NeuraCities
All posts
GISZoningAI GISLand use

AI for zoning: code lookup vs zoning analysis on the map

The NeuraCities Team 9 min read

"AI for zoning" covers two jobs that are easy to confuse. One is reading the zoning code: what does R-2 allow, what is the side setback, is a duplex a permitted use. The other is analyzing zoning on the map: which parcels are zoned for apartments, how many acres that adds up to, and how much of it sits in the floodplain.

The first is a question about text. The second is a question about geography. Different tools are built for each, and knowing which one you need saves a lot of frustration. This guide explains both, the caveats that apply to any AI answer about zoning, and where NeuraCities fits.

Two different jobs: reading the code and analyzing the map

Code lookup (text)Zoning analysis (map)
Typical question"What can I build on an R-2 lot?" "What is the rear setback in C-1?""Which parcels are zoned for multifamily?" "How much commercial zoning is in the floodplain?"
What the AI readsThe adopted zoning ordinance, plans and related documentsThe zoning district layer, parcels and other GIS layers
What comes backA plain-language answer, ideally with a citation to the sectionA map, a table of parcels, acreage totals and charts
Who asksResidents, applicants, permit staff, real estate due diligencePlanners, housing and economic development staff, consultants, developers
What can go wrongMisreading a definition, missing an overlay or an amendmentAn outdated zoning layer, missing overlays, parcels split across districts

Most real projects need both. A housing study asks the map question first (where is multifamily allowed, and how much land is that) and then the text question for the short list (what density and height does that district actually permit).

AI that reads the zoning code

Zoning ordinances are long documents, with definitions in one chapter, a use table in another and dimensional standards in a third. That makes them a natural fit for language models trained or grounded on a community's own adopted documents.

  • EuclidHL, from the planning firm Houseal Lavigne and listed as an Esri partner solution, is an AI planning and zoning assistant that lets residents, stakeholders and city staff ask planning questions in plain language. It answers from the city's planning documents, zoning code and GIS data, and can be embedded on a city website.
  • Zone Genie, from Applied Spatial Analytics Platform (ASAP), is described on its Esri partner listing as checking permit submittals against local zoning rules and returning consistent results for staff and applicants.

Tools of this kind are usually set up per community, because the answer depends entirely on which ordinance is loaded and whether it is the current one. When you evaluate one, ask what happens after an amendment is adopted, whether answers cite the section they came from, and how overlays and conditional uses are handled.

General chatbots are a different matter. A general model that has not been given your town's adopted code will answer from whatever it absorbed about zoning in general, and "R-2" means different things in different towns. Use them to understand concepts, not to read a specific ordinance.

AI that analyzes zoning on the map

The map side answers questions about where and how much. These are the questions a council, a housing committee or a comprehensive plan update keeps asking:

  • Which parcels are zoned X? Every parcel in a district, or a group of districts, with a table of addresses, owners and acres.
  • How much land is zoned for multifamily? Total acres by district, and the share of the town's residential land it represents.
  • How much of it is vacant? The same totals, filtered to vacant or underused parcels from the assessor's land use codes.
  • Zoning versus flood risk. Which commercial or residential districts fall inside the FEMA 100-year floodplain, and how many parcels.
  • Zoning near transit. Parcels within a walk of a bus or rail stop that are still zoned single family only.
  • Rezoning impact. If a corridor were rezoned, how many parcels and acres would change, who owns them and how many are vacant.

Researchers have also been mapping zoning across many jurisdictions at once. The National Zoning Atlas, directed by Sara Bronin, maps where single-family homes, apartments and accessory dwelling units are allowed, along with lot size, parking and height rules, and covered roughly 36 states in whole or in part as of August 2024. For a single town, though, the most current source is still the local zoning layer.

Where NeuraCities fits: the map side

NeuraCities is an AI GIS. You ask the zoning question in plain English, and its agents find the zoning, parcel and hazard layers, run the spatial operations and return a map, a table, charts and a written answer with sources. It works with the zoning map layer (the district boundaries and their codes), not the ordinance text. For what a district permits, read the adopted code or use a tool built for it.

A typical zoning question runs like this:

  1. 1The orchestrator declares a checklist: parcels, the zoning layer, any other layers the question needs and the result it will return.
  2. 2The Data Agent searches the city and county open data portals for zoning and parcels, plus federal sources such as the FEMA flood map and Census ACS, and probes each layer for coverage before downloading it.
  3. 3The GIS agent joins zoning to parcels, filters by district code, measures area, and overlays or buffers other layers as asked.
  4. 4The answer comes back as a map, a parcel table you can export to Excel, a chart of acres by district and a summary listing every source.

