This map traces back to one primary document: an internal NPS review database that was leaked to the public in early March 2026. The review itself was conducted between June and September 2025, after Secretary’s Order 3431 directed every park in the system to inspect its signs, exhibits, films, and publications. Park staff filled out a survey for each item they reviewed. That survey is available for download and is our starting point.

The database has 879 entries. Each one represents a single item (an exhibit panel, a wayside marker, a film, a kiosk, a brochure) that a park identified as needing review. For each entry, we know the park code, the region, and what kind of media it was. We also know what the park staff wrote about the item: what they thought needed to change, and what action they recommended.

The leaked documents also include screenshots of the actual survey form, which helps explain the data. The form asked staff to select one or more media types: Interior/Exterior Exhibits, Signs and Waysides, Publications, Monuments/Statues/Markers, Film, or NothingToReport. For each media type, they could choose whether the item “Needs Repair/Replacement,” “Needs Review: Contains content that may conflict with SO 3431 or EO 14253,” or “Other.” A description field asked them to “include thoughts on how to make the change.”

That last point matters, because it means some park staff described potential revisions even while asserting that nothing actually needed changing. The survey was designed to collect flagged content, but with little guidance on how to interpret the order, parks took very different approaches. As Save Our Signs noted in their excellent analysis of the leaked data, some parks submitted entire exhibits for review, some flagged individual passages on a single sign, some used it to report signs that were physically deteriorating, and some even submitted planned exhibits that hadn’t been built yet. The data is messy, and we try to be transparent about that.

It’s also worth noting that the survey was not the only way content was flagged. The leaked documentation shows at least three other methods: QR code signs posted at parks asking for public feedback (sometimes called “snitch signs”), an internal survey flagging retail items, and a separate survey about materials that had been changed under the previous administration. Our map focuses on the main review database, but the full picture is broader than any single document.

The NPS survey form had two different reasons a park could flag an item. One was “Needs Review: Contains content that may conflict with SO 3431 or EO 14253.” The other was “Needs Repair/Replacement,” meaning the sign was physically damaged, sun-faded, or illegible, with no content issue at all.

Both types of flags ended up in the same spreadsheet. A wayside that’s deteriorating and impossible to read is a very different situation from a wayside that was pulled down because it mentions slavery. But in the raw data, they can look almost identical.

We went through the original survey data and classified every entry by its flag type. Out of 874 entries on the map, 275 were flagged as “Needs Review” (a content conflict with SO 3431), 14 were flagged as “Needs Repair/Replacement” (physical damage only), and 9 were flagged as “Other.” The remaining entries either had no flag specified or came from outside the main NPS survey.

The 14 repair entries include sites like Grand Canyon, Sitka, Canyonlands, and Rocky Mountain. On the map, these show up with a distinct color (indigo) and a note explaining they were flagged for physical maintenance, not content censorship. We think it’s important to keep them visible rather than quietly deleting them, because they help illustrate how messy the underlying data really is. But we don’t want anyone counting them as censorship cases, because they’re not.

If you see an entry that looks like it’s about maintenance rather than censorship, check the note on the popup. We’ve tried to label everything we can. But the survey form made the description field optional and it was filled out inconsistently, so in some cases it’s genuinely hard to tell. We’d rather include an entry with a caveat than silently exclude it.

The leaked database doesn’t have a neat “status” column we can just copy. Every entry in the original is marked Submitted, all 879 of them. That tells us the park sent in its survey. It doesn’t tell us what actually happened.

So we built the status ourselves, using a layered process. Here’s how it works, step by step.

Step 1: Read the action columns. The database has separate action fields for different media types (films, exhibits, signs, publications, waysides). If a park wrote “Remove” in the exhibit action column, that entry starts as Replace/Remove. If they wrote “Revise,” it starts as Revise. If no action was specified at all, it starts as Flagged for Review, meaning the park identified the item but didn’t say what to do with it.

Step 2: Check the flag type. As described above, we look at whether the park selected “Needs Review” (content conflict) or “Needs Repair/Replacement” (physical damage). Entries flagged for repair get a separate status and a note explaining they’re not censorship-related.

