What you get

Eight, but not all of them everywhere. A screen is only a screen where the data behind it exists, so a project type whose hard rules measure something a county does not publish is marked unavailable there rather than run against nothing.

Available in every county we cover (37 assessed): General non-residential, Battery storage (BESS), Data center, Industrial / manufacturing, Exploratory (no hard screen).

Available in some counties — the number is what it is today, and rises as sources are added:

• EV fast charging — 8 of 37 counties
• Small distribution solar — 10 of 37 counties
• Large distribution solar — 10 of 37 counties
• Utility-scale solar — 35 of 37 counties

DC fast charging (Level 3). The screen is built around highway-corridor demand — it requires a count location carrying at least 10,000 vehicles a day within three miles, and a measured distance to a mapped lower-voltage line, because a fast-charging site needs real service capacity rather than a wall outlet. That distance is scored rather than screened: two miles is the figure this profile is built around, and a site beyond it ranks lower instead of dropping out.

It is not a Level 2 screen. Workplace, multifamily and destination charging go almost anywhere there is existing parking and a panel with headroom, so the questions that decide those sites are electrical and contractual rather than locational — which is not what this tool measures, and we would rather say so than let you read a corridor result as advice about a parking garage.

Those screens have hard rules that measure distance to a lower-voltage line, or traffic volume on the nearest highway. We do not use utility circuit geometry for the first — almost no utility publishes it — but a national public dataset of lines at or below 69 kV, which we treat as a location proxy and label as one. Its coverage is real but uneven: some counties have dozens of such lines, some have none at all, and a county with none cannot support those screens however good the rest of its data is.

Traffic counts are the other half, and those are published state by state rather than nationally — there is no single national source, so each state is wired separately.

Where the layer is missing, every parcel would come back “review required”, which is a page of non-answers. We would rather say the screen cannot run here than sell you a result that means nothing.

Yes, and unevenly, because the sources are unevenly available. Traffic counts are being added state by state from each state DOT. Gas and water evidence has a national and a service-area path. Fiber routes have no free national dataset at all, and distribution-circuit hosting capacity is published by individual utilities or not at all — for some markets it genuinely may never exist.

The counts above are the live state of it. If a project type shows a number below the total, that is the work remaining, not a rounding.

That we could not rule the parcel in or out on the evidence available — not that it failed. Unknown is never treated as favourable anywhere in the scoring, so a missing input holds a parcel for review instead of quietly passing it. A county with thin data produces more of these.

Each county is rebuilt from its publishers’ live services; Coverage shows per-county build state and the scoring version applied. Parcel rolls and zoning are typically annual, published on each county’s own cycle, so a very recent split or rezoning may not have reached the source we read.

Yes, and it is the part most screening tools do not do. Counties are re-imported as their publishers republish, and every site saved to a project is re-screened against the new build rather than left as the answer you got the first time. When something that mattered moves — a rezoning, a parcel split, a constraint layer that now covers land it did not — the change is recorded against the site and shown in the project, with what it was and what it is now.

A result you are looking at is computed when you open it, not read back from the last time somebody ran it. That is also why a site can stop qualifying: nothing here is carried forward in order to protect a conclusion you already saw.

It is not an interconnection study, a title search, an environmental review, or legal advice on what a jurisdiction will permit. Distances to lines and substations are locational proxies, not statements about available capacity — see the Site-Screening Disclaimer. It is meant to decide which parcels are worth a week of real diligence, and to show its evidence so you can check the reasoning.

Not yet, and the honest answer is worth more than an optimistic one. What exists today is the foundation: the screening engine, the scoring rules and the memo logic are shared code that both this website and the built-in assistant call — the interface is a client of the engine, not the place the thinking lives. Opening that same engine to a customer’s own agent is a deliberate design goal, and new capability is built as an action both a person and an authorized program can invoke.

It matters because of what a result already is here. A screened parcel is a structured record rather than a page to read: a recommendation, the risks that could end the site, what is verified, what remains unknown, and why — each with its source and date. Unknown is a value, never a blank or a zero, which is what makes the output safe for software to act on rather than something a person has to interpret first.

What is missing is access, and it is missing on purpose. A programmatic interface should not be handed out before it is authenticated, metered and reliable, so there is no key to request today and no date promised here. What the surface looks like is written down in full — the endpoints, the response envelope and the refusal rules are in the API and MCP reference, which describes it without offering a way onto it. If you are building something that would use it, the suggestions page reaches us — we would rather hear the use case while the surface is still being shaped than after it is fixed.

Coverage is added county by county, driven by what people ask for. Tell us the county and the project type on the suggestions page and we will tell you honestly whether the data exists to support it — for some markets it does not, and we would rather say so than add a county that answers nothing.