In 2024 the EPA set the first nationwide legal limit on PFAS in drinking water: 4 parts per trillion for PFOA and for PFOS, the two most-studied "forever chemicals." A part per trillion is roughly one drop in 20 Olympic swimming pools. The fact that the limit is that low tells you how toxic these compounds are at almost any dose.
> Correction, October 3, 2026: an earlier version of this study said the worst 2013-2015 systems now read near zero. A unit error in our database understated every 2023-2025 reading by a factor of 1,000. With the error fixed, five of the ten worst historic systems test non-detect for PFOA and PFOS and one was not retested, but four, including Saipan, still had samples above the 4 ppt limit. The table and answers below are corrected.
The country has now been tested against that idea twice, a decade apart. The first national look ran from 2013 to 2015; the most recent, and by far the largest, ran from 2023 to 2025. Putting the two side by side is the clearest picture we have of where PFAS in tap water actually stands – and it is more hopeful at the top of the list, and more sobering across the middle, than either round alone suggests.
The worst PFAS readings in the EPA's testing history are from the first round, 2013-2015: 162 water systems recorded PFOA or PFOS above today's 4 ppt limit, topping out at 7,000 ppt on Saipan. In the newest round, 2023-2025, five of the ten worst systems tested non-detect for PFOA and PFOS (a sixth was not retested), but four, including Saipan, still had samples above the 4 ppt limit. But contamination did not go away – it moved from a few catastrophic sites to a wide, low-level background. An analysis of the 2023-2025 data by the lab network Pace Analytical found roughly 8% of tested systems over the 4 ppt limit for PFOA and about 8.9% over it for PFOS, and the NRDC found that about 136 million Americans get water from a system where PFAS was detected at some level.
Before this decade, national PFAS data was patchy – states tested on their own schedules with their own methods. Two EPA programs changed that:
This difference is the single most misread thing about PFAS data. A 7,000 ppt reading from 2014 and a 3 ppt reading from 2024 are not the same kind of measurement, and a system that looks "clean" now was often filthy a decade ago. Comparing the raw numbers without the dates attached is how bad rankings get built.
Here is the top of the UCMR3 list – the worst PFAS ever recorded in nationwide EPA testing. Because the old methods could only detect PFOA above ~20 ppt and PFOS above ~40 ppt, every system on it was, by definition, far over today's 4 ppt limit:
| Water system | State | Worst reading (2013-2015) | | --- | --- | --- | | Commonwealth Utilities Corp. | Saipan (MP) | PFOS 7,000 ppt | | Artesian Water Company | DE | PFOS 1,800 ppt | | Security WSD | CO | PFOS 1,300 ppt | | Warminster Municipal Authority | PA | PFOS 1,090 ppt | | Horsham Water & Sewer Authority | PA | PFOS 1,000 ppt | | Oatman Water Company | AZ | PFHxS 730 ppt | | Warrington Township | PA | PFOS 667 ppt | | Issaquah Water System | WA | PFOS 600 ppt | | Suffolk County Water Authority | NY | PFOS 530 ppt | | Oakdale, MN | MN | PFOS 439 ppt |
A 7,000 ppt PFOS reading is 1,750 times the 4 ppt federal limit. The pattern almost never breaks: the cluster of Pennsylvania systems (Warminster, Horsham, Warrington) sits beside former naval air stations; Colorado's Security system is downstream of Peterson Space Force Base; Oakdale, Minnesota traces to decades of 3M production. The worst PFAS in America has almost always been a firefighting-foam or industrial story.
