Exercising With a CGM: What Glucose Does During Workouts and How to Read It

Wearing a CGM through a workout is one of the most interesting things a sensor can show you, and one of the easiest things to misread. The graph is a delayed picture of a moving target. This guide explains what the research actually says about glucose and exercise in type 2 diabetes, what sensors can and cannot do at speed, how to read a workout graph without overreacting to it, and who should be talking to a clinician before they start experimenting.

I am not a physician, and this is general education. Everything below comes from the sources listed at the end, and where I could only read a summary or an abstract, I say so. Several of the best sources are focused on type 1 diabetes, and I flag that where it matters.

Who should read this, and who should talk to a clinician first

Most of this guide is for adults with type 2 diabetes or prediabetes who are curious about their own glucose patterns. The group that needs the most care is anyone taking insulin or other medicines that can cause low blood sugar. Mayo Clinic advises people who take insulin or other medicines that can lower blood sugar to test 15 to 30 minutes before exercising, and says not to exercise if you needed help recovering from a serious low in the past 24 hours. Those are Mayo’s statements for people on those medicines, and the exact plan, including whether and how your medicine changes on workout days, is a decision for your prescriber. I do not give dosing or medication-adjustment advice here.

There is also an eligibility point on the sensor side. The FDA review document for the Dexcom Stelo, the first over-the-counter CGM, describes it as for people 18 and older who are not on insulin, says not to use it if you have problematic hypoglycemia or are on dialysis, and says users should not take medical action based on its output without consulting a healthcare professional. It also has no glucose alerts or alarms; the app shows notices such as signal loss or sensor expiry, and readings display between 70 and 250 mg/dL, with values outside that range shown as “Below 70” or “Above 250.” So for exercise, an over-the-counter sensor like that is a learning tool, not a safety monitor. Our guides on which CGM fits which goal and Stelo versus FreeStyle Libre Rio go through the differences.

What a CGM actually measures during exercise

A CGM does not measure blood glucose. It measures glucose in interstitial fluid, the fluid between cells, and estimates blood glucose from it. When glucose changes slowly, the two track closely. When it changes quickly, the fluid reading follows the blood with a delay. The Accu-Chek educational page on lag time says that in steady conditions the lag is about 5 to 20 minutes depending on sensor placement, can stretch to 12 to 24 minutes during rapid change such as after eating, exercising or taking insulin, and can reach up to 40 minutes depending on the sensor. A diaTribe article by Dr. Francine Kaufman summarizing a consensus position statement on CGM and exercise gives similar numbers: about 5 minutes at rest and 12 to 24 minutes during exercise, and it notes that exercise also raises the sensor’s average error, so readings are less accurate.

Table note: The Stelo numbers come from the FDA review document for that one product. Other sensors have their own accuracy data and I did not review them for this guide. The “average error” figures are mean absolute relative difference against a lab comparator in the study described in that document, which is a different thing from how a sensor behaves on your arm during a hard interval.

Here is the practical translation. If your glucose is falling quickly while you exercise, your actual blood glucose may already be lower than the number on the screen. If it is rising quickly after a hard effort, the screen may be behind in the other direction. The sensor is most trustworthy when the line is flat and least trustworthy at the exact moments people most want an answer. If you feel shaky, weak, dizzy or confused, Mayo lists those as signs to stop and check, and a symptom outranks a number that may be 12 to 24 minutes old. A fingerstick is the usual way to confirm; our guide on CGM versus a fingerstick meter compares what each is good for.

What research says glucose does with different kinds of exercise

The cleanest statement I found comes from the consensus summarized by diaTribe, and it is worth repeating who it is for. It is a position statement from several diabetes organizations that the diaTribe article describes as focused on type 1 diabetes, likely applicable to people with type 2 diabetes who are on intensive insulin regimens and use a CGM. I did not read the statement itself, and the diaTribe page did not name its authors or give a date, so I cannot attribute it more precisely than that. With those limits, it says mild to moderate aerobic exercise such as brisk walking, running, swimming or cycling usually lowers glucose, while intense aerobic or anaerobic interval exercise, such as weightlifting or sprinting, tends to stabilize or raise it. Mayo Clinic makes a similar point from the other side: for people checking before a workout, blood sugar can rise during strength training or high-intensity interval training.

