Ask

ATLAS turns it into an explicit query, runs that query in plain code, and shows you both — so you can see exactly which question was answered and correct it when the reading is wrong. No number on this page is written by a language model.

What Ask can answer

How this differs from a chatbot

• Your question is turned into a small, fixed query object. That object is shown to you every time.
• The query is executed by ordinary code over the same data the rest of ATLAS uses. A language model never computes a figure and never chooses an answer.
• If the question cannot be mapped to something in the data, ATLAS says so and lists what it can answer. It does not answer a nearby question instead.
• Every result carries the source and the geography the figures are actually published at.

What this does not show

• The interpreter matches keywords. It does not understand the question, and it can read one wrongly — which is exactly why the query it built is shown to you every time, and why correcting it is part of using this page rather than a failure of it.
• It answers only from the fields listed above. Asked anything else, it declines and says so rather than answering a nearby question it can handle.
• It cannot answer "why". It retrieves published counts and rates; it establishes no cause, and two measures moving together here is not evidence that one drives the other.
• The answers are measurements, not assessments. They do not establish need, service capacity, availability, access or eligibility for anywhere they describe.
• Results are drawn at the geography each source publishes. A figure attached to a county is one measurement for that whole county, not a figure for any division inside it.