Lifestyle-related illnesses, such as diabetes and obesity, have high social and economic costs. Globally, more than two-thirds (68.8%) of adults and almost 3 in 4 men (74%) are considered to be overweight or obese (1). In 2008, the costs related to obesity in the US alone were estimated to be $147 billion (2). The situation with respect to diabetes is no better. In 2015, 415 million people were estimated to have diabetes worldwide (1 in 11 adults), and the annual costs are estimated to be $673 billion (12% of global health expenditure) (3).

One of the major contributing factors to such illnesses is poor dietary habits, in particular diets high in sugar, carbohydrates, and fat and diets low in fiber (4–6). Considerable research attention and practical intervention measures have been taken in an attempt to improve dietary choices. One such measure has been to encourage people to cook for themselves at home. For example, both the US Government’s ChooseMyPlate initiative (7) and the UK Government’s Change4Life program (8) advocate home cooking. Poorer cooking skills, less-frequent preparation of home-cooked food, and more frequent consumption of pre-prepared foods have been associated with restricted quality of diet and obesity (9–11). Further evidence supporting the endorsement of home cooking comes from research demonstrating that the amount of food consumed away from home is linked with higher Body Mass Index and lower fruit and vegetable intake (12), leading the authors of one study to conclude that strategies are needed to encourage more cooking among the general population (12). The healthiness of home cooking will, nevertheless, depend on what people cook and how they cook it. These two variables correlated with sociodemographic factors and obesity (13). Thus, the solution may be more complicated than simply getting people to cook—we need to understand sources of cooking inspiration and the role these play in the dietary decisions people make.

Howard et al. (14) investigated the nutritional properties of two common sources of food. They compared recipes published by well-known UK celebrity chefs with leading UK supermarket ready meals, which are often presented as being unhealthy (7, 8). Ready meals, more commonly known as “TV dinners” in the United States, are pre-prepared main courses that can be reheated in their container, require no further ingredients, and need only minimal preparation before consumption. Surprisingly, Howard et al. (14) found that in some respects the ready meals were healthier than recipes. More ready meals than recipes met the WHO goals for fiber density and percentage of energy derived from carbohydrate and fat, although more ready meals than recipes exceeded the recommended sodium density. Thus, the source and, in particular, the content of the recipe seem to be more important than the type of meal (i.e., whether it is a ready meal or home cooked).

A further common source of cooking inspiration is the Internet (15). The food website, which, at the time of writing, claims to be the world’s largest food-focused social network, is Allrecipes.com. The site has a community of 40 million home cooks accessing 3 billion pages annually across 19 sites in 24 countries with recipes available in 13 languages (16). The British version of Allrecipes.com was named as the Daily Mail’s top pick for healthy eating websites, highlighting the “sophisticated search engine” and claiming that “… diabetics, coeliacs and even those specifically wanting to increase their fibre intake – are all catered for” (17).

The recommendation of a popular tabloid newspaper combined with government and media encouragement may persuade members of the public that cooking recipes sourced from the Internet is an approach likely to improve their diet, despite this no systematic study having comprehensively assessed the nutritional content of online recipes.

Using methods previously applied to evaluate the healthiness of ready meals and recipes published by celebrity chefs (14), we analyze the energy, protein, carbohydrate, fat, sugar, fiber, and salt content of recipes uploaded to Allrecipes.com and determine whether the nutritional content complied with national and international recommendations. We use the data collected from the previous analysis as a basis of comparison (14).

We carried out a cross-sectional analysis of the nutritional content of 5,237 recipes from the food website Allrecipes.com. The global version of the site was chosen rather than the British version as the British version alone included insufficient data for analysis. This is unlikely to be a significant source of bias as the global version of the site is popular in Britain as well as worldwide: the web analytics service Alexa.com ranks Allrecipes.com as the most popular cooking community platform (18).

