Docket ReportLive through Oct 6
The Roadless Rule record: what half a million comments say
The Forest Service asked the country whether to rescind the 2001 Roadless Rule. 504,587 comments answered. Nearly a third came through 137 form-letter campaigns, and 98.4% of those copies say keep the rule. This is the full record, read by machine and checked by hand, with the count updated until the docket closes on October 6.
1. The count
504,587 comments, in two windows a year apart, and the last week still open
Docket FS-2025-0001 has taken comments twice. The first window was the scoping notice in the fall of 2025: 223,862 comments posted between Sep 1, 2025 and Sep 20, 2025. The second opened when the proposed rule and its draft environmental impact statement posted on Aug 20, 2026. It was due to close on Sep 21, 2026; the agency extended it to Oct 6, 2026. That window holds 267,405 comments so far.
Comments per day as posted by regulations.gov. The agency posts in batches, sometimes days after a comment arrives, so a tall bar is a posting day, not a day the public wrote more. The 2000 rule drew 1.6 million comments and 95% support; that figure counted everything received, including bundled signatures, so it isn’t the same measure as a posted count and we don’t set the two side by side as equals.
How we know: the platform walks the docket every twenty minutes through the comment window and stores every comment with its text and attachments. Method.
2. Campaigns and unique voices
137 form-letter campaigns account for 30.8% of the record
A campaign here is a group of 20 or more near-identical comments, the kind an organization drafts and its members send. The platform finds them by text, not by sender, so a campaign is counted whether or not anyone signed it. 137 such groups hold 151,319 comments. The 18 biggest, each over a thousand copies, hold 138,012 between them (the twelve largest are in the table below), and the largest single campaign runs to 26,941.
The other 339,948 comments are outside any campaign. Under the Administrative Procedure Act the agency may answer a campaign once; it has to answer every substantive unique comment. That is why the second number matters more than the first.
The twelve largest campaigns. Each excerpt is the opening of the campaign’s exemplar comment, quoted from the stored text and linked to it on regulations.gov. The full table of 137 is in the download.
How we know: exact duplicates by hash of the normalized text, near duplicates by MinHash over five-word shingles, verified at an estimated Jaccard similarity of 0.8 or higher. Method.
3. Which way the record leans
Of 149,564 campaign copies we can read a position from, 99.6% say keep the rule
We can state stance with confidence where the text is shared: every copy in a campaign says what its exemplar says. 148,936 campaign copies ask the Forest Service to keep the 2001 rule. 628 ask it to rescind, all of them from 2 campaigns, the largest 404 copies. 936 copies are one-liners that don’t name the rule or the proposal (“Save our forests.”), and 819 are comments whose whole text is an attachment. Counting every campaign copy, readable or not, the keep share is 98.4%, the figure in the tape at the top; the 99.6% here is the same count over the copies that state a position.
For the 339,948 comments outside campaigns the honest answer is: not yet. The platform has classified 96,001 individual comments so far, and among them 97.2% support keeping the rule and 1.4% support rescission. But those are the earliest comments in the record, the slice organized mailing lists dominate, not a random sample, so we don’t publish that share as the docket’s. A stratified sample of the unique comments is scheduled once the window closes, and this page will carry the result with its margin.
How we know: campaign stance comes from the exemplar’s machine classification where one exists (115,010 copies) and from a hand read of the 36,309 copies whose exemplars had none, read Sep 28, 2026 by the desk. Method.
4. What the comments argue
Water, wildlife and wildfire on one side; wildfire again on the other
Across the 96,001 comments the platform has classified, these are the arguments made most often, with the number of comments that make each one. The rescission side is small enough to list in full: 1,310 classified comments, and their case is almost entirely fire.
The case for rescission, as the record makes it (1,310 classified comments):
How we know: each classified comment carries the arguments it makes, extracted by model and stored as labels; the counts are over comments, so one comment making three arguments counts once for each. This band covers the classified slice only. Method.
