The FDA put this so plainly that the quote belongs at the top:
When a product is advertised to disinfect in seconds, it likely means that it gives off an unsafe level of UV-C radiation.
In this category the promise of speed is a warning label, not a feature. It is a rare thing for a regulator to name a marketing claim as a marker of danger.
Start with the physics though, because it decides everything else.
Line of sight, and nothing else
UV-C works like light, because it is light. It does not flow around obstacles, does not soak in, does not work around a corner. A photon either reached the cell or it did not.
For a smooth hard surface that is an excellent technology. Food has a problem: food has texture.
A stoma on a lettuce leaf is an opening, and a bacterium inside it receives a dose of zero. The fold at the base of the leaf is shade. A hair on a raspberry is shade. A pore in apple skin is shade. Chemistry works poorly in those places too, but at least it flows in. Light flows nowhere.
How much that matters, in numbers
The best measurement on this question comes from Huang, de Vries and Chen in 2018 (Int J Food Microbiol, volume 283). They ran UV directly in water with agitation, which is far kinder than a still item under a lamp: the produce turns, the water moves the shadow, the dose spreads more evenly.
UV alone, no chemistry, against Salmonella:
| Produce | Reduction, log | Surface |
|---|---|---|
| Tomato | over 3.6 | smooth, continuous |
| Iceberg | 2.6 | dense leaf, little texture |
| Carrot | 2.0 | rough |
| Blueberry | 1.7 | smooth, but with a calyx well |
| Spinach | 0.9 | thin leaf, stomata, folds |
Same rig, same dose, same Salmonella. Tomato against spinach is a fourfold gap in logs, which is a hundredfold gap in surviving bacteria.
That is line of sight expressed as data. The more complex the surface, the less UV buys you. The authors’ own recommendation tells the story: for spinach they recommend chlorine, and they leave UV on its own only for tomato.
Note the conditions once more. Their water was realistically dirty, COD raised to 2000 mg/L, turbidity above 60 NTU, and turbid water absorbs UV. In a countertop box there is no water at all, the produce lies still, and half its surface faces the floor of the chamber. How water clarity, dose and flow rate decide whether a UV unit works at all is explained in the wiki: UV-C water treatment and UVT.
An honest caveat: texture does not explain everything
Shade is a convenient universal explanation, which is a reason to check it.
Lightbown and DiCaprio in 2025 (Front Plant Sci, volume 16) did something clever: they cast exact polymer replicas of romaine and spinach leaf topography, inoculated them with virus, and compared the casts against real leaves. If geometry decides everything, a cast should behave like a leaf.
Their conclusion: “surface topography has limited impact on virus removal from leafy greens.” Binding mechanisms, both specific and non-specific, are doing work as well.
That study concerns chemical sanitizers rather than UV, and for light the line of sight problem remains a geometric fact. Still, the caution is worth carrying.
The same paper produced a figure worth keeping on its own. Spinach at 50 ppm sodium hypochlorite: 15-day-old leaves gave 1.97 to 2.28 log, 45-day-old leaves gave 0.36 to 0.52. A mature leaf shelters contamination roughly five times better than a young one, under identical treatment. Nobody prints leaf age on the bag.
Why the combined numbers look better
In that same study, UV together with peracetic acid held residual Salmonella in the wash water below the detection limit of 2 CFU/mL in almost every replicate.
Notice what improved. The water. Not the leaf. That is exactly the logic industry applies to wash water sanitizers: treat the water so that one contaminated leaf does not contaminate the rest of the batch.
Now the device itself
In July 2022 the FDA issued a safety communication about UV wands, updated in August 2023. The wording:
Some UV wand products give off at a distance of about two inches, as much as 3,000 times more UV-C radiation than the exposure limit recommended by the International Commission on Non-Ionizing Radiation Protection.
Three thousand times the international exposure limit, at two inches.
The injuries the FDA names: erythema and a burn-like skin reaction, and photokeratitis for the eyes. Time to injury, verbatim, “after a few seconds of use.” Photokeratitis feels like sand in the eyes and hurts badly, typically some hours after exposure, by which point connecting cause to effect is difficult.
The list of products FDA tested includes recognizable names: Sharper Image UV Sanitizing Portable Wand, several OttLite models, UVILIZER Flip, SurfaceSoap UV, 59S Wand Disinfector.
The features these unsafe products had in common, as the FDA lists them:
Lack of safety information, claims to disinfect in seconds, and the lack of any means to protect users.
Read the middle item again. Speed of the promise ties directly to emitted power, and power ties to risk. A device that honestly needs thirty minutes is safer than one promising ten seconds.
Boxes against wands
A closed UV box with a lid interlock does solve the eye and skin problem: the lamp cuts out when you open it. On user safety that is a different category, and lumping it in with wands is unfair.
What a box does not solve is shade. A tomato in a box rests on the floor of the chamber, and the underside of the tomato receives nothing. A bag of salad in a box is several hundred leaves, of which the light sees the top two.
Where a box genuinely earns its place is smooth hard objects you can lay out singly and turn over: keys, a phone, glasses. Food does not fit that description.
What to do instead
- Running water and friction. Mechanics do not depend on line of sight: fingers and a brush reach where a photon cannot.
- Do not soak the batch in one bowl. Standing water without a sanitizer spreads contamination across the batch.
- Dry after washing. An FDA recommendation, and it removes a further share of surface bacteria.
- Leave bagged washed greens alone. Details in do triple washed claims mean anything.
- For higher-risk eaters, the answer is heat. UV is not a kill step on textured produce, and no household soak is either.
Bottom line
- UV-C acts only along line of sight. Stomata, folds and hairs are shade, and in shade the dose is zero.
- On one rig: tomato over 3.6 log, spinach 0.9. The difference is texture, not lamp power.
- Combined with chemistry, UV improves the cleanliness of the wash water rather than the leaf.
- FDA measured consumer UV wands at up to 3,000 times the international exposure limit at two inches, with skin burns and photokeratitis after seconds.
- A promise to disinfect in seconds is itself a marker of an unsafe emission level. That is the FDA’s own framing.
- A closed box removes the risk to your eyes and leaves the shade exactly where it was.
Neighboring gadgets and methods: ozone produce washers, baking soda, vinegar, hydrogen peroxide, ready-made sprays. The wider picture sits in the leafy greens food safety hub.
