GovBot Earthquakes & Landslides

ChatGPT plugin in preparation — not published. Its independent no-auth service is being prepared and verified for client testing. Hosted transport checks and actual ChatGPT acceptance are separate steps.

Dated, sourced briefs from public earthquake and landslide records

GovBot Earthquakes & Landslides is an independent app being prepared for ChatGPT. Ask about a U.S. state, a public research point and radius, or a USGS event ID. These workflows are designed to help produce a brief that notes sources and retrieval times for the figures it can trace, and identifies evidence that is unavailable or unknown in the records. The plugin has not been published in ChatGPT; client testing and release review remain separate from source and transport checks.

What you can prepare

Earthquake catalog brief

Counts, magnitude bands, decades and largest events for a state, or for a research point and radius, over a UTC date range. Event rows carry USGS IDs, UTC origin times, magnitudes with their magnitude types, and links to USGS event pages. The total matching count is shown next to the number of rows actually returned.

Selected-event shaking explanation

For one USGS event: the catalog origin, ShakeMap maximum intensity and ground motions with units exactly as published, and voluntary Did You Feel It? responses — kept apart so you can see what each kind of evidence does and does not mean.

Landslide inventory brief

Mapped landslide inventory records within 0.1–25 km of a research point, with contributing inventories, confidence classes and occurrence date ranges. Record counts are shown alongside the number of rows listed.

Source, date and unit limits

Read these limits before relying on any brief. The workflows are designed to help produce sourced briefs with unavailable evidence identified, but they do not guarantee that every figure or answer is complete, current, or fully attributed.

Dates are not interchangeable
Earthquake times are in UTC. An event’s origin time, its source update time and the time GovBot retrieved it are different dates, and briefs label each one. Ordinary results may come from a cache up to 10 minutes old; boundary, layer and ShakeMap metadata up to 24 hours old.
Magnitudes carry a type
Each magnitude is reported with its type, such as Mw, ml or mb. Different magnitude types are not directly comparable.
Significance is an index
The USGS significance index ranks published events. It is not a magnitude threshold, a damage estimate, or a probability of a future earthquake.
Units are shown exactly
Peak ground acceleration (PGA) and velocity (PGV) appear with the units in the ShakeMap metadata — labels such as g, %g and cms are not relabeled or converted. If units are not published, they remain unknown and the value is not interpreted.
Shaking evidence is regional
Did You Feel It? responses are voluntary reports, not a census of who was affected. ShakeMap values are regional estimates combining models and recordings. Neither is a measurement at a particular building.
Landslide records are records
The USGS inventory compiles many sources. Point and polygon records can overlap, confidence classes describe mapping evidence, and dates are often ranges. A record — or the lack of one — does not establish site risk or stability.
Missing is not zero
Catalogs and inventories are incomplete, especially for older and smaller events. A zero count, an empty field or a source error is not evidence that there is no hazard.
Not a warning service
No predictions, current warnings, or building, site, geotechnical, legal or insurance determinations. For current alerts, follow official agencies and local authorities.

What to share — and what not to

The tools accept a U.S. state, or public research coordinates (WGS84 latitude and longitude) with a radius in kilometers, plus UTC dates, a minimum magnitude, a landslide confidence class or a USGS event ID. There is no address lookup and no device location.

You do not need to share your home address or precise personal location. A city-center or other public reference point is enough for radius research.

Example prompts

  1. Prepare a historical earthquake catalog brief for California from 2000-01-01 to 2024-12-31 UTC, magnitude 4.5 and above: total count, magnitude bands, counts by decade and the five largest events with their magnitude types and USGS links.

    Earthquake catalog brief · state scope

  2. Explain USGS event ci38457511. Keep the catalog origin, the ShakeMap values with their exact units and product version, and the Did You Feel It? response count separate, and tell me what the records leave unknown.

    Selected-event explanation · USGS event ID

  3. Using the public research point 47.6062, -122.3321 (WGS84, Seattle city-center reference) with a 5 km radius, prepare a landslide inventory brief with point and polygon record counts, the nearest records, their confidence classes, date ranges and contributing inventories.

    Landslide inventory brief · public research point and radius

These are example requests, not recorded results. No demonstration has been published.

Tools

search_earthquakes_near_point
Total matching count and one ordered page of catalog events around a research point and radius.
summarize_earthquakes_for_area
Whole-area counts, magnitude bands, decades and largest events for a state or a research point and radius.
get_earthquake_details
Origin, magnitude, review status, ShakeMap and Did You Feel It? summaries for one USGS event ID.
list_significant_earthquakes
Events at or above a USGS significance index, worldwide or for a state.
search_landslides_near_point
Inventory record counts and nearest records around a research point.

Independent test endpoint: https://earthquakes-api.govbot.ai/mcp — public HTTPS readiness is verified separately before client testing.

Sources

GovBot is independent. It is not affiliated with or endorsed by the U.S. Geological Survey, the Census Bureau, any other government agency, or OpenAI. Source data remain attributed to their publishers.

Contact

Questions or corrections: hello@govbot.ai