What does the research say when the outcome is the 24-hour CGM trace in people with type 2? The best summary I found is a 2020 meta-analysis in Frontiers in Endocrinology by Munan and colleagues. It pooled 28 studies that used CGM, 23 short-term and 5 longer-term. The short-term studies included 373 participants. Compared with a no-exercise control condition, mean 24-hour glucose was lower by 0.5 mmol/L (95% confidence interval 0.3 to 0.7), which is about 9 mg/dL, with a range of roughly 5.4 to 12.6 mg/dL on my conversion. Time above the high range fell by 94 minutes per day, about 1 hour 34 minutes, and glycemic variability, measured as mean amplitude of glycemic excursions, fell by 0.41. Fasting glucose and time spent low did not change significantly. Those are averages across studies of different designs, with high heterogeneity, and the authors rated some included trials as high risk of bias.

For longer exercise programs, the picture is murkier. The five longer-term studies did not show a statistically significant drop in mean 24-hour glucose against control (a difference of 0.9 mmol/L, about 16.2 mg/dL, with an interval that ran from a large drop to a small rise), though glucose was lower than before the program began. Three of those five studies had no control group. I would not read that as “training does not help.” It is a reminder that CGM-measured effects of one workout are better documented than long-term effects, and that this particular review could not settle the longer question.

One subgroup finding gets repeated online, so it deserves a careful sentence. In that meta-analysis, exercise done fasted or in the morning was associated with larger drops in 24-hour glucose than exercise done in the afternoon. The authors say participants were not randomly assigned to exercise timing, so the data cannot show timing caused the difference, and some of those analyses were not pre-specified. Treat it as an idea to test on your own graph, not a rule.

For intervals, the evidence I could verify is small. A 2018 randomized four-condition crossover study in the European Journal of Applied Physiology (Metcalfe and colleagues) followed 11 men with type 2 diabetes, average age 52 and HbA1c 7.0%. Each did no exercise, 30 minutes of continuous exercise, ten one-minute intervals at about 90% of maximum heart rate, or two 20-second all-out sprints, while wearing a CGM under a standardized diet. Compared with no exercise, the two-sprint protocol lowered mean 24-hour glucose by 0.58 mmol/L (about 10.4 mg/dL), which was statistically significant. The 30-minute continuous session lowered it by 0.37 mmol/L (about 6.7 mg/dL), a trend that did not reach significance, and the ten-minute interval protocol also showed 0.37 without significance. Time above the high range fell in all three exercise conditions. The authors call for further study, and 11 people is a small sample. It is a useful reminder that short, hard efforts can work in the data, not evidence that you should do sprints.

Reading a workout graph without overreacting

A workout graph can look dramatic, and the right question is usually “what is the trend and what else changed?” rather than “what is the number?” Trend arrows and rate-of-change indicators are defined by each manufacturer, so I will not quote their numeric thresholds here; check your own app’s guide. The consensus summary suggests thinking in ranges and trends together. For people using CGM with insulin, it describes a target of 90 to 180 mg/dL during exercise, or 126 to 180 for prolonged aerobic exercise, and 80 to 180 afterward, with a suggestion to stop exercising below 70 mg/dL. Those numbers are written for type 1 diabetes and insulin users. They are not targets for everyone, and your own range is something to set with your clinician.

Here is a worked example with made-up numbers, built so the arithmetic is visible. Imagine a CGM that reports every 15 minutes during a 60-minute walk. Readings: 9:00 (before) 142 mg/dL, 9:15 138 mg/dL, 9:30 121 mg/dL, 9:45 104 mg/dL, 10:00 (end) 96 mg/dL, 10:15 99 mg/dL, 10:30 108 mg/dL. Between 9:15 and 10:00 the displayed glucose fell from 138 to 96 mg/dL, a drop of 42 mg/dL over 45 minutes, or about 0.93 mg/dL per minute. If the 12 to 24 minute lag from the sources above applied and that rate had continued, actual blood glucose could differ from the displayed number by roughly 11 to 22 mg/dL. That is an illustration of why lag matters, not a prediction, because real glucose rarely falls in a perfectly straight line. Notice what the example does not tell you: it does not tell you whether 96 mg/dL is “good.” It tells you the direction, the speed, and why a person who feels fine at 96 and falling might still choose to check.

Table note: This is simple multiplication with an assumed constant rate and the 12 to 24 minute range from the Accu-Chek page, to show the size of the effect. Real lag varies by sensor, placement and person.