This is the part that gets lost when a decade-old number is quoted as if it were current. The systems at the top of that 2013-2015 list were among the first to act. Utilities under legal and public pressure installed granular activated carbon and reverse-osmosis treatment, switched water sources, or blended supplies. When the EPA came back to test in 2023-2025, every one of them was far below its peak, but not all of them were below the new limit (highest single 2023-2025 PFOA or PFOS sample shown; Oatman, AZ was not retested, so the next-worst system, Hyannis, MA, takes its place):
| Water system | 2013-2015 peak | 2023-2025 reading | | --- | --- | --- | | Commonwealth Utilities Corp. (Saipan) | PFOS 7,000 ppt | PFOS 32.1 ppt | | Artesian Water Company (DE) | PFOS 1,800 ppt | PFOA 23.1 ppt | | Security WSD (CO) | PFOS 1,300 ppt | non-detect | | Warminster Municipal Authority (PA) | PFOS 1,090 ppt | non-detect | | Horsham Water & Sewer Authority (PA) | PFOS 1,000 ppt | PFOS 33.8 ppt | | Warrington Township (PA) | PFOS 667 ppt | non-detect | | Issaquah Water System (WA) | PFOS 600 ppt | non-detect | | Suffolk County Water Authority (NY) | PFOS 530 ppt | PFOS 17.4 ppt | | Oakdale, MN | PFOS 439 ppt | non-detect | | Hyannis, MA | PFOS 430 ppt | non-detect |
That is the honest headline of the newest data: treatment works, and six of the ten systems in this table are now non-detect for PFOA and PFOS. But Saipan, Horsham, Artesian and Suffolk County still had 2023-2025 samples several times the 4 ppt limit, even after cutting their peaks by more than 95%. If someone quotes you a "7,000 ppt" or "1,800 ppt" tap-water number in 2026, they are quoting the 2014 crisis, not the 2024 result.
A few caveats keep this honest. "Non-detect" here means no PFOA or PFOS above the reporting limit: several of these systems still reported other, currently unregulated PFAS, such as PFBA at 105 ppt in Oakdale and PFPeA at about 64 ppt in the Security district. First, a handful of small systems from the old list (for example, Oatman, AZ) were not in the 2023-2025 sample, so their current status is simply unmeasured, not confirmed clean. Second, these are single samples, and the EPA judges compliance on a running annual average, so a sample above 4 ppt shows real exposure but is not by itself a violation. Third, ten famous systems are not the whole country, which is the whole point of the next section.
The top of the list got better; the middle did not. The best national read of the 2023-2025 results comes from independent analyses of the full EPA dataset, since that dataset is still being finalized:
So the shape of the problem changed. In 2013-2015 the story was a few systems at catastrophic levels. In 2023-2025 it is a very large number of systems at low-but-detectable levels, with a meaningful slice still over the 4 ppt line – just rarely at the four-figure readings that made headlines a decade ago. The EPA itself notes the current figure is an undercount: about 5% of expected results had not yet been reported, and a single sampling day can miss contamination.
Here is a detail almost no consumer coverage explains, and it trips up even data professionals. UCMR5 tested for lithium at the same time as the 29 PFAS, and lithium shows up at far higher concentrations than PFAS, often tens of thousands of parts per trillion (tens of parts per billion) versus single or double digits.
If you sort raw 2023-2025 readings by concentration without separating lithium out, the top of the list is almost entirely lithium, not PFAS. We caught this in our own data and it is a known issue – Pace Analytical specifically flagged the "lithium concerns" in interpreting UCMR5. So when you see a "worst water" ranking with current readings in the high hundreds or thousands of ppt, check whether the chemical is actually PFOS or PFOA, or whether lithium is quietly inflating the number. Lithium has its own health questions, but it is not a PFAS, and the 4 ppt limit does not apply to it.
Put the two traps together and you have the two ways PFAS rankings get faked: quoting a 2013-2015 peak as if it were current, or letting lithium ride to the top of a 2023-2025 list. A genuine PFOS or PFOA reading over the limit today is usually in the single digits to low hundreds of ppt.
A 4 ppt limit and an 8% exceedance rate can feel abstract. Here is the practical translation:
If you are on a public system, there is a real chance PFAS has been detected in your water, and a smaller but real chance it is over the new limit. Systems that are over the line have until the EPA's compliance deadline (2029 under the 2024 rule; in May 2026 the EPA proposed letting systems extend it to 2031) to install treatment – years away. And if you are on a private well, no one is testing for you at all.
The good news is the fix at home is well understood, because it is the same technology the worst systems used to clean themselves up. Reverse osmosis removes 90% or more of PFAS, and a filter certified to NSF/ANSI 53 or 58 for PFOA and PFOS reduction handles it at the tap. Boiling does nothing – it actually concentrates PFAS. For the specifics, see our guide to the best PFAS water filters and how to choose one for your situation. To see which systems still rank worst on current data, check the worst-cities ranking.
