Skip to content

Know how your buildingbehaves in an earthquake

A few photos, your location, and a couple of details — that's all QuakeWise needs to give you a clear, honest read on the shaking your home faces and how its structure is likely to respond.

No fabricated scores. Every number is traced to its source.

Grounded in GEM hazard data + TBDY-2018 / FEMA P-154

Built on data you can check

We don't invent risk. Each estimate is grounded in published hazard models and recognized engineering codes — and we show our sources.

GEM global hazard

Peak ground acceleration comes from the Global Earthquake Model's published hazard, not a guess.

TBDY-2018 & FEMA basis

Structural scoring follows the Turkish seismic code and FEMA P-154 rapid-screening logic.

Named provenance

Every figure shows where it came from. If data is missing, we say so — never a placeholder.

How it works

From a point on the map to a structural read

A guided flow that asks only what it needs, revealing more as it learns about your building.

  1. Locate

    Share your location or drop a pin. We pull the seismic hazard for that exact point.

  2. Capture

    Add a photo or two. Vision AI reads the structure type, era, and visible condition.

  3. Describe

    Confirm a handful of building details. The form auto-fills from what we already know.

  4. Analyze

    We combine hazard and structure into a per-magnitude damage read with confidence bands.

  5. Act

    Get the magnitude your building is safe up to, plus ranked, practical recommendations.

Five clear stages — each one unlocks only when the previous is ready.

What we assess

The structures we model

Your building is matched to a standard HAZUS Model Building Type — the engine carries published fragility curves for reinforced concrete, steel, masonry and timber systems, and nothing outside that set.

Illustration of a reinforced-concrete frame building
C1 · C2 · C3

Reinforced concrete

Moment frames, shear-wall systems, and concrete frames with masonry infill — the dominant stock across Turkish seismic zones.

Illustration of a steel-framed building
S1

Steel

Steel moment-frame structures, modelled with their own ductile fragility set.

Illustration of a masonry building
URM · RM1

Masonry

Unreinforced bearing-wall masonry (the most vulnerable) and reinforced masonry, each with distinct fragility.

Illustration of a timber-framed building
W1

Timber

Wood light-frame construction — the most ductile and least vulnerable of the modelled systems.

Building types follow the HAZUS / FEMA P-154 taxonomy used by the engine. If your structure falls outside this set, the result says so rather than guessing.

What you get

A damage curve, not a vibe

See expected performance across earthquake magnitudes, with an explicit safe-up-to-M threshold.

Ranked recommendations

Concrete, prioritized actions — from inspection to retrofit — matched to your building's weak points.

Honest by design

Confidence bands on every result, and an explicit 'unavailable' state whenever data is missing.

How the science works

Risk = hazard × exposure × vulnerability

No black box. The engine composes three published, citeable layers, then states every approximation it makes along the way.

01

Hazard — the shaking at your exact point

We read the peak ground acceleration (PGA) for your coordinates from the Global Earthquake Model hazard map — the level with a 10% chance of being exceeded in 50 years (a 475-year return period). Local soil conditions are then applied with TBDY-2018 site factors, because soft ground amplifies shaking.

Source: GEM Global Seismic Hazard Map 2023.1 · TBDY-2018 site factors

Diagram of soil classes used in seismic site amplification
02

Exposure — what your building is, and its weak points

Vision AI and a short form establish the structural type, era, and visible deficiencies. A soft (open) ground storey, vertical or plan irregularity, or a steep slope each shift the building toward a more vulnerable fragility set — exactly the features FEMA P-154 rapid screening looks for.

Basis: HAZUS Model Building Type + FEMA P-154 screening features

Diagram of a soft-story building deficiency
03

Vulnerability — a fragility curve, not a single score

Each building type and design level has published HAZUS lognormal fragility curves. Evaluated at the site shaking, they give the probability of each damage state — and as the magnitude rises, that probability climbs along the curve. A weak pre-2000 building reaches high damage probability far sooner than a modern code-conforming one.

Basis: HAZUS-MH lognormal fragility (published medians, β ≈ 0.64)

Diagram of escalating earthquake damage levels
04

Fusion — combine the evidence, honestly

The engineering prior is combined with any further evidence through a reliability-tempered Bayesian step: a signal we trust less is weighted down, never blindly believed. Every result carries an uncertainty band, and where we have not calibrated locally, that band is widened rather than hidden.

Method: reliability-weighted Bayesian late fusion · always with a confidence band

Diagram of vertical structural irregularity

Four damage states

No damage
Minor
Major
Destroyed

Every estimate is expressed over these four states — the shared damage ontology used end to end. The output is a probability across them, with a band, never a single false-precise number.

Two methods, cross-checked

Every assessment runs through a transparent physics engine — and, in beta, AI models that cross-check it. We never hide which produced a number, or how confident it is.

Primary

Physics engine

Site hazard from the GEM 2023 global seismic model — the ground shaking with a 10% chance of being exceeded in 50 years — combined with published HAZUS fragility for your building type by reliability-tempered Bayesian fusion. Deterministic, inspectable, every number traced to its source.

Beta

AI models

A vision language model reads your building photos to suggest structural attributes — live today. A neural model trained on post-earthquake building data is being added as a beta cross-check alongside the physics engine; by design it only ever informs our calibration and never overrides the validated physics result.

Beta

Hands-off voice

Speak your way through the assessment: QuakeWise reads each step aloud and listens for your answer, in natural ElevenLabs voices. The only things it needs from you are location permission and your photos — it fills in the rest, and you confirm.

0.81≈ 81% likelihood

A calibrated probability, not a letter grade

QuakeWise answers with a probability between 0 and 1. A reading of 0.81 means an estimated 81% likelihood of that damage level under the earthquake scenario you chose — calibrated against real past earthquakes and shown with its confidence band, so the number means what it says.

No fabricated scores. Every number is traced to its source — and beta features are labeled, never dressed up as proven.

See where your building stands

A first read takes a few minutes. No account required to start.

Start your assessment

Informational estimate — not a replacement for a licensed engineer.