Honey and Diabetes: What the Research Shows

Measured against table sugar under controlled conditions, honey behaves like what it is: a sugar. About 82% of it by weight is sugars, chiefly fructose and glucose; the two human studies that measured its glycaemic index put every honey tested in the medium or high band, none in the low; and in the one trial that fed adults honey, sucrose and high-fructose corn syrup in turn, the three proved metabolically indistinguishable over two weeks. The record has a genuinely contested corner, and this page reports it in full: one meta-analysis of controlled trials found a small fall in fasting glucose on honey, at low certainty and mostly against usual-diet rather than sugar comparators, while a later umbrella review found fasting glucose moving the other way. What follows walks through the composition, the glycaemic-index measurements, the head-to-head trial, the three syntheses that disagree, and how UK guidance and the regulator read the same evidence.

What any of it means for an individual’s diet, medication or glucose targets is a conversation for that person and their diabetes team — a GP, a diabetes nurse, a registered dietitian. What the research measured is set out below.

Plate I · PaintingRecord

What is honey actually made of?

Sugar and water, in that order. The standard reference analysis (USDA FoodData Central, record 169640) gives 100 g of honey as 82.4 g of carbohydrate, of which 82.1 g is sugars: about 40.9 g fructose, 35.8 g glucose and 0.89 g sucrose. The rest is 17.1 g of water and 0.3 g of protein, at 304 kcal, with vitamins and minerals at trace level.

Fructose raises blood glucose less sharply than glucose does, and the popular case for honey rests on that difference — but honey holds roughly as much glucose as fructose, and the measurements below were taken on whole honey.

Is honey low-GI?

Not in any human study that has measured it — though the measured values spread wider than a single study suggests, and the spread is worth setting out in full. The glycaemic index is measured directly in people: a food portion providing 50 g of carbohydrate is eaten, blood glucose is tracked, and the response is expressed against the same person’s response to 50 g of pure glucose, scored 100.

Ischayek and Kern (2006) determined the glycaemic index of four US honeys — clover, buckwheat, cotton and tupelo — in 12 healthy adults of mean age 24.5 years, each honey served as a 250 mL solution providing 50 g of carbohydrate and measured against the same participant’s response to 50 g of glucose. The reported means were 69.2 for clover, 73.4 for buckwheat, 73.6 for cotton and 74.1 for tupelo, with standard errors of 6.4 to 8.2. No statistically significant difference between the honeys appeared, and no relationship between glycaemic index and the fructose-to-glucose ratio was detected; the authors read this as showing that “small differences in fructose-to-glucose ratios do not substantially impact honey glycemic index”.

A second study measured two other honeys and came out lower. Robert and Ismail (2009) served eight healthy volunteers — five men, three women, aged 24 to 44, of normal BMI — 50 g carbohydrate portions of Malaysian wild honey and of Australian honey, against a glucose reference tested three times. The mean index was 65 ± 7 for the Malaysian wild honey and 59 ± 5 for the Australian; the two did not differ from each other, but both sat significantly below glucose at 100 (P < 0.001). The authors concluded that both are “intermediate GI foods”. Eight volunteers is a very small panel, and none of them had diabetes.

Diabetes UK sets out the conventional bands: low GI is “55 or below”, medium “56 to 69”, high “70 or more”. Across the two studies, the six measured values run from 59 to 74.1: three fall in the high band, three in the medium band, and none in the low band. The same page states the framework’s limits: GI “doesn’t take into account your portion sizes which can have a bigger impact on your blood sugar levels”, and “the amount of carbs you eat has a bigger effect on blood sugar levels than GI alone”. Both study populations were small groups of healthy adults, not people with diabetes.

Fig. I · Six values on the band scaleDiagram

Six measured glycaemic-index values · Diabetes UK bands · glucose reference 100

Empty — no measured value falls in the low band

5570100

Robert & Ismail (2009) · eight healthy volunteers · against a glucose reference tested three times

• Australian honey59 ± 5Medium band
• Malaysian wild honey65 ± 7Medium band

Ischayek & Kern (2006) · four US honeys · 12 healthy adults · 50 g of carbohydrate against 50 g of glucose

• Clover69.2Medium band
• Buckwheat73.4High band
• Cotton73.6High band
• Tupelo74.1High band

Standard errors of 6.4 to 8.2 are reported across these four together; the page gives no figure for any one of them. Each bar is therefore drawn solid to ± 6.4 and dashed out to ± 8.2.