PFAS is one of several exposures we track at national scale. Our companion analyses cover the others: The State of Lead in US Drinking Water maps where lead still leaches from aging service lines, and The State of Lead Paint in US Homes estimates how many homes still carry the pre-1978 lead-paint risk.
### How many US water systems exceed the EPA PFAS limit? In the newest round of EPA testing (2023-2025), an analysis by Pace Analytical found about 8% of tested public water systems over the 4 ppt limit for PFOA and about 8.9% over it for PFOS. The EPA considers the figure an undercount, since testing is about 95% complete and single samples can miss contamination.
### What was the highest PFAS level ever found in US tap water? The highest reading in nationwide EPA testing was PFOS at 7,000 ppt – 1,750 times the 4 ppt limit – recorded at the Commonwealth Utilities Corporation system on Saipan during the 2013-2015 round (UCMR3). That system has since installed treatment, and its highest 2023-2025 sample was 32.1 ppt PFOS: more than 99% lower, but still about 8 times the limit. It is important to attach the date: 7,000 ppt is a 2014 crisis number, not a current one.
### Have the worst PFAS water systems been cleaned up? Partly. Of the ten systems with the highest PFAS readings in 2013-2015, five tested non-detect for PFOA and PFOS in the 2023-2025 EPA data (a sixth, Oatman, AZ, was not retested), including Warminster and Warrington, PA and Colorado's Security district. Four, Saipan, Horsham (PA), Artesian Water (DE) and Suffolk County (NY), cut their levels by more than 95% but still had samples above the 4 ppt limit. Treatment works, but not everywhere is finished, and the much larger problem is the number of systems with lower-level detections.
### How many people have PFAS in their drinking water? The NRDC's analysis of the 2023-2025 data found roughly 136 million of the 280 million Americans tested so far are served by systems where PFAS was detected. The NRDC also found PFAS above EPA thresholds in 79% of congressional districts, and it notes the true number is likely higher.
### What is the EPA limit for PFAS in drinking water? The EPA's 2024 rule sets an enforceable limit of 4 parts per trillion (ppt) each for PFOA and PFOS, plus limits on PFHxS, PFNA, and GenX (10 ppt) and a mixture limit using a Hazard Index. Systems over the limit must reduce it by the 2029 compliance deadline. In May 2026 the EPA proposed rescinding the PFHxS, PFNA, GenX and Hazard Index limits and letting systems extend the PFOA/PFOS deadline to 2031.
### Why do some PFAS rankings show numbers in the thousands? Two reasons. Either the ranking is quoting a 2013-2015 peak (when the worst sites read in the thousands) as if it were current, or it is letting lithium – a non-PFAS analyte UCMR5 measured alongside PFAS, at far higher concentrations – inflate the list. A genuine PFOS or PFOA reading over the limit today is usually single digits to low hundreds of ppt.
### How do I find out if my water has PFAS? Look up your system. We publish the EPA PFAS results for thousands of water systems by ZIP code, city, and state. Check your water at the KnowYourExposure home page to see the specific compounds and levels detected where you live.
The country now has its clearest picture of PFAS in tap water: a decade ago the danger was a few systems at catastrophic levels; today it is a near-universal, lower-level background with the worst historic sites far below their peaks, though several are still over the new limit. You cannot fix the national pipeline, but you can find out exactly what is in your own water and filter it. Start by looking up your ZIP code.
This analysis draws on the EPA's two national PFAS testing programs – the third Unregulated Contaminant Monitoring Rule (UCMR3, 2013-2015) and the fifth (UCMR5, 2023-2025) – aggregated and de-duplicated in our own database. Lithium is excluded from every PFAS aggregate (it is not a PFAS). Population-served figures follow the NRDC's analysis of the UCMR5 data. Full sources, definitions, and our public corrections log are on our research page. Data as of August 2026.
Cite this analysis (APA):
> Know Your Exposure Research Team. (2026). The State of PFAS in US Drinking Water (2026). KnowYourExposure. https://knowyourexposure.com/blog/state-of-pfas-in-us-drinking-water-2026
Journalists and researchers are welcome to cite these figures with a link back. For a custom data pull or a quote, contact [email protected].