Crime Data Assistant

Crime statistics come from multiple datasets covering different time periods, detail levels, and offense types. These sources approach crime data differently and can sometimes provide different answers to the same question. The assistant will now automatically pick the best source. More information on sourcing is provided below.

See Trends

Measure Outcomes

Dig into Details

Compare Differences

About this tool & the data sources

What is the Crime Data Assistant?

The Crime Data Assistant is an AI-powered research tool that lets you ask questions about crime in the United States using plain English. Instead of downloading spreadsheets or navigating government websites, you type a question like “How have murders trended in Chicago since 2020?” and get back an interactive chart, a data table, and an explanation — all generated in real time.

Under the hood, an AI model translates your question into a database query, runs it against real crime data, and formats the results. Always remember – I am a bot and may make mistakes from time to time. Verify findings with official sources such as the Crime Data Explorer.

How to Read Results

1

Explanation

A plain-English summary of what the data shows, generated by AI with context about limitations.

2

Chart

An interactive line, bar, or area chart. Hover over data points for exact values.

3

Data Table

Sortable raw numbers. Click column headers to sort. Export as CSV.

The Five Data Sources

Each database covers different time periods, detail levels, and collection methods. Your question is automatically routed to the best source.

Current TrendsHistoricalNIBRSSupplementary HomicideHate Crime
SourceAgency DataFBI Uniform Crime ReportFBI NIBRS Master FilesFBI Supplementary Homicide ReportsFBI Hate Crime Statistics
Time Range2017–present1930–20241991–present1976–present1991–present
GranularityMonthlyAnnualIncident-levelIncident-levelIncident-level
CoverageHundreds of agencies10,000+ agenciesVaries by yearVaries by year~15,000 agencies (voluntary)
Crime TypesPart I UCR (7 offenses)Part I UCR + staffing52+ NIBRS offensesHomicides onlyBias-motivated crimes
Best ForRecent trendsLong-term trendsDeep analysisHomicide analysisHate crime / bias trends

Understanding Each Data Source

Current Trends

This data comes from The Crime Index, a project by AH Datalytics that collects monthly crime data directly from police department websites, the FBI's Crime Data Explorer, and public records requests. It covers the 7 traditional Part I UCR crime categories: violent crimes (murder, rape, robbery, and aggravated assault) and property crimes (burglary, theft, and motor vehicle theft). Note that some reported data may be preliminary, subject to change, and may not precisely match figures that will eventually be formally reported to the FBI.

Why it exists: The FBI publishes annual crime estimates with a delay. This dataset fills the gap by providing near real-time monthly data from hundreds of the largest police departments to establish local, statewide, and national crime trends. It's the best source for answering “what's happening right now?”

What “national sample” means: Several hundred agencies report consistently every month. When you ask for “national” trends, the assistant uses only this consistent sample so that month-to-month changes reflect actual crime changes, not agencies dropping in and out of the data.

Historical Library (FBI Uniform Crime Reports)

The Uniform Crime Reporting (UCR) program is the longest-running crime data collection in the U.S., dating back to 1930. The FBI collects annual crime counts from law enforcement agencies nationwide and publishes them as annual reports.

Why it exists: This is the gold standard for historical crime analysis. If you want to know how crime has changed over decades — or compare a city's crime rate in 1990 vs. 2020 — this is the dataset to use.

Key caveat: The data you see here reflects what agencies submitted to the FBI though not all agencies have reported in all years, especially before 1960. Additionally, not every crime is reported to law enforcement every year.

NIBRS (National Incident-Based Reporting System)

NIBRS is the FBI's modern, incident-level crime reporting system. Instead of just reporting annual totals, agencies submit detailed records for every individual crime incident — including information about victims, offenders, weapons, drugs, property, injuries, and how the case was resolved.

Why it exists: The older Summary Reporting System only captured aggregate offense counts. NIBRS captures a fuller story of each incident. This makes it easy to answer questions like “how many vehicle burglaries were there in Texas last year?” or “what percentage of carjacking victims are male?”

Key caveat: NIBRS participation has grown slowly. In the early 1990s, only a handful of states reported. By 2021, the FBI required all agencies to switch to NIBRS, and coverage is nearly 90% of the U.S. population. When looking at trends over time, always consider whether an increase reflects more crime or simply more agencies reporting. An agency reporting an increase in crimes from one year to the next may have simply started reporting via NIBRS in the middle of a year.

What it can't do: The database has total jurisdiction population but not population broken down by race, gender, or age. This means you can calculate per-capita crime rates for a city or state, but you cannot calculate rates like “murders per 100K females” because the female population isn't available.

Supplementary Homicide Reports (SHR)

The Supplementary Homicide Reports (SHR) are a specialized FBI data collection focused exclusively on homicides. Since 1976, agencies have submitted detailed records for each murder and manslaughter — including victim and offender demographics, weapon used, victim-offender relationship, and the circumstance of the killing.

Why it exists: While NIBRS and UCR count homicides, the SHR provides richer detail about each killing. You can answer questions like “what share of murders involved a firearm?”, “how has domestic homicide trended since 1990?”, or “what is the age breakdown of murder victims in Texas?”

Key caveats: Not all agencies report SHR data every year. Most notably, Florida did not submit SHR data for most years between 1997 and 2020. Alabama and some other states have inconsistent reporting. National counts from SHR will undercount actual homicides.

About offender data: Approximately 40% of homicide incidents have no recorded offender information. Having offender demographics recorded does not mean the case was solved or cleared — this dataset does not measure clearances. When viewing offender breakdowns, results reflect only cases where offender demographics were recorded at the time of reporting.

Common Pitfalls

Crime data is nuanced. Additionally, I am a bot and may sometimes make mistakes. Keep these in mind when interpreting results and always verify findings with official sources if you're uncertain.

  • Data is preliminary and collected directly from police departments — it has not been validated by the FBI and is subject to revision.
  • National figures use a fixed “national sample” of several hundred agencies. This tracks trends but does not represent a national total.
  • An agency not appearing in the data may simply not be in the sample — it doesn’t mean they have zero crime.
  • Crime categories follow the traditional UCR Part I structure (murder, rape, robbery, aggravated assault, burglary, larceny, and motor vehicle theft).
  • Reporting is voluntary — not all agencies reported in all years, especially before 1960.
  • The FBI changed the definition of rape in 2013 (from “forcible” to a broader definition). Comparing rape rates across that break requires caution.
  • Data reflects only crimes reported to police. Unreported crimes are not captured.
  • NIBRS participation varied wildly by state and year, especially before 2021. An increase in crime counts may reflect more agencies reporting, not more crime.
  • The database has total jurisdiction population only — not population by race, gender, or age. Per-capita rates by demographic group cannot be calculated.
  • Queries on raw tables (100M+ rows) can be slow. The assistant uses pre-built summary tables when possible, but very specific queries may take longer or time out completely.

Tips for Better Results

Be specific

“Murder rate in Houston 2020–2024” beats “crime in Texas.”

Try different tabs

Each tab draws from a different database. If one can’t answer your question, another might.

Specify a time range

Say “2019 to 2024” instead of “recently” for more relevant results.

Ask follow-ups

The assistant remembers context. Try “now show that as a rate per 100K.”

Built by AH Datalytics. Data is provided as-is for research purposes. Always verify critical findings against the FBI Crime Data Explorer. Not affiliated with or endorsed by the FBI.