5. Who filed on the record
The agency published no names, so the record’s organizations come from the letters themselves
Regulations.gov carries no filer name or organization for any comment on this docket; the Forest Service anonymized the field at the source. What can be read is the letterhead: the organizations that name themselves in the opening of a comment or an attachment. A name-match pass on September 28 found mentions, not filers: a group cited inside a member’s comment is not a filer. The letterhead pass that tells the two apart runs next, and this band fills with names and links when it lands. Until then the record’s only named side is the rescission campaign below.
The rescission side’s own words. The largest campaign asking the Forest Service to rescind runs to 404 copies and opens:
How we know: a dated pass over the first 1,500 characters of every comment and attachment in the 2026 window, matched against a list of organizations the desk expected on this docket, as of Sep 28, 2026. A group that filed under another name isn’t in it; tell us and we’ll add it. Method.
6. Where the comments come from
Only 6.4% of comments say what state they’re from, and Montana leads the ones that do
The docket carries no state field for any comment. The platform reads a state only when the writer names one in the text (31,317 comments so far), so this is a distribution over the comments that volunteer it, not over the record.
How we know: the geography label on each classified comment, over the classified slice. Method.
7. The attachments
15,921 files on 5,023 comments, 170,571 pages of them
The substantive record, the science, the legal briefs, the county resolutions and the tribal letters, mostly arrives as attachments. The platform fetched every file in the 2026 window: 5,941 yielded text directly, 3,768 were scans or photographs, and 580 of those have been recovered by optical character recognition so far. 6,206 are formats the platform doesn’t read, mostly images of handwritten letters, and 6 failed to fetch.
How we know: every attachment’s file, page count and text status are on the platform. Method.
Method and limits
What we read, how, and what this page doesn’t claim
- The record. Every comment and attachment on regulations.gov docket FS-2025-0001, walked through the public API every twenty minutes during the comment window, stored with its text. Coverage is stated on the page against the docket’s own count; the platform held 97.4% of the docket when these figures were computed. The remainder is comments the agency has posted since the last walk.
- Posting days. Dates are the day regulations.gov posted the comment, in Eastern time. The agency posts in batches. A day with no bar is a day nothing posted, not a day nobody wrote.
- Campaigns. Exact duplicates are grouped by a hash of the normalized text. Near duplicates are found with 64-permutation MinHash over five-word shingles, banded for candidate pairs and verified against a representative at an estimated Jaccard similarity of 0.8 or higher, then joined by union-find. A group of 20 or more is a campaign. The threshold is a judgment; the download carries every group of 20 or more so a reader can cut it elsewhere.
- Campaign stance. A campaign’s stance is its exemplar’s. Where the exemplar has a machine classification (a “card”) that stance is used, covering 115,010 copies. The 36,309 copies whose exemplars had no card were read by the desk on Sep 28, 2026 from the exemplar’s own words; one-line exemplars that name neither the rule nor the proposal were left unclear rather than guessed, and attachment-only exemplars were left unread.
- Individual stance, arguments, geography. The platform’s classifier has read 96,001 comments (87,347 from the 2025 window, 8,654 from 2026). It reads the record in posting order, so that slice is the earliest part of each window, which is where organized campaigns concentrate. No figure from the classified slice is presented as a figure for the docket. A stratified random sample of the comments outside campaigns, sized for a margin under one point, is scheduled after the close.
- Organizations. A dated pass over the opening 1,500 characters of every comment and attachment in the 2026 window against a list of names. It finds organizations that name themselves; it cannot find the ones that didn’t, and it is a snapshot, not a live read.
- What is not claimed. A stance share for the whole docket. Any filer identity for an individual. That posted equals received. That the 2000 rule’s 1.6 million is comparable to this count without the received-versus-posted caveat.
- Reproducibility. Every count re-derives from the platform on every render, on a thirty-minute cache, and the page stamps the time. After the close the figures lock on the day the final walk completes and the page says so.
Cite as
Downloads
Summary tables, CC BY 4.0, credit “More Than Just Parks” with a link to this page. The comment texts themselves are public record at regulations.gov.
Before the close
The docket is open until October 6
Our Roadless Rule map shows every inventoried roadless area the proposal would open, and its comment builder tells you which arguments the record already has thousands of and where a unique one still counts.
See the map and file a commentCorrections
No corrections yet. If a figure on this page is wrong, write ask@morethanjustparks.com and we’ll fix it and log it here.