Drop your thermometer’s probe into a properly built ice bath and it should read 32°F — at sea level, in Denver, anywhere. That’s the whole test: five minutes, no gear beyond crushed ice, water, and a tall glass. If the reading misses by more than the thermometer’s stated accuracy spec, adjust it (dial models and some digitals), cook with a known offset, or replace it — many digitals can’t be adjusted at all. Then point your suspicion at the oven, because independent test kitchens have measured even healthy ovens swinging about 25°F above and below their setpoint. Catching that lie takes a leave-in probe and 30 to 60 minutes — a lazy-Sunday job rather than a five-minute one, and worth every minute.

Why your thermometer drifts

The USDA’s Food Safety and Inspection Service sets the baseline plainly: most food thermometers give a reading accurate within 2 to 4°F — and that’s when they’re behaving. Drops onto tile, thermal shock, a hard season of grill duty: that’s how drift happens. ThermoWorks draws a useful line between the two families. Dial (bimetallic-coil) thermometers drift from drops and ordinary kitchen mishaps and need regular re-checks; digital thermometers rarely need recalibration, though the company still recommends double-checking them every so often.

Rarely isn’t never. On the BBQ Brethren forums, a competition cook discovered his Thermapen reading roughly 14°F off during a boiling-water check — and the forum’s collected advice was to run the ice test first and swap the battery before blaming the sensor, since a dying battery can impersonate a broken thermometer. Independent testing backs up the stakes: when Serious Eats tested 18 instant-read thermometers, the top performers read ice water at 32 to 33°F and boiling water around 211°F, while the worst were off by 2 to 3°F — and were the slowest of the bunch, too. Accuracy and build quality tend to travel together.

The ice-bath test (the five-minute fix)

The ice bath is the gold standard of home accuracy checks for one beautiful reason: properly made, it reads 32.0°F within about a tenth of a degree, per ThermoWorks — and unlike boiling water, it reads 32°F at any elevation. No math, no exceptions. Here’s how to build one that actually holds that temperature:

  1. Fill a tall glass or small container to the very top with crushed ice. Crushed ice or small cubes pack tighter than big cubes, leaving fewer warm pockets of water.
  2. Add cold water to about half an inch below the top of the ice. The ice should stand proud of the waterline — a mound of ice with water worked into it, not cubes floating in water. Let it rest for about a minute, then stir.
  3. Insert the probe at least 2 inches into the center of the slurry. USDA guidance specifies a minimum 2-inch immersion, touching neither the sides nor the bottom of the container. Aim for the vertical middle.
  4. Stir gently and wait at least 30 seconds, until the reading settles. It should read 32°F.

The slurry trap. A glass of ice water is not an ice bath. ThermoWorks warns that a casual glass — a handful of cubes bobbing in water — can read as much as 12°F (6°C) too warm, which makes an accurate thermometer look broken and a broken one look fine. If you take one thing from this guide: mostly ice, just enough water to fill the gaps.

The boiling-water test (and why altitude changes everything)

The classic second opinion checks the hot end of the range — useful, but it carries an asterisk the ice bath doesn’t.

  1. Bring a pot of clean water to a full rolling boil. Use a pot at least 4 inches deep, per ThermoWorks, so the probe can be properly immersed. USDA guidance suggests distilled water for true accuracy, because dissolved minerals — and any salt — shift the boiling point.
  2. Insert the stem at least 2 inches, centered in the water. Keep it off the bottom and sides; touching the pot means measuring the metal, which runs hotter than the water.
  3. Wait at least 30 seconds. At sea level, boiling water reads 212°F.

The altitude rule: just under 1°F per 500 feet

That “at sea level” is doing heavy lifting. The USDA’s rule: each 500-foot increase in elevation lowers the boiling point of water by just under 1°F. At 2,000 feet, water boils at about 208°F; at 7,500 feet, around 198°F. ThermoWorks pegs Denver at roughly 202°F, and if you’d rather skip the arithmetic, its free boiling-point calculator computes the target for your location.

This trips up real cooks constantly. In a DiscussCooking thread, an owner living at 7,000 feet reported water boiling at about 199°F and wondered whether the thermometer needed recalibrating. It didn’t — that’s exactly what water does at 7,000 feet.

When to skip this test

If you live at meaningful elevation, or your water is heavily mineralized, the boiling test adds variables the ice bath simply doesn’t have. The consensus in those owner threads matches our read: the ice bath is the more reliable home check, because it’s the same 32°F in Key West and Santa Fe. Treat boiling water as a second data point when conditions are right, not as the primary verdict.

Reading your results: adjust, offset, or replace

Know the spec before you touch anything

Here’s the step nearly every calibration guide skips: a thermometer reading 33°F in an ice bath may be working exactly as designed. Every model has a stated accuracy spec, and a reading inside it is a pass, not a problem. ThermoWorks makes the point with its own instruments — the ThermoPop’s ice-point spec is ±2°F, so anything from 30 to 34°F means leave it alone, while a Thermapen is rated to ±0.7°F and gets judged against a much tighter window. Look up your model’s spec in the manual or on its product page before adjusting anything.