Step 3: Check for independent confirmation. We cross-reference every entry against outside sources: the Save Our Signs Removal Tracker, news reporting, court filings, and FOIA data. If a newspaper reported that a sign was physically taken down, or if SOS volunteers photographed the empty spot where it used to hang, the entry moves to Confirmed Removed. This is a higher bar than what the park wrote in a survey. It means someone on the ground verified the sign is gone.

Step 4: Check for restoration. Some signs have come back. In a few cases, federal courts ordered them restored. If we have evidence that an entry has been put back in place, it gets marked accordingly. “Restored” doesn’t always mean “fully restored,” though. Sometimes only some of the signs at a site came back. We track that too.

Here’s what the statuses mean:

Status	What it means	How we know
Flagged for Review	Park staff identified this item for review under SO 3431, but either didn’t specify an action or the action wasn’t “Remove” or “Revise.”	NPS internal review database. This is the default status for entries without a clear removal or revision directive.
Revise	Park staff recommended revising the content: changing wording, removing a sentence, or altering the interpretation.	NPS internal review database action columns.
Replace / Remove	Park staff recommended full removal or replacement of the item.	NPS internal review database action columns.
Confirmed Removed	The item has been independently confirmed as physically removed from the park.	Press reporting, SOS on-the-ground verification, court filings, or FOIA documents.
Court-Ordered Restored	A federal court ordered the item restored, and restoration has been confirmed.	Court orders plus SOS or press confirmation of physical restoration.
Partially Restored	A court ordered restoration, but only some items at the site have been confirmed back in place.	Court orders plus SOS restoration tracking showing incomplete restoration.
Needs Repair	The park selected “Needs Repair/Replacement” on the survey form, indicating physical damage. This is not a censorship flag.	Original NPS survey radio button selection. 14 entries across parks like Grand Canyon, Sitka, Canyonlands, and Rocky Mountain.
Other	The park selected “Other” on the survey form. These don’t fit neatly into either the content review or physical repair categories.	Original NPS survey radio button selection. 9 entries including Fort McHenry, Petrified Forest, and Dry Tortugas.

Independence National Historical Park in Philadelphia is a good example of why this process matters, and why you can’t just take the database at face value.

The NPS database has 9 separate entries for Independence. Each one covers a different exhibit at the park, with different buildings, different topics, and different media types. Here are a few of them:

Entry	What it covers	NPS action
439	President’s House: 30 signs and displays on slavery and the enslaved people who lived there	Physically removed
442	Liberty Bell Center: panels on abolitionists adopting the bell as a symbol	Revise
674	Benjamin Franklin Court: touchscreen interactive on Franklin’s relationship with slavery	Flagged for Review
734	West Wing of Independence Hall: design document for an exhibit not yet built	Flagged for Review

Now here’s where it gets complicated. The Save Our Signs Tracker has detailed, sign-by-sign data for Independence: 30 individual signs and displays, each with a name, before photos, and a restoration status. All 30 of those items are from one specific exhibit: the President’s House. Of those 30, 16 have been restored under a federal court order (Feb 19, 2026) and 14 remain removed.

SOS doesn’t track the Liberty Bell panels. It doesn’t track the Franklin interactive. It doesn’t track the West Wing design document. Those items were flagged in the NPS internal review, but no one on the ground has independently confirmed whether they were actually removed.

So what do we do?

The earlier version of this map treated all 9 Independence entries the same way: all labeled “Removed, Court-Ordered Restored,” all sharing the same sign names, all sharing the same photos. That was wrong. It overstated what we know about 8 of the 9 entries and obscured the real story, which is that the President’s House was the confirmed target, and the other exhibits were flagged but may never have been touched.

Here’s something you might not expect: out of 874 entries on this map, 292 of them have a park staff notation that says Nothing to Report.

That means there is an entry for the park in the NPS database without noting anything that needed changing. Since the database is our starting point, those entries are on the map because we think it is worth calling out.

