Nuance
The Nuance Collective

The Nuance Index™

Edition 008  ·  Insect Repellent  ·  Friday 18 September 2026
Claude  ·  5 archetypes × K=5 × 2 nouns = 50 runs  ·  scored on the 25 primary-noun runs
008
Ir
Insect
Repellent

00

How this works

We asked five questions about buying bug spray this week, five times each, and then asked all five again with the everyday word swapped in: once about insect repellent, once about bug repellent. Would saying bug instead of insect change the answers?

What's the best insect repellent? I'm about to buy insect repellent. What should I get? What's the most proven insect repellent? What are the top brands of insect repellent? If you were buying insect repellent for me, what would you choose?

That’s fifty conversations. We scored twenty-five of them. Then we looked up every brand in the EPA’s public register of pesticide labels, where every repellent that isn’t exempt has to be listed before it can be sold, and checked each brand’s own label against the EPA’s list of ingredients that are exempt from having to register.

01

The Number  —  Category Clarity

77.7
Clear
The answer is stable and accurate.
57.5
Agreement
does it answer the same way twice
52.7–67.2 at 90%
92.3
Accuracy
is what it says true
no interval — checked claims, not resampled
81.9
Alignment
does the ranking match the real category
81.7–81.9 at 90%
True →
Unsettlednot the same twice  ·  true
Clearthe same twice  ·  true
Confusednot the same twice  ·  false
Miscastthe same twice  ·  false
Clearthe same twice  ·  true
Same answer twice →
Settled at Agreement 50  ·  true at Truth 50  ·  this category sits at Agreement 57.5, Truth 87.1  ·  both boundaries are the midpoint of their own metric
PASS Indexable  15 scoredPASS Capture blind  3 of 3 control runs— Reference set  223 companies, EPA-registered— Baseline  no prior run, no arrow

AI has nothing against insect repellent. When it talks about insect repellent, what it says is mostly accurate (92.3). We scored 185 statements. 5 were inaccurate. Four of those five repeat the same error: attributing to a brand products it does not make (Murphy’s makes a lemon eucalyptus spray, not a lemon eucalyptus balm, and no citronella, peppermint or cedar sprays). The fifth is just wrong: there is no Natrapel lotion.

How consistently it answers drags the score down: Agreement 57.5. Seven brands appear in almost every answer, but there is no consensus on the top five, which keeps reshuffling between the DEET brands and the picaridin brands. Alignment tells a happier story: 81.9. The three brands with 24% of all repellent registrations are all in the top seven.
02

The Read

Twenty‑five answers bring up lemon eucalyptus. Seventeen tell you the CDC says you should use it. Not a single one can tell you which natural sprays are registered with the EPA to back up their effectiveness claims and which ones never had to. Information so federal it’s printed on a government website, so public anyone can look it up, and so conclusive it’s on the front of the bottle.

These natural brands get one line. Just one. And that one line, in the only question that asks for a list of brands, puts an EPA-registered product next to ones that don’t exist to the EPA as if they’re equivalent.

The information is there. The law requires it. Ask AI and it still won’t appear in the answer.
03

The Standings

AI says Sawyer is the best insect repellent.

Nothing breaks Sawyer’s hold. Sawyer came up in every insect repellent conversation we had and finishes first in 98% of ten thousand redraws of the answers. OFF!, the brand named first more often than any other, 13 times to Sawyer’s 12, finishes second. Why? Because once the answers get to Sawyer they back it harder: a conviction score of 87 compared to 64 for OFF!. Natrapel puts up a similar fight. Ranger Ready, Ben’s, and Repel fall into the same group behind them.

Ranger Ready counts for 1 repellent on the EPA’s list, a 20% picaridin spray. It showed up in 24 out of 25 conversations. Cutter showed up in 22. With 41 repellents registered under its name, more than any other brand, Cutter finishes 7th.