The data describing the Allrecipes.com recipes was downloaded in Summer 2015 and contains 7648 recipes published between the years 2000 and 2010 on the Allrecipes.com website as main dishes.

Rather than sampling from the dataset (as was done with the ready meals and celebrity recipes), we used all of the recipes available, which meet the criteria applied in the previously published analysis (14). This was important as we wished to draw comparisons with the meals analyzed in the previous work. The following restrictions were applied:

we included only main dishes consisting of at least 225 g per portion, matching the restriction used in Howard et al.’s analyses.

Table 1 shows how these filters influenced the number of recipes analyzed. For the other categories of meal (TV chef recipes and supermarket ready meals), the data were provided by the lead author of an earlier study (14). Full details of the inclusion and exclusion criteria and sampling methodology for these recipes can be found in their paper (14). Ethical approval was not required for this analysis of published, publicly available information.

Table 2 shows the nutritional content per portion of Internet recipes created by users in Allrecipes.com compared to TV chef recipes and supermarket ready meals. The TV chef recipes contain more energy (2,530 v 2,113 kJ (Internet) v 2,066 kJ (ready)) and more protein (37.46 v 29.50 g (Internet) v 27.85 g (ready)) than both Internet recipes and ready meals (all comparisons sig. p < 0.01). Internet recipes are significantly lower in terms of carbohydrate (35.20 v 49.48 g (TV) v 51.05 g (ready), p < 0.01) and sugar (5.50 v 8.25 g (TV) v 6.80 g (ready), p < 0.01, p = 0.02) content, but significantly higher than TV chef recipes in terms of sodium density (829 v 660 mg, p < 0.01). There is no significant difference between the sodium content of Internet recipes and ready meals, which are known for unhealthy high levels of salt (21, 22) (829 v 800 mg, p = 0.59). In terms of fats and saturated fats, recipes from TV chefs contain the most (fat: 27.06 g, saturated fat: 9.20 g), but ready meals contain the least (fat: 17.20 g, saturated fat: 6.80 g) the difference between ready meals and Internet recipes being highly significant (fat: 17.20 v 24.80 g, p < 0.01; saturated fat: 6.80 v 8.80 g, p < 0.01). The fiber content of Internet recipes is low in comparison to the meals from the other sources (significantly lower than both TV chef meals 3.00 vs 3.44 g, p = 0.01 and ready meals 6.45 g, p < 0.01).

Table 3 and Figure 1 summarize the percentage of recipes/meals of different types, which meet the criteria established by the WHO in terms of the individual nutritional properties. In this case, the values are normalized with respect to the total number of calories in one portion. The data show that only 6 of the 5,237 Internet recipes (0.11%) and none of the sampled TV chef recipes or ready meals meet all 7 criteria. A small number of recipes of each type (5.9% of Internet recipes, 7% of TV chef recipes, and 1% of ready meals) do not meet any of the criteria at all. However, the majority recipes of all three types meet one or two criteria. A total of 46% of Internet recipes meet one criterion compared to 42% of TV chef recipes and 27% of ready meals. As the number of criteria to be met increases, the percentage of Internet recipes meeting these drops at a higher rate than the other types of meal. Figure 1 shows that Internet recipes and TV chef recipes met a similar mean number of WHO criteria (1.76 and 1.77, respectively; SDs 1.17 and 1.10, respectively). Ready meals met a mean of 2.37 criteria (SD 1.21). Using Mann–Whitney tests, the small difference between Internet recipes and TV chef recipes is non-significant (p = 0.64), but ready meals met significantly more WHO criteria than Internet recipes (p < 0.01).