If it’s adjustable

Dial thermometers are the most fixable: USDA guidance notes that many have a calibration nut under the dial head. With the probe held in the ice bath, grip the nut with pliers or a small wrench and turn until the needle reads 32°F. Digitals vary by sensor type — the USDA says thermocouple models can be calibrated and some thermistor models can too, so check the manual for a CAL or reset function.

If it’s not — or it’s too far gone

Some thermometers can’t be brought back in line. The USDA is direct about it: oven cord thermometers and thermometer forks cannot be calibrated, and a thermometer that can’t be adjusted may need replacing. That leaves two honest paths. You can note the error and cook with a mental offset — reads 2°F high in ice, subtract 2°F at the stove — which works right up until the night you forget mid-sear. Or you retire it. (Owners in that BBQ Brethren thread also note ThermoWorks offers re-certification for its own instruments, an option worth asking about before binning an expensive probe.)

If yours failed the test and can’t be adjusted, replacement costs less than one ruined brisket — and far less than a food-safety gamble. Our guide to the best kitchen thermometers walks the whole field, but the short version: the budget-friendly ThermoPop 2 is rated ±1°F across everyday cooking temperatures and ships with a NIST-traceable calibration certificate, while the spend-more Thermapen ONE reads in one second at ±0.5°F across its key ranges, with the same traceable cert. We weigh that upgrade — and the cheaper alternatives — in Thermapen vs ThermoPop vs the budget picks, and you can read how we evaluate gear on our testing methodology page.

How often should you check?

There’s no official interval for home cooks, so here’s the honest layering. ThermoWorks’ position is relaxed: dials drift from drops and other kitchen mishaps; digitals rarely need recalibration but deserve a double-check every so often. The BBQ community — which stakes competition entries on these tools — runs a stricter playbook, and the calibration guides circulating there suggest checking a thermometer when it’s brand new, after any drop, at the start of grilling season, every three to four months, and after any cook where the results didn’t match the numbers. Given the test costs five minutes and a tray of ice, the stricter habit is easy to endorse. And once more, from the forums: put in a fresh battery before you diagnose drift.

Your oven lies too (and the thermometer isn’t the problem)

Why the number on the dial is fiction

Your oven does not hold 350°F. It cycles its heating elements around 350°F, and the swings are bigger than most cooks suspect. ThermoWorks’ measurements: at a 350°F setpoint, the actual peaks and troughs can run as high as 400°F and as low as 300°F, and most home ovens sit 25 to 50°F off their setpoint. America’s Test Kitchen’s findings are just as damning — ovens set to the same temperature can vary from one another by as much as 90 degrees, and even a well-calibrated oven cycles roughly 25°F above and below its target. The built-in sensor won’t save you either: ATK notes it sits in an out-of-the-way spot that doesn’t reflect the temperature at the center rack, where your food actually lives.

None of this means your oven is broken — cycling is how thermostats work. It means the dial is an average at best, and a single glance at an oven thermometer tells you almost nothing, because you might be catching the peak or the trough of a 100°F cycle.

The 30-to-60-minute oven audit

The fix is measuring over time. ThermoWorks’ oven-test protocol, lightly simplified:

  1. Clip an air probe — a leave-in probe with a grate clip — to the middle of the center rack.
  2. Set the oven to 350°F and let it preheat and cycle for 30 minutes. Ignore this window; the oven is finding its rhythm.
  3. Note the lowest and highest readings across a second 30-minute window, then average them. The difference between that average and 350°F is your oven’s true offset.

A leave-in probe makes this nearly effortless — our leave-in probe thermometer guide covers the options. The two-channel Square DOT, which ThermoWorks pitches, in as many words, as a lie detector for ovens and smokers, has a 15-minute rolling Average Mode that does the math for you and carries an America’s Test Kitchen winner badge; the ChefAlarm pairs a leave-in probe rated to 700°F with a built-in timer and handles the same job. No leave-in probe? A basic dial oven thermometer works too — just take repeated readings across the 30-to-60-minute window instead of trusting one look.

What to do with the offset

If the audit says your oven runs 25°F cool, there are two fixes. The everyday one: adjust your setpoint — a recipe calling for 350°F gets dialed to 375°F, permanently. The tidier one: many ovens have a built-in calibration offset in their settings that shifts every setpoint at once; your manual will say. And one reassurance from the USDA while you’re at it: unlike boiling water, oven temperatures are not affected by altitude — sea-level oven times and temperatures apply whether you’re roasting in Miami or Boulder.

Why a few degrees matters (the food-safety stakes)

Calibration would be a nerdy hobby if the targets were fuzzy. They’re not. The USDA’s safe minimum internal temperatures are exact floors, and a thermometer that reads just 5°F high will call chicken done while the meat sits below the line.

FoodUSDA safe minimum internal temperature
Beef, pork, veal, lamb — steaks, chops, roasts; fresh ham145°F, plus a 3-minute rest
Fish145°F
Ground meats (beef, pork, veal, lamb)160°F
Egg dishes160°F
All poultry, including ground165°F
Leftovers and casseroles165°F

Two more USDA points worth pinning to the fridge: the bacterial danger zone runs from 40 to 140°F, and color is not a doneness indicator — a burger cooked to 160°F is safe regardless of the shade inside. For the full breakdown, including where chefs prefer to pull meat relative to the USDA floor and who should stick strictly to the official numbers, see our meat temperature chart.

Sources