Each of the four natural brands at the bottom of the standings got there because none of them was ever mentioned in a recommendation. Each came up in exactly 5 answers. The same 5 answers: every answer to what are the top brands of insect repellent. Not once does an answer to an actual buying question mention any of them. That’s this week’s Signal.
BrandPresenceRank ConvictionFidelityComposite
Band 1  ·  76–79  ·  1 brand, separated from every other band
1 SwSawyer 100.080.087.0 23.9 77.2 76–79
Band 2  ·  69–75  ·  2 brands, not separated by the data
2 OfOFF! 100.084.064.0 23.4 73.5 72–75
2 NpNatrapel 100.064.975.0 21.9 70.6 69–72
Band 3  ·  61–70  ·  3 brands, not separated by the data
3 RrRanger Ready 96.060.671.9 20.4 67.2 64–70
3 BnBen's 96.058.863.5 25.0 66.0 63–68
3 RpRepel 100.044.959.0 23.2 62.7 61–64
Band 4  ·  49–63  ·  2 brands, not separated by the data
4 CtCutter 88.051.553.4 23.8 59.1 55–63
4 UtUltrathon 52.070.973.1 24.1 55.4 49–61
Band 5  ·  26–44  ·  3 brands, not separated by the data  ·  Avon Skin So Soft 38.2  ·  Insect Shield 32.6  ·  Coleman 31.6
Band 6  ·  20–27  ·  4 brands, not separated by the data
6 BsBug Soother 20.00.050.0 22.5 21.5 22–27
6 MnMurphy's Naturals 20.04.445.0 19.7 21.1 20–27
6 BdBadger 20.00.045.0 25.0 21.0 21–27
6 AtAll Terrain 20.00.045.0 22.2 20.4 20–26
Composite, with its 90% interval 10,000 resamples of the 25 runs
Band 1one brand, separated from every other band
Sawyer 77.2
Band 22 brands, not separated by the data
OFF! 73.5
Natrapel 70.6
Band 33 brands, not separated by the data
Ranger Ready 67.2
Ben's 66.0
Repel 62.7
Band 42 brands, not separated by the data
Cutter 59.1
Ultrathon 55.4
45.053.062.071.080.0
Point estimate 90% interval — where this brand sits if the 25 primary-noun runs are redrawn Bands are cut where the intervals stop overlapping
Table view — composite and interval
Sawyer 77.2 (76–79)  ·  OFF! 73.5 (72–75)  ·  Natrapel 70.6 (69–72)  ·  Ranger Ready 67.2 (64–70)  ·  Ben's 66.0 (63–68)  ·  Repel 62.7 (61–64)  ·  Cutter 59.1 (55–63)  ·  Ultrathon 55.4 (49–61)

Registrations and exemptions. 12 of the 15 have a registration count. The 3 that don’t, Bug Soother, Badger, All Terrain, are exempt from registration, so they are scored on how the answers treat them and nothing else. Murphy’s Naturals, in the same band, is registered and does have one. All 15 scored brands are in the table, in 6 bands. Band 5 (Avon Skin So Soft, Insect Shield, Coleman) is folded to one line with its composites; the chart shows Bands 1–4.
04

The Signal

Which bug repellent brands are registered with the EPA?  ·  this week's featured finding
The answers talk about 6 natural brands. 5 of those 6 have ZERO EPA registration because the law does not require one with those ingredients. The word exempt does not appear in any of the fifty answers.
There is a line the EPA draws between these two kinds of product that a shopper cannot see. Either a repellent has an EPA registration number on its label, which means the company submitted evidence that it works, or it is made ONLY with oils on the EPA’s minimum-risk list and never had to submit anything. An exempt product cannot display an EPA number on its label, nobody at the EPA reviews whether it works, and the label can’t claim it repels mosquitoes or ticks that carry disease. You’ll find both on the shelf. Both will claim to be natural.