A second worked example shows how to read a day, not a moment. Suppose a hypothetical day has 96 sensor readings, one every 15 minutes, and 14 of them are above 180 mg/dL, which is 3.5 hours. The share in range (70 to 180) would be 85.4%. If a walk is associated with 4 fewer readings above 180 on a similar day, one hour less, the share in range would be 89.6%. That is arithmetic with invented inputs. The point is that “time in range” is a day-level number, so one workout moves it a little and consistency moves it more, and a CGM is a way to see whether your own days look like that. Our guide on how to read your CGM data explains time in range, averages and variability.

After the workout: delayed lows, nighttime, and compression lows

If you take medicines that can lower blood sugar, the hours after exercise matter. Mayo Clinic says low blood sugar can happen even 4 to 8 hours after exercise, and that tougher workouts have a longer effect. A CGM is useful here because it can show a slow downward drift that a single afternoon fingerstick would miss. For people using a CGM with alerts, the consensus summary suggests raising the low alert at night after exercise and having carbohydrate available; I am noting that as a topic to discuss with your clinician rather than a plan.

A different problem is the false low. The University of Michigan Depression Center’s CGM knowledge base explains that pressure on a sensor, such as sleeping on it, reduces the fluid the sensor samples and can produce a falsely low reading. It typically looks like a steep, sudden drop on a graph that was otherwise steady, with no exercise or fast-acting insulin to explain it. Its prevention advice is to tell people about compression lows, avoid sleeping on the sensor, and place it where you are unlikely to lie on it when the manufacturer allows. If a hard workout day is followed by a mystery low at 3 a.m. and you were lying on the sensor, compression is a possible explanation, but confirming it with a fingerstick is the safer way to find out.

Sweat and movement create their own problems. A sensor that peels half off during a run or a swim stops giving useful readings. Our article on what to do when a CGM sensor will not stick covers skin prep, overpatches and removal using manufacturer guidance.

A simple workout log for your CGM

If you want to learn from your own graph, write down a few things that change from day to day. A log is also what you hand your clinician when the question is “why did my glucose do that?”

A caution about the log: do not turn it into a score. One bad day on the graph is a data point, not a failure, and our post-meal glucose tracker can hold the same kind of notes for meals if you want them side by side.

Where Charles is with exercise

Charles has said that being diagnosed changed how he treats exercise. He has described realizing he did not have enough muscle mass, starting on a bike, then lifting at a gym with a friend, then running into limited range of motion and tight hamstrings. He is training for his first 5K, the Sweet Tea 5K in Somerville on November 11, after about eight weeks of preparation. I am not sharing any CGM results from that training here. His own experience is in running with type 2 diabetes and strength training with type 2 diabetes.

Scenarios

If you have prediabetes

You may be using an over-the-counter sensor mainly to see how your body responds to food and movement. A reasonable way to use it is to compare similar days: the same walk on a day you ate earlier versus later, or a strength day versus a rest day. The meta-analysis above suggests short-term exercise can lower average glucose in type 2 diabetes, but it did not study prediabetes, so I would not assume the same size of effect. Keep the sensor in perspective: it shows patterns; your clinician interprets them.

If you have type 2 and are not on medicines that cause lows

Low blood sugar during exercise is less of a concern in this group, but not impossible, and symptoms still count. The more useful questions are about patterns: does your glucose tend to rise during hard intervals and come back down afterward? Does a 10-minute walk after dinner flatten the evening bump? Our guide on walking after meals goes through the research on that specific habit.

If you take insulin or a medicine that can cause lows

Your clinician should set your starting range, your stop point, your carbohydrate plan and any change to your medicines on exercise days. The type 1 consensus numbers above are context, not instructions. Do not change a dose because of anything you read here. If you have a prescription CGM with alerts, ask whether your alert thresholds should be different around workouts.

If you already wear a CGM and are new to exercise

Start easy. A walk, a short bike ride or a gentle strength session gives you a graph that is easy to read, with a smaller chance of a confusing spike from an all-out effort. Add one variable at a time. Our posts on yoga, cycling, swimming and HIIT take each activity in turn.

Common mistakes

Treating the number as the whole story. A reading is a delayed estimate. Check the trend and how you feel.

Reacting to a compression low. A steep, sudden drop after you lay on the sensor may be false. Confirm with a fingerstick before treating, if your clinician has told you how to handle lows.

Using a sensor with no alerts as a safety net. The Stelo has no glucose alarms, per the FDA review document, so it will not warn you.

Comparing different conditions. A fasted morning run and an evening session after dinner are different experiments.

Quoting type 1 targets for type 2. The consensus ranges above were written for a different population.

Skipping the log. Without notes you will not remember whether it was the workout, the meal or the bad night that moved the line.

