological areas, sit on council registers rather than in national data. Ask the conservation officer.

Ecological designations and buffer distances

Sites of Special Scientific Interest have mapped boundaries, and so do Special Areas of Conservation, Special Protection Areas and Ramsar sites. Impact risk zones published around them indicate the development types that trigger consultation with the statutory nature conservation body, and those zones extend well beyond the designated boundary.

Below that level the mapping thins out quickly. Local Wildlife Sites are held by county record centres. Priority habitat inventory polygons are modelled from national datasets and are indicative rather than surveyed, so an ecologist will routinely confirm or reject them on the ground. Ancient woodland has a national inventory, but the buffer expectation applied in decisions is a policy matter rather than a mapped one.

And no dataset records whether a bat roost, a badger sett or a breeding bird is present. That requires a survey, in season, by somebody competent, and for European protected species a licensed ecologist.

Mandatory biodiversity net gain applies regardless of designation, so a clean ecological search does not remove the need for a habitat baseline.

What no constraint found does not mean

Take the phrase literally. No constraint was found in the datasets that were queried.

Several categories of constraint sit outside national data entirely.

  • Tree Preservation Orders, many of which exist only as scanned orders in a council register.
  • Article 4 directions with their precise boundaries, which vary in how well they are published.
  • Contaminated land register entries and historic landfill within influencing distance.
  • Critical drainage areas and the lead local flood authority's own records.
  • Section 106 obligations and restrictive conditions imposed by an earlier permission on the land.
  • Local plan site allocations, settlement boundaries and neighbourhood plan policies.
  • Nutrient neutrality and water neutrality catchments, which have been extended more than once.
  • Highway adoption status, public rights of way and private access rights.

None of these appear reliably in an automated check, and several of them can stop a scheme. The council's local validation checklist is the document that tells you what will actually be demanded, and it is authority-specific.

Confirming a result with the authority

Two routes, and both are worth using.

A written pre-application enquiry gets you a considered answer with a named officer, at a published fee. It is the right route where the constraint position affects value.

A direct enquiry to the relevant specialist team is quicker and free in most authorities. The tree officer for protection orders. The conservation officer for curtilage, setting and local listing. The lead local flood authority for surface water records. The contaminated land officer for register entries.

Ask specific questions. "Is there a Tree Preservation Order affecting the trees on the northern boundary of the site at the address below" gets an answer. "Are there any constraints on this site" does not.

Where a drafted document is being prepared alongside, the constraint findings should be checked into it rather than assumed, and the errors worth hunting for are set out in factual errors to look for in a drafted planning document. A pre-application meeting is also the natural place to test the constraint position, as using a draft to prepare for a pre-application meeting explains.

Recording the check with a date

Treat the output as a dated record rather than as a standing fact.

Save the result with the date, the datasets queried and the boundary used, because the boundary matters. A check run on a red line that excludes the access strip will miss a constraint on the access strip. Where the site boundary changes during a project, rerun the check.

Then rerun it before submission in any case. Designations are made, extended and revoked, and an application supported by an eighteen month old constraint search invites a validation query that costs a fortnight.

Automated checks give a starting position quickly, and the report recommendation that follows is a suggestion rather than a validation list. Enter a postcode on PlanScope to see the likely bundle, browse the report types at all 66 report types by category, and see pricing for the free report, the £50 per report option and the Professional tier.

PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval. This is general guidance rather than legal advice, so check the position with your local planning authority or a qualified professional before relying on it.

Frequently asked questions

Which datasets does an automated constraint check query?

PlanScope checks Environment Agency flood data, Historic England listed buildings and conservation areas, and Natural England Sites of Special Scientific Interest layers, then applies threshold rules to suggest likely reports. Those are authoritative for what they cover, and between them they are a partial picture of what affects an application.

Does Flood Zone 1 mean the site will not flood?

No. The flood zones describe the probability of flooding from rivers and the sea, ignoring defences, so a site behind a substantial defence still shows in Zone 3. Surface water risk is mapped separately and affects more sites in practice. Groundwater and sewer flooding are not nationally mapped at site scale.

Can a dataset tell me if a building is curtilage listed?

No. Curtilage listing depends on what structures were within the curtilage of the listed building at the relevant date, which is a legal question decided on the facts rather than a mapped line. Ask the conservation officer. Harm to setting is also material and has no mapped extent at all.

What is missing from a clean automated result?

Tree preservation orders held only as scanned council records, Article 4 boundaries, contaminated land register entries, critical drainage areas, section 106 obligations and conditions from earlier permissions, local plan allocations, neighbourhood plan policies and nutrient or water neutrality catchments. Several of these can stop a scheme.

How often should a constraint check be repeated?

Whenever the site boundary changes, and again before submission. Designations are made, extended and revoked, and national flood mapping has been rebuilt more than once. An application supported by an eighteen month old search invites a validation query that can cost a fortnight.

Sources and further reading

constraintsdatasetsdue diligenceflood zonesheritage

This article is general guidance about the planning system in England and, where stated, the rest of the UK. It is not legal or professional advice. Check your local planning authority’s validation requirements and take professional advice for your specific site.

More in AI and Planning

View all →
AI and Planning7 min read

Where site judgement cannot be drafted in a design statement

A draft can carry the structure, the policy references and the checklist. Context, character and the access arrangements that depend on levels come from looking.

Read article →
AI and Planning6 min read

Explaining an AI-assisted draft to a client or a committee

Sooner or later a client, an officer or an objector asks how a document was produced. A planner who has thought about the answer in advance gives a better one.

Read article →
AI and Planning7 min read

Why planning applications are returned as invalid

Validation is a completeness check against the national requirements and the council's own local list. Here are the defects that cause a return and how to deal with an invalidity notice.

Read article →
AI and Planning7 min read

Drafting to an appeal deadline without cutting corners

Appeal timetables do not move, and a statement of case submitted late is normally returned unread. Here is where drafting buys time and where only evidence will do.

Read article →
AI and Planning7 min read

Version control when report sections are regenerated

A report assembled from independently produced sections is only consistent if somebody checks it after the last edit. The dependencies that break, and the controls that catch them.

Read article →
AI and Planning7 min read

What AI drafts get right, and wrong, in planning reports

A drafting tool knows what a Transport Statement contains. It does not know what the traffic count said. Where the real review time goes on an AI-assisted planning document.

Read article →