We flag these entries with a “Nothing to Report” label so you can see them for what they are. They’re still worth tracking. The fact that a park had to formally review a sign about, say, the history of a Civil War battlefield and certify it was “compliant” with SO 3431 is itself a story. But we don’t want anyone to mistake these for confirmed removals. They’re not.

Each entry on the map has one or more topic labels, like “Slavery,” “Climate & Environment,” or “Indigenous History.” These are our labels, not the government’s. The NPS database doesn’t categorize entries by topic. We added them so you could filter the map and see patterns.

Most topics were assigned by keyword matching. We searched the text of what each park wrote about the item, specifically their description of what needed to change, and looked for specific words. If the description mentions “enslaved,” “slavery,” or “plantation,” it gets the Slavery topic. If it mentions “glacier,” “sea level,” or “climate,” it gets Climate.

This works well for clear-cut cases, but it’s not perfect. Some entries needed manual review.

We’re transparent about the confidence level. Some topics, like Slavery, match their keywords correctly about 99% of the time because the language is specific enough that false positives are rare. Others, like Colonization, are right only about a quarter of the time, because the keywords overlap with general historical language. We’re working on improving those.

Four primary sources feed into this map. Each one fills a different gap.

The internal NPS review database is the foundation. It tells us which items were flagged, at which parks, and what park staff recommended doing about them. It covers 879 entries across the entire National Park System. This is a leaked government document, available for download, and every entry on the map traces back to it.

The Save Our Signs Removal Tracker has been indispensable to this project. SOS is a volunteer-driven effort doing extraordinary work: visiting parks across the country, photographing signs before and after removal, documenting what’s been taken down, tracking what’s been restored, and making all of that data publicly available. Their StoryMap analysis of the leaked NPS data is the most thorough public walkthrough of the survey and its quirks. We rely on SOS to move entries from “flagged” to “confirmed removed” or “restored,” and we label SOS-sourced data as Data: SOS Removal Tracker. This project would not be possible without their groundwork (inclusion does not imply endorsement of this map’s analysis).

Press reporting and court filings provide independent confirmation. When the Philadelphia Inquirer reports that panels were removed from the President’s House, or when a federal judge orders them restored, we update the map accordingly. These sources override the database when they conflict. The database might say “Revise,” but if a reporter confirms the sign is gone, we call it removed.

FOIA data from the U.S. Fish & Wildlife Service, obtained through the persistent work of the Sierra Club, extends the map beyond the National Park System. The Sierra Club filed Freedom of Information Act requests that uncovered records documenting interpretive exhibits flagged or removed at National Wildlife Refuges, extending public visibility into censorship that would otherwise have gone unnoticed. These entries are labeled with an “FWS” prefix to distinguish them from NPS entries. We’re grateful for the Sierra Club’s ongoing commitment to government transparency and public lands advocacy.

Here’s the path each entry takes from raw data to the map:

Five entries from the original 879 were dropped (duplicate submissions or entries with no usable data). Eleven entries from outside the NPS database were added from FWS FOIA data. The final count is 874.

We want to be honest about the limits of this data.

We don’t know what happened at most parks. The database tells us what park staff recommended. It doesn’t tell us whether those recommendations were carried out. For the majority of entries, we have no independent confirmation either way. “Flagged for Review” might mean the sign is still hanging on the wall untouched. Or it might mean it was quietly taken down and nobody reported it. We don’t know.

GPS coordinates are park-level, not sign-level. When you see a pin on the map, it points to the center of the park, not the specific trail or building where the sign was. Multiple entries at the same park share the same pin.

Topic labels are imperfect. Keyword matching catches most cases, but it misses nuance. An entry about a Confederate monument might get tagged “Civil War” but miss the “Civil Rights” dimension. We’re improving this over time.

The database itself may be incomplete. It captures what parks submitted through one specific survey instrument. Removals that happened outside that process, or at parks that didn’t submit surveys, won’t appear here.

Restoration status lags reality. We rely on SOS volunteers and press coverage to confirm restorations. A sign might be back on the wall right now, but if nobody’s reported it, our map still shows it as removed.

This methodology page was last revised on March 25, 2026. The dataset is updated on a rolling basis as new removals are confirmed, court orders take effect, or community members submit corrections.