How it works

AeroLeads scans satellite imagery of an area and uses AI vision to find properties with pools, solar panels, or landscaping in poor condition — then turns each find into a street address. The result is a targeted lead list for home-service businesses: pool care, solar maintenance, landscaping.

How a scan works

1. 1

   Pick an area

   A zip code, or a rectangle you draw on the map.

2. 2

   Fetch imagery

   The area is tiled into high-resolution satellite images.

3. 3

   AI detection

   Claude scans each tile against a labeled grid and names the cell holding each feature.

4. 4

   AI verification

   Every detection's close-up crop gets a second, independent AI check.

5. 5

   Street addresses

   Verified detections are matched to the nearest parcel's address.

What the labels mean

✓ verified

Every lead passed a second, independent AI check: the close-up satellite crop of the detection was re-examined with a stricter prompt, and only confirmed detections become leads. The note next to the checkmark is that check's own explanation.

Property type chips

The verification step also classifies the property. Residential means a single-family home; Multi-family means an apartment, condo, or townhome complex. The two can be hard to tell apart from above (an apartment tower and a hotel look alike), so treat the distinction as approximate. Public pools, swim clubs, schools, and parks are classified Excluded and never become leads. Which types become leads depends on the feature — solar leads are homeowner-only, while pool and landscaping leads include complexes and businesses.

Confidence percentages

The verification check's own certainty about the feature (not the initial detection's). Leads below 70% are dropped before you see them.

Evidence thumbnails

The actual satellite crop the AI verified — the region of imagery containing the detection. Click one to enlarge. If the thumbnail doesn't show what the note claims, don't trust the lead.

Coverage and cost

A scan covers an area tile by tile — each tile is one satellite image about 170 m across, and each costs real money to fetch and analyze. Bigger coverage means more tiles, which means more cost and more time. The presets trade these off: Quick covers the central blocks of a zip, Standard the core neighborhoods, Deep the widest preset area, and Full zip every tile in the zip. Drawing an area scans exactly the rectangle you outline — often the cheapest option when you know the streets you care about.

During beta, the dollar figures shown are live estimates of the underlying API cost (imagery + AI analysis) — not a price.

Honest limitations

• Detection quality depends on the satellite imagery — its recency, season, and resolution. A pool installed last month may not be in the imagery yet.
• Addresses come from Google's geocoder. Detections near property lines are assigned to the nearest address point or dropped when it's ambiguous, but an address can still occasionally be off by a lot number.
• Accuracy is validated manually against satellite imagery. (Validation of the current pipeline is in progress; measured precision will be published here.)