What is dietary filtering, and how does an app flag ingredients?

What is dietary filtering?

Dietary filtering is the removal of recipes that break a rule you have set, so a search for dinner never shows you something you cannot eat. Set vegan, and recipes with chicken disappear.

It works cleanly on a catalogue, where every recipe is tagged in advance. It works less well on a personal library, because the recipes came from a video caption and a photographed cookbook page, and nobody tagged them.

The deeper limitation is that filtering treats a recipe as pass or fail. Most recipes are neither: they need 1 ingredient changed. Removing them from view means never cooking dishes that were 1 substitution away from working.

What does flagging do differently?

Flagging reads the ingredient list against your rules, marks the specific lines that break them, and proposes a replacement for each. The recipe stays visible and becomes a decision instead of a dead end.

In practice that turns a library of 200 recipes into 200 candidates rather than 40. A stir-fry with soy sauce is a gluten-free dinner as soon as you know to use tamari, and the app naming that line is what closes the gap between knowing your restriction and cooking around it.

What are the ingredients people miss?

The obvious ones look after themselves; nobody halal-cooking misses pork. The problems are ingredients whose names give nothing away, and they are consistent enough to list.

• Gelatine in marshmallows, some yoghurts and many desserts. Matters for halal, kosher, vegetarian and vegan cooking.
• Fish sauce and anchovy in Thai curries, Worcestershire sauce and Caesar dressing. Catches vegetarians constantly.
• Whey and casein in bread, crisps and processed foods, which are dairy despite the labels not saying milk.
• Malt and malt extract in cereals and some sauces, which is barley, so not gluten-free.
• Alcohol in vanilla extract, some sauces and many desserts, which matters for halal cooking.
• Mono- and diglycerides and some emulsifiers, which may be animal-derived and are not identifiable from the recipe.

That last one is the honest boundary of what software can do. An app reads a recipe's ingredients; it cannot know which brand of margarine you bought. Adapt recipes for food allergies covers the label-reading half.

What makes a good substitution?

A swap has to match what the ingredient was doing, not just what category it belonged to. Replacing butter with olive oil works in a tray of vegetables and fails in shortbread, because in one it is fat and in the other it is structure.

1. 1Work out the function. Fat, structure, moisture, acid, salt, thickening or browning.
2. 2Match that function first. Sweetness matters less than whether the dish sets.
3. 3Check the water content. Swapping butter for oil adds moisture; swapping cream for coconut milk changes it too.
4. 4Expect a taste change and accept it. Tamari for soy sauce is close, and coconut cream for dairy cream is noticeable.
5. 5Do not restructure baking. In cakes and pastry, use a swap intended for the job rather than improvising.

The general skill is covered in substitute ingredients, and the dairy case specifically in dairy-free substitutes. Both matter more than any app feature, because you are the one deciding whether the result is worth eating.

How does this work on your own recipes?

Differently from a catalogue, and this is where most apps quietly fall short. A tagged database knows in advance which recipes are vegan. A library you built from videos, screenshots and cookbook photos has no tags at all.

So the check has to happen at the ingredient level, on text extracted from the source. That places a dependency worth understanding: dietary flagging is only as good as the import that produced the ingredient list. If a video's caption omitted the fish sauce and nobody read the audio, no amount of rule-checking will catch it, which is part of why what is a recipe importer treats extraction quality as the foundation.

It also means the check improves as your library gets tidier. Recipes with complete ingredient lists get flagged accurately. Recipes imported as a title and 3 lines of caption do not, and they are worth reimporting if a real restriction depends on them.

Where does automation stop?

It stops at the packet. Software reads a recipe and can tell you that a line is a problem, that a line might be a problem depending on brand, or that nothing looks wrong. Only the third of those is a conclusion, and even that assumes the import read the recipe correctly.

For preference-driven eating, that is comfortably enough. For a serious allergy it is a filter that narrows the work rather than a substitute for reading labels, and anyone cooking for a household with several requirements should read cook for multiple diets for the practical routine.

It is worth separating 3 levels of requirement, because they need different amounts of caution. A preference, such as eating less dairy, is well served by flagging alone. A religious or ethical rule, such as halal