Table 4 shows the number and proportion of each meal type that met each nutrient-specific WHO goal. More Internet recipes meet the WHO guidelines for protein than other classes of meal (10.88 v 7% (TV), p < 0.01, 10.86 v 9% (ready), p < 0.01). However, fewer Internet recipes met the criteria for fat (14.28 v 24 (TV) v 37% (ready), p < 0.01), saturated fat (25.05 v 33 (TV) v 34% (ready); p < 0.01), and fiber (when compared to ready meals 16.50 v 56%; p < 0.01). In terms of salt, significantly more Internet recipes met the criteria for sodium density than ready meals (19.63 v 4%, p < 0.01), but significantly fewer than the TV chef meals (19.32 v 36%, p < 0.01). For sugar, no differences between Internet recipes and TV recipes can be observed (81.1 v 81% (TV), p = 0.86). However, Internet recipes significantly differ from ready meals (81.1 v 83% (ready), p < 0.01).

Table 5 shows the traffic light assessment for the different recipe types according to modified Food Standards Agency guidelines (20). The FSA guidelines are based on macronutrient properties normalized by portion size. According to these guidelines, Internet and TV chef recipes have almost equal numbers of red labels (45 v 47% (TV)), but both TV recipes and ready meals have more green labels than Internet-sourced recipes (36 v 42 (TV) v 41% (ready)). Figure 2 shows these data averaged over all recipes of each type to provide simulated front of package labels for an average Internet recipe, an average TV chef recipe, and an average ready meal using a design based on FSA guidelines. Internet and TV recipes are classified as having high fat; all three categories are labeled as having low sugar, and both ready meals and Internet recipes are considered to contain medium salt.

Overall based on the FSA guidelines, the healthiest category is the ready meals and the unhealthiest category is Internet recipes, which is the same overall conclusion drawn when using the WHO guidelines.

To establish how stable these values are over time, we calculated the same statistics for each year of data available to us. Table 6 shows the nutritional content per portion of Internet recipes created by users in Allrecipes.com for the years 2000 to 2015. Similarly, Figure 3 presents the simulated front of package labels, based on the guidelines from the FSA (20), for an average Internet recipe created by Allrecipes.com users for each year during that time period. Overall both depict stable trends over time. The FSA labels for macronutrients, for the average recipe based on annually uploaded recipes, are the same for every year in the dataset. Figure 4 demonstrates the percentage of Allrecipes.com recipes meeting different numbers of WHO criteria overall, and Figure 5 depicts the same information at a macronutrient granularity. Both figures show limited annual variation but present an overall consistent trend.

Internet recipes sourced from Allrecipes.com tend to be high in protein, fat, saturated fat, and sodium, low in carbohydrate, and within the recommended range for sugar according to World Health Organization’s nutritional guidelines for the avoidance of diet-related diseases (6).

When compared to meals based on television chef recipes and ready meals from leading UK supermarkets, the Internet recipes were the least healthy. Significantly, fewer were within the recommended ranges for fat, saturated fat compared to meals from the other sources, and significantly fewer Internet recipes met the recommended range for fiber than ready meals. In terms of sodium density, significantly more Internet recipes met the criteria than ready meals, but significantly fewer than the TV chef meals. Internet recipes did, however, meet the criteria for protein content significantly more often than the other categories.

Internet recipes were also more likely to achieve red traffic light labels according to the criteria of the UK Food Standards Agency (FSA) (20). By investigating the criteria for each year of data collection, we revealed that the characteristics of recipes uploaded to the site are very stable on average. Thus, our findings, taken together with those of Howard et al., add weight to the argument that encouraging people to cook more at home might not, by itself, be enough to improve nutritional habits. This is because some common sources of cooking inspiration themselves promote unhealthy meals.

CT collected and analyzed the data. DE and CT drafted the manuscript, and both authors contributed to the study design and interpretation of results. SH commented on successive drafts of the manuscript and supported CT with the study design. CT and DE will act as guarantors.

All authors have completed the ICMJE uniform disclosure form at http://icmje.org/downloads/coi_disclosure.zip (available on request from the corresponding author) and declare: no support from any companies for the submitted work; no relationships with any companies that might have an interest in the submitted work in the previous three years; and no non-financial interests that may be relevant to the submitted work.