In a real Nashville run, NeuraCities screened Davidson County's 286,920 parcels for vacant land over 10 acres within half a mile of an interstate and outside the FEMA 100-year floodplain, then joined the county's 5,992 zoning district polygons. The question explicitly asked that each parcel appear once even where it touches several zoning districts, a common trap in zoning joins. Seven parcels qualified:

Land useAcresZoningArea
Vacant rural land41.5AR2AJoelton
Vacant industrial land28.4IRGoodlettsville
Vacant rural land21.3AR2AJoelton
Vacant residential land18.1RS20Goodlettsville
Vacant rural land15.8AR2AWhites Creek
Vacant rural land15.7AR2ANashville
Vacant rural land11.3R10Goodlettsville
Nashville: seven vacant parcels outside the floodplain
Live map from a real NeuraCities run. Drag to pan, use two fingers or Ctrl + scroll to zoom.
Live
The real Nashville run: parcels, the FEMA 100-year floodplain and zoning districts from maps.nashville.gov, interstates from OpenStreetMap. Seven vacant parcels over 10 acres within half a mile of an interstate and outside the floodplain qualified, each shown with its zoning district.

Zoning questions to ask NeuraCities

Each of these has a decision behind it. Replace the city and the district codes with your own:

Which commercial zones in Salt Lake City fall inside the FEMA 100-year floodplain? Show the parcels and total the acres by zoning district.
How many acres in [town] are zoned for multifamily housing, by district, and how many of those acres are on vacant parcels?
Which parcels within a quarter mile of a bus or rail stop in [city] are zoned single family only? Count them and total the acres.
If every parcel fronting [corridor] between [street] and [street] were rezoned to mixed use, how many parcels and acres would change, how many are vacant and who are the largest owners?
Which census tracts in [city] have the highest share of renters, and how much of each tract's land is zoned to allow apartments?

Transit questions follow the same pattern as the Charlotte Opportunity Zones study, which counted 482 bus and light rail stops across 17 zones. Swap the zones for zoning districts and the count becomes "how much single family zoning sits next to transit".

Tip: name the districts
Use the district codes exactly as they appear in your zoning layer. If you are not sure which codes allow multifamily, ask NeuraCities to list the distinct district codes in the layer first, then decide from the adopted code which ones to include and say so in the question.

Caveats for any AI zoning answer

  • AI interpretations are drafts. Whether a code lookup tool says a use is permitted or a map analysis totals the acres zoned for multifamily, the result is a starting point for a planner, attorney or zoning administrator to confirm, not a determination.
  • The zoning layer may lag the adopted map. Rezonings are adopted by ordinance, and the GIS layer on an open data portal is updated afterwards, sometimes with a delay. Check the layer's source and date in NeuraCities, and ask the planning office for the current file if it is old. Your uploaded layer takes priority over the public one.
  • Overlays change the answer. Historic districts, floodplain overlays, airport zones and form-based overlays often live in separate layers. A parcel's base district alone may not tell you what is allowed, so ask for the overlay layers too.
  • Parcels cross district lines. Split-zoned parcels will appear twice in a naive join. Ask for one row per parcel, as in the Nashville example, or for the area of each parcel inside each district.
  • Small towns may not publish zoning at all. If the Data Agent cannot find a zoning layer, NeuraCities says so instead of guessing. Ask the town or county for the shapefile and upload it.

Bringing your own zoning data

Many towns keep their zoning map as a county GIS file or an engineer's export. Upload it as a Shapefile, GeoJSON or File Geodatabase and it is saved to My Database for every future conversation. If you keep district standards in a spreadsheet (minimum lot size or maximum units per acre by district code), upload it too and ask for it to be joined to the zoning layer on the district code. That turns "acres zoned multifamily" into a rough count of units the zoning allows, built from your own table.

Results export as Shapefile or CSV for the GIS team, as Excel tables, and as print-ready maps in PNG, JPG, SVG or PDF. Share the conversation by link with the planning commission, or turn it into a dashboard page for a staff report. Save a zoning question as a workflow with {{variables}} such as {{city}} or {{district_codes}} to rerun it after the next rezoning.

Frequently asked questions

Can AI read a zoning code?

Yes. Zoning assistants such as EuclidHL from Houseal Lavigne are set up on a community's adopted ordinance and planning documents and answer questions in plain language. Treat their answers as drafts and confirm them against the adopted code or with the zoning administrator.

What is AI zoning analysis?

It is using AI to run GIS analysis on the zoning map: finding parcels in a district, totaling acres zoned for a use, overlaying zoning with flood zones or transit, and estimating what a rezoning would change. The AI plans the analysis and GIS operations compute the results.

Does NeuraCities read zoning ordinance text?

No. NeuraCities works with the zoning map layer: the district boundaries and codes, joined to parcels and other layers. For what a district permits, use the adopted ordinance or a code lookup tool built on it.

How do I find which parcels are zoned for multifamily?

In NeuraCities, ask for the parcels in your city whose zoning district is one of the codes that allow multifamily, with a table of acres. It joins the zoning layer to parcels and returns a map and table with sources. Confirm which district codes allow multifamily from the adopted code.

Is the zoning map on my city's open data portal up to date?

Not always. Rezonings are adopted by ordinance and the GIS layer is updated afterwards, sometimes with a delay. Check the layer's date, and ask the planning department for the current file if you need certainty.

Can AI tell me if my project is allowed?

It can give a useful first answer, and some tools check permit submittals against local rules. A final determination still comes from the local zoning administrator, so use AI to prepare the question and the evidence, not to replace the review.

More guides on ai and gis

Ask which parcels are zoned for what, and get a sourced map and table back.

Run a zoning analysis

Have a question about your own data?

Talk to us, or read the common questions.