How Weverit Reports Are Made

Last updated: May 2026

Weverit publishes two kinds of content: building reports and news articles. They're produced differently, and we want to be transparent about both.

Building reports

Every report on a NYC residential building is generated automatically from public records. We don't write these — we organize what's already public into one place.

Where the data comes from

All data is sourced from official NYC, state, and federal public records. We re-fetch each source on a schedule matched to how often it changes — typically daily to monthly.

What we add

We don't generate the underlying numbers. They come directly from public records. What we add is:

1. Aggregation — pulling data from multiple official APIs into one page so you don't have to query each portal separately

2. Plain-English summary — a short AI-written paragraph that describes what the data shows, generated from the actual numbers on each report

3. Composite grade — a letter grade computed as a weighted average of seven dimensions, each scored 0–100 from the underlying records:

   • Health (30%) — open HPD violation count and severity class mix
   • Safety (25%) — recorded incident counts near the building
   • Rent stabilization (10%) — whether DHCR records list stabilized units
   • Schools (10%) — presence of nearby public elementary schools
   • Price (10%) — listing rent vs. ZIP median, when a listing is provided
   • Parking (8%) — count of off-street parking facilities within walking distance
   • Affordability (7%) — area income vs. typical rent burden

   Dimensions with no available data are dropped from the average and the remaining weights re-normalize. Score bands: ≥92 = A, ≥85 = A−, ≥80 = B+, ≥70 = B, ≥60 = C+, ≥50 = C, ≥40 = D, below 40 = F.

4. Source links — every dataset section in the report links back to the originating record

What we don't do

• We don't manually edit individual building reports
• We don't suppress or downweight findings based on landlord requests
• We don't accept payment to remove or modify reports
• We don't sell, share, or otherwise transmit your search history

News articles

Some articles on Weverit are written by humans. Some are drafted by AI and edited by a human before publication. Either way, the editorial standards are the same.

How AI-assisted articles are made

1. Topic identification. A topic is identified — either from patterns in the data we observe, or from current housing news relevant to NYC tenants.
2. Data extraction. We query NYC public records for the specific numbers, addresses, and patterns the article will discuss.
3. AI drafting. Anthropic's Claude API generates a first draft based on the extracted data and our editorial style guide. The AI is configured to use only the numbers we provide — it cannot invent statistics or attribute fabricated quotes.
4. Human editing. A human editor reviews the draft, verifies every number against its source, fixes prose, removes any AI over-confidence, and adds context the AI missed.
5. Source verification. Every numerical claim is checked against its public record before publication.
6. Methodology paragraph. Each AI-assisted article ends with a short paragraph describing what data was used and how.

What AI does and doesn't do at Weverit

AI does:

• Synthesize public records into readable prose
• Suggest article structure and section headings
• Draft opening paragraphs and transitions
• Format data into tables and lists

AI does not:

• Generate numbers (all figures come from public datasets)
• Invent buildings, addresses, or owner names
• Quote real people unless the quote is from a verified public record
• Make legal claims or accusations
• Decide what stories get published

Editorial principles

• Every numerical claim links to its public source
• Methodology is disclosed in every AI-assisted article
• We correct errors publicly, not silently — if a published article contains a factual mistake, it's corrected with a visible note
• We don't publish articles that we haven't independently verified
• We don't publish to please landlords, brokers, advertisers, or any party other than NYC tenants reading the article

Corrections

If you find an error in a building report or a news article, write to [email protected] with the URL and the correction. We'll respond within 48 hours.

Why we publish methodology

Most AI content on the web doesn't say it's AI. Most data aggregators don't say where data came from. Most publishers don't say how they edit.

We do, because tenants making decisions about where to live deserve to understand how the information they're reading was produced.