Reported meanThe published ± value, per honeyStandard error, published only as a range across four honeys

Low · 55 or below · no measured valueMedium · 56 to 69 · three valuesHigh · 70 or more · three values

Diabetes UK, on the framework’s limits: GI “doesn’t take into account your portion sizes which can have a bigger impact on your blood sugar levels”, and “the amount of carbs you eat has a bigger effect on blood sugar levels than GI alone”.

Is honey better than sucrose for blood glucose?

The best-controlled comparison found no difference — and its design is what makes it the best-controlled. Raatz, Johnson and Picklo (2015) ran a randomised crossover trial in 55 adults — 28 with normal glucose tolerance, 27 with impaired glucose tolerance — who each consumed, in random order, 50 g of carbohydrate a day from honey, sucrose, or high-fructose corn syrup containing 55% fructose, two weeks per sweetener with a washout between. A crossover means every participant took all three sweeteners in turn and served as their own control, which strips out the differences between people that can blur a parallel-group trial.

Glucose, insulin, HOMA-IR (a calculated index of insulin resistance) and the incremental areas under the curve for glucose and insulin on oral glucose tolerance testing did not differ by treatment. Body weight was unchanged. Triglycerides “increased significantly from pre- to post-treatment in response to all sugars tested”, and C-reactive protein rose in the impaired-tolerance group on all three. The authors’ conclusion: fourteen days of the three sweeteners “resulted in similar effects on measures of glycemia, lipid metabolism, and inflammation”.

Two weeks is short, 55 participants is few, and impaired glucose tolerance is not diagnosed diabetes. Within those limits, the three came out alike.

What have trials in people with diabetes found?

Mixed — and the three syntheses of this literature disagree with one another in ways that reward a close look.

The most tightly controlled synthesis is by Ahmed and colleagues (2023), a systematic review and meta-analysis restricted to controlled trials of at least one week: 18 trials, 33 trial comparisons, 1,105 participants, median duration 8 weeks and median age 41.2 years. Pooled, honey reduced fasting glucose by 0.20 mmol/L (95% CI −0.37 to −0.04). It also reduced total cholesterol by 0.18 mmol/L, LDL cholesterol by 0.16 mmol/L, fasting triglycerides by 0.13 mmol/L and alanine aminotransferase by 9.75 U/L, and raised HDL cholesterol by 0.07 mmol/L. Subgroup differences appeared by floral source and by processing: robinia, clover and raw honey were the subgroups in which the fasting-glucose and total-cholesterol reductions showed up. The authors’ conclusion was that honey “may improve glycemic control and lipid levels when consumed within a healthy dietary pattern”, and they went further than current guidance does: “The results do not support the consideration by policymakers and those who issue guidelines to designate honey as a free sugar”.

Four things bound that result, all recorded in the review itself, and each is a lesson in how to read a pooled number. The certainty of evidence was graded low for fasting glucose and for every outcome above except HDL cholesterol, which was graded high. Fasting glucose was downgraded for substantial unexplained heterogeneity (I² = 76.8%) and for publication bias — a trim-and-fill adjustment, the standard correction for missing negative studies, removed the statistical significance of the fasting-glucose effect altogether. The comparator in 70% of trial comparisons was the participant’s usual diet; only 15% used sucrose and 6% high-fructose corn syrup, so this is mostly honey measured against no honey, not against table sugar. And only 21% of the studies were in people with type 1 or type 2 diabetes, against 42% in healthy participants of mixed weight. A pooled shift of 0.20 mmol/L, at low certainty, that does not survive a publication-bias adjustment is a slender foundation.