Ignoring how you feel. Mayo lists shaky, weak, dizzy or confused as reasons to stop and check, whatever the screen says.

Myth versus fact

Questions to bring to your clinician

What glucose range should I aim for before, during and after exercise? Do any of my medicines raise my risk of a low after a workout, and for how many hours? What should I carry and how should I treat a low? At what number should I stop? Should my sensor alerts be set differently on workout days? Is an over-the-counter sensor appropriate for me, or do I need a prescription one? Would my kidney, eye or heart history change what kind of exercise is safe?

If you want to share a graph, bring the log from above and a few days of readings, not a single screenshot.

Frequently asked questions

Does exercise lower blood sugar in type 2 diabetes?

Often, yes. A 2020 meta-analysis of short-term CGM studies found that exercise was associated with about 0.5 mmol/L (roughly 9 mg/dL) lower 24-hour glucose and 94 fewer minutes per day above the high range. Averages hide wide differences between people, and hard intervals can raise glucose temporarily.

How accurate is a CGM during exercise?

Less accurate than at rest. A diaTribe summary of a consensus position statement says sensor error rises during exercise and lag can reach 12 to 24 minutes, and the Accu-Chek lag page says up to 40 minutes on some sensors. Confirm with a fingerstick if you feel unwell or the reading does not match how you feel.

Why does my CGM show a low after a workout?

It may be a real post-exercise drop, which Mayo says can occur 4 to 8 hours after exercise, or a false compression low if you were lying on the sensor. A sudden steep drop on an otherwise steady graph is the pattern the University of Michigan page describes for compression. Confirm before treating.

Does HIIT raise blood sugar?

It can during or right after, according to the consensus summary and Mayo. In one small crossover study of 11 men with type 2, a two-sprint session lowered 24-hour glucose compared with no exercise, while a ten-interval session did not reach statistical significance. That is not enough to generalize.

Should I exercise fasted or after eating?

The 2020 meta-analysis found larger average drops with fasted or morning exercise, but timing was not randomized, so it cannot say fasting caused the difference. If your medicines can cause lows, ask your clinician before exercising fasted.

Can I use Stelo for workouts?

The FDA document says Stelo is for adults 18 and older who are not on insulin, has no alerts, and says not to use it with problematic hypoglycemia or dialysis. It can show your pattern, but it will not warn you of a low.

What glucose level is too low to exercise?

Mayo says to stop exercising at 70 mg/dL or lower, or if you feel shaky, weak, dizzy or confused, and its table lists under 90 mg/dL as possibly too low to start exercising safely. Your own thresholds should come from your clinician.

How long does the lag last?

About 5 minutes at rest in one source, 5 to 20 minutes depending on placement in another, 12 to 24 minutes during rapid change, and up to 40 minutes on some sensors.

Do sensors fall off when I sweat?

Sometimes. Sweat, friction and water can loosen the adhesive. See our article on sensors that will not stick for manufacturer-based tips.

How do I know if my graph change was from the workout?

You usually cannot from one day. Compare several similar sessions and note meals, sleep, stress and medicines in a log.

Sources

• Munan M et al., Acute and chronic effects of exercise on continuous glucose monitoring outcomes in type 2 diabetes: a meta-analysis. Frontiers in Endocrinology, 2020 (28 studies; read in full text on the publisher site)
• Metcalfe RS et al., Extremely short duration interval exercise improves 24-h glycaemia in men with type 2 diabetes. European Journal of Applied Physiology, 2018 (abstract record read via a bibliographic database)
• diaTribe, Dr. Francine Kaufman: Exercise well with your CGM (summary of a consensus position statement; the statement itself was not read)
• Accu-Chek Australia: lag time and CGM
• Mayo Clinic: diabetes and exercise, when to monitor your blood sugar (January 9, 2024)
• FDA 510(k) review K234070: Dexcom Stelo
• University of Michigan Depression Center knowledge base: CGM reading low values at night (August 2023)
• AJMC: American Diabetes Association issues recommendations for physical activity and exercise (October 31, 2016)

I could not open the PubMed or PubMed Central records for several of these, so figures come from the publisher or news pages named. The numbers in the worked examples are invented for illustration.

About the author: Charles Kirkland is the founder of Metabolic Clarity Labs and author of The Journey Back. He is not a physician. This is general education, not a medically reviewed treatment plan.

Sources checked October 31, 2026. Discuss personal decisions with a qualified clinician. Do not start, stop, or change prescribed treatment based on this article.