625 ILCS 5/11-601(c) names one of them.
The API returns exactly that unit, with its own text, the text it completes, and where it sits in the section.
No second model call and no parsing on your side.
pinpoint object beside it:
Where it works
The price does not change.
A resolve with a pinpoint is the resolve price, a body with a pinpoint is the body price, and
includePinpoint adds nothing to a search.
A pinpoint that finds no such unit is still a resolved section and is charged like one.
What a pinpoint can look like
The pinpoint is read from the end of the citation. Labels are parenthesised, outermost first, and a label that contains a hyphen or a dot is one label.
Hyphenated and decimal labels are never split.
(f)(1.5) is the unit 1.5 inside (f), not the unit 5 inside 1 inside (f).
Ranges and lists ((a)-(c), (a), (b)) resolve to the first unit.
The block says so in kind, and lists the others in also.
Ask for each one with GET /us/statutes/section/{actId}/body?pinpoint=....
State-specific forms are read too, for example Minnesota subd., Kentucky Section 8(2)(a), New Hampshire II(a)(1), Wisconsin letter inserts such as (am) and (1m), and Spanish forms for Puerto Rico.
The older subsection string on the section is unchanged.
It is dot-joined (f.1.5), so it cannot tell a unit 1.5 from 1 then 5.
Read pinpoint.path when you need the unit itself.
Always read matched
A pinpoint can name a unit the section does not have.
The call still succeeds, because the section resolved.
The block says what happened instead of answering with the whole section.
On a miss,
available lists the labels that do exist at the level where the lookup stopped, so a mistyped pinpoint can be corrected without another call.
reason is no_such_subsection when the outline is known and has no such unit, and no_structure when we could not recover the section’s outline.
In the second case the section text is still at /body.
If the pinpoint could not be looked up at all (the section text was unavailable, or the lookup ran out of time), the response carries pinpointNote instead of pinpoint.
The section itself resolved and is charged as usual, and retrying usually answers.
How far to trust it: structure
Every pinpoint says where its outline came from and how far to trust it.
We never present a guess as exact.
A section whose outline we cannot recover is reported as
none, not as a section with no subsections.
structure.source is artifact when a pre-built outline of this section was used, which is only done after its hash matched the exact text we serve, and request_time when the outline was parsed from the text on this request.
structure.textSha256 is the SHA-256 of the plain text that charStart and charEnd index into.
Hash your copy of plain to confirm the offsets apply to it.
Coverage varies by state and by section.
Long definitions sections and sections with tables are the most likely to be heuristic.
Check matched and structure.quality on every call and treat them as part of the answer.
Search: which subsection matched
Search ranks whole sections. WithincludePinpoint: true, each hit also says which subsection it matched, in a pinpoints object keyed by actId.
located, pinpoint ((c)(1)), a citation you can paste back into resolve, the unit’s text, the leadIn it completes, and its character offsets.
matchedBy is passage when the unit lies inside the passage the search matched, and query when the unit was chosen by your query’s words alone, which is what later pages of a query get.
When several sub-units matched equally well, scope is parent and the block names their common parent instead of choosing between them.
A hit that cannot be named says why in reason (no_single_unit, no_structure, and so on).
That is a verdict about the hit and not a failure.
It never changes which sections are returned, their order or their scores, and the search is charged the same either way.
Coverage by state
Coverage varies by state, so this table gives the measured rates where we have them and says where we do not. The percentages come from a full build of outlines over every section of the state, graded bystructure.quality as described above.
As of 2026-10-26.
Those percentages measure the build, not a given call.
Where the outline for a response came from is in
structure.source, and the grade is in structure.quality.
Read both on every response.
- Illinois. In the build, 68,219 of the 72,251 sections have an outline taken from the publisher’s own markup.
Sections that hold several versions of the text in one record, and long sections whose enumerations run inside a sentence, are the likeliest to be
heuristic. Inserted labels such as(d-1)and(1.5)are kept as single labels. A call can instead parse the outline from the section text, and that parse is graded on its own, so readstructure.quality. - Minnesota. In the build, 26,325 of the 27,747 sections have an outline taken from the publisher’s own markup, and a call can use a pre-built outline of the section.
A subdivision may open at
(1)without a paragraph (a) before it. - Missouri. The outline is parsed from the text of the section, one paragraph per line.
A definitions section has no numbered subsection level, so in
Mo. Rev. Stat. 556.061(1)the unit (1) sits directly under the section. - Wisconsin. The outline is parsed from the text of the section.
Letter inserts such as
(1m),(2g)and(am)are each one label, and repealed provisions leave gaps in the sequence, which is normal. - Iowa. The outline is parsed from the text of the section, where each paragraph is printed with the labels of the units above it.
A unit that only introduces numbered paragraphs can have an empty
text, with the content in itschildren.