An umbrella review published in 2025 reaches an equally low ceiling from a different direction. Norouzzadeh and colleagues reviewed the existing systematic reviews and the primary randomised trials of honey, royal jelly and propolis — 69 randomised trials and 3,544 participants across all three products, 16 of the trials on honey — grading the evidence by AMSTAR-II, GRADE and ICEMAN. Pooled, honey was associated with a fall in HbA1c (weighted mean difference −0.25, 95% CI −0.45 to −0.05; I² = 58.6%), rated very low certainty of evidence. The reviewers then undercut their own result: the effect appears to be driven by a single study, and “when this study was excluded, there was no significant change in HbA1C levels” (WMD −0.23; 95% CI −0.51 to 0.03). Their dose–response modelling put the largest HbA1c reduction at a daily 10 g (WMD −0.37; 95% CI −0.56 to −0.18), diminishing at 50 g and above.

The same modelling runs against honey elsewhere. Each daily gram was associated with a systolic blood pressure rise of 0.09 mmHg (95% CI 0.03 to 0.14); fasting glucose rose at a daily 20 g; and interventions of 12 weeks or more “were associated with increased FBG and AST levels” — fasting blood glucose and a liver enzyme. The authors’ summary sentence carries both halves: “10 g of honey daily may lower Hemoglobin A1C but adversely affect systolic blood pressure, Aspartate transferase, triglycerides, fasting blood glucose, and high-sensitive C-reactive protein”. They add that “caution is advised for long-term use due to potential adverse effects on FBG, TG, liver enzymes, BP, and inflammatory markers”, and record high heterogeneity “likely due to variability in participants’ dosages, durations, and health statuses”. That 10 g figure is a point on the reviewers’ modelled dose–response curve, not the result of any single trial.

The third synthesis is narrower in method. A review by Akhbari and colleagues (2021) identified 13 trials of eaten honey against metabolic outcomes in adults with type 2 diabetes and in non-diabetic adults. Eight of the 13 had no placebo or control group; on the Jadad scale, five were of acceptable methodological quality and eight of low quality. The reviewers concluded that honey “might have no significant effects on the modulation of metabolic profiles in nondiabetic subjects”, and that “a high intake of honey might increase glucose levels and worsen other metabolic parameters in patients with T2DM” — qualified immediately by their own caution that “due to substantial heterogeneity in study design and limited clinical trials, results, however, should be interpreted with great caution”.

At least two of the 13 ran the other way, and belong in the record. The review reports that in Rasad and colleagues (2014) — a randomised double-blind trial in 60 healthy young adults, 80 g of honey in 250 mL of water a day against sucrose — honey “decreased serum levels of FBS with no changes in blood pressure”; and that in Yaghoobi and colleagues (2008) — 60 overweight or obese participants, 55 completing, 70 g a day for 30 days, again compared with sucrose — honey “reduced only BMI and FBS with no changes in lipid profile in healthy subjects”. Neither trial was in people with diabetes, neither ran longer than six weeks, the review records directions of change rather than effect sizes for either, and both sit in the set of 13 in which 8 had no placebo or control group and 8 were of low quality on the Jadad scale.

Set side by side, the three do not converge, and the disagreement itself is the finding. The controlled-trials meta-analysis reports a small pooled fall in fasting glucose — 0.20 mmol/L, low certainty, mostly against usual-diet comparators, and not surviving a publication-bias adjustment. The 2025 umbrella review reports fasting glucose moving the other way, alongside a very-low-certainty fall in HbA1c that disappears when a single study is removed. The narrative review’s reading for type 2 diabetes also runs against honey, on a trial set in which 8 of 13 had no control group. What none of the three delivers is a large, long, controlled trial in people with diabetes comparing honey against the sugar it is supposed to replace. The general honey evidence page reads the wider metabolic record; the page on honey and weight takes the calorie question.

What do Diabetes UK and the NHS actually say?

Both are quoted, not paraphrased. Diabetes UK opens its page on sugar with “You don’t need to cut out sugar from your diet if you have diabetes”, then places honey: added sugars “are called ‘free sugars’ and they are also present in pure fruit juices, smoothies, syrups and honey”. It gives a ceiling — “The maximum recommended daily amount of sugar is 30g for adults – which works out at just seven teaspoons a day”.

The NHS free-sugars page is identical on the point, listing “Sugars in honey, syrups (such as maple, agave and golden), nectars (such as blossom), and unsweetened fruit juices, vegetable juices and smoothies”, then adding: “The sugars in these foods occur naturally but still count as free sugars.” Its limits run from 30 g of free sugars a day for adults down to 10 g at age 1.

Two further lines complete the picture. NICE’s guideline on type 2 diabetes in adults (NG28) opens its dietary section by asking clinicians to “provide individualised and ongoing nutritional advice from a healthcare professional with specific expertise and competencies in nutrition”, and adds a recommendation to “discourage adults with type 2 diabetes from using foods marketed specifically for people with diabetes”. Diabetes UK’s glycaemic-index page closes: “speak to your GP about a referral to a registered dietitian.” That is the address for the individual version of this page’s question.

How did the regulator read honey and blood sugar?

As not proven — and the reasoning is instructive. Under retained Regulation (EC) 1924/2006, only claims authorised on the Great Britain nutrition and health claims register may be used in commercial communications. The register holds eight honey entries; every one is non-authorised, and none concerns blood glucose. Most were rejected because honey “is not sufficiently characterised for a scientific assessment of this claimed effect”, per the EFSA assessments cited in the register (EFSA Journal 2011;9(6):2243 and 2010;8(2):1484) — the assessors’ point being that honey varies too much from jar to jar for any single claimed effect to be pinned to it.

The register does authorise blood-glucose claims — for other things, which shows what the evidential bar looks like when it is cleared. Fructose carries one: “Consumption of foods containing fructose leads to a lower blood glucose rise compared to foods containing sucrose or glucose”, usable only where glucose or sucrose has been replaced by fructose so that their content falls by at least 30%. Honey is not such a food; in its natural state it holds roughly as much glucose as fructose. Beta-glucans from oats and barley, pectins, resistant starch and the sugar replacers hold similar authorised claims under stated conditions. Honey holds none.

Claims about reducing the risk of a disease sit in a separate category under Article 14, requiring specific authorisation. None exists for honey, or for any bee product. Wordings of the form “suitable for diabetics” have been assessed for other foods — glucomannan, acacia gum, wheat dextrin, dietary fibre — and rejected as unsubstantiated every time.

Fig. II · The post-prandial glucose columnDiagram

Great Britain nutrition and health claims register · post-prandial glucoseAuthorised · with conditions

• FructoseAuthorised

  “Consumption of foods containing fructose leads to a lower blood glucose rise compared to foods containing sucrose or glucose”

  ConditionUsable only where glucose or sucrose has been replaced by fructose so that their content falls by at least 30%.

• Beta-glucans from oats and barleyAuthorised

  ConditionA similar authorised claim under stated conditions. This page does not set those conditions out, so none is printed here.

• PectinsAuthorised

  ConditionA similar authorised claim under stated conditions. This page does not set those conditions out, so none is printed here.

• Resistant starchAuthorised

  ConditionA similar authorised claim under stated conditions. This page does not set those conditions out, so none is printed here.

• The sugar replacersAuthorised

  ConditionA similar authorised claim under stated conditions. This page does not set those conditions out, so none is printed here.

HoneyNo entry

The frame is empty — honey holds no claim in this column at all

Every entry above is authorised under stated conditions honey does not meet.

Eight honey entries stand on the register; every one is non-authorised, and none concerns blood glucose.

Nor does the fructose condition reach it: honey is not such a food — in its natural state it holds roughly as much glucose as fructose.

Only claims authorised on the register may be used in commercial communications.

What else is in the record?

One safety line is absolute. NHS guidance is that honey occasionally contains bacteria that can produce toxins in a baby’s intestines, “leading to infant botulism, which is a very serious illness”, and that honey is not to be given to a child under 1 year old. The same page adds a line that runs through this article: “Honey is a sugar, so avoiding it will also help prevent tooth decay.” Allergy to eaten honey is rare, documented at case-report level, and not specific to diabetes; the general honey evidence page carries that record and the source behind it.

The record, then: a composition that is four-fifths sugars; six measured glycaemic indexes, none of them low; a two-week three-way comparison in which honey, sucrose and high-fructose corn syrup behaved alike; three syntheses that contradict one another on fasting glucose at low or very low certainty; and a register that authorises a blood-glucose claim for fructose but nothing for honey. What the evidence does not contain is a demonstration that honey is a different kind of sugar. How that bears on an individual’s condition, medication or blood glucose targets is a conversation for their diabetes team.