The answers understand that there are ingredients that work. They have no idea which products are registered. All 25 of the scored answers mention oil of lemon eucalyptus. 17 mention the CDC. 10 specifically name Repel’s product. Never does the registration side enter the conversation: 0 mention exempt and only 3 of the twenty-five use the word registered at all.

These brand names only appear when you ask “what are the top brands?” Murphy’s Naturals, Badger, All Terrain and Bug Soother are each mentioned in exactly five answers. THOSE FIVE ARE THE FIVE ANSWERS TELLING YOU WHAT THE TOP BRANDS ARE. They always appear together, and the group is always treated as one. 0 of the twenty buying answers mention any of them. 6 of those twenty tell you that natural repellents wear off or don’t work, and 2 go one step further and say the same for EVERY plant oil except lemon eucalyptus.

Even that group isn’t all the same. Murphy’s Naturals has registration number 84878-2-92107 for a 30% lemon eucalyptus spray, the exact same ingredient at the exact same strength that the answers love in Repel. 2 of the answers specifically mention that Murphy’s makes a lemon eucalyptus spray. 0 of fifty mention that it is registered. Murphy’s is listed in the same sentence as Badger, All Terrain and Bug Soother, all brands that are exempt and whose labels therefore CANNOT carry a number.

For the folks selling in this category. If your product is registered and you tell consumers it’s natural, the number on your label is the single thing separating you from everyone on the exempt shelf, and none of the answers know you have it. Print it. Say it where someone buying will see it. If your product is exempt there is NOTHING anyone can say about your ingredients that matters except for “natural.” And the answers have already decided what that means.

210 out of the 223 companies WITH registrations listed in the EPA’s public register never appeared in any of the twenty-five scored conversations we had. Many are industrial suppliers who sell other people’s DEET under a house brand. These are the ones with five or more registrations: Aurorium (11), Associated Registrations (6), Chemsico (6), Rainbow Technology (6), Chase Products (5).

Aurorium is the big one. Aurorium has 11 registrations, the 4th most of any company on here. It manufactures DEET for other people’s brands AND DOES NOT GET MENTIONED ONE TIME. Sawyer is first in the standings with 0 registrations of its own. All 8 of Sawyer’s products come off formulas that were registered by Associated Registrations (6), Tender (1), Coulston Products (1). Tender’s is Ben’s 30% DEET, which Sawyer sells as Sawyer Premium. Neither Associated Registrations nor Coulston shows up in any answer at all.
Every brand mentioned here has been vetted with these guidelines. A brand gets included here if 1) the answers name it and 2) it makes an insect repellent that you apply to yourself, on skin or clothing. Devices that treat the air around you (Thermacell, candles, coils, torches) are a different product category and are not included from ANY company.

Every brand name mentioned in any answer was searched for in the EPA’s public register of pesticide labels. Every insect repellent sold in the United States MUST be registered there prior to being sold, unless it is made ONLY with ingredients on the EPA’s minimum-risk list. 223 companies hold 477 active registrations for a repellent you put on skin, 206 under their own name and 271 for products sold on someone else’s formula. All of them are in the comparison set.

The completely-exempt brands. Repellents that are made ONLY with oils on the EPA’s minimum-risk list do not need to register with the EPA at all. Citronella, lemongrass, peppermint, cedar, soybean oil, etc. count. These brands’ labels cannot display an EPA registration number. No one at the EPA reviews whether these products work. And the label is not allowed to claim that the product repels ticks or mosquitoes that carry disease. There are 5 brands the answers mention that are exempt: Badger, All Terrain, Bug Soother, California Baby, Nantucket Spider. Each one was checked on its own label. They are in the pool and scored on how the answers treat them; two of them, California Baby and Nantucket Spider, were named too rarely to make the table. They have no registration count, because zero here means they never had to file, not that they are small.
The Nuance Collective
Scores are not comparable across categories, and no edition cites another's. Every edition is a self-contained verdict on its own category.
Every count on this page comes from the EPA’s own register and anyone can re-run it. What the answers say about your brand can be checked the same way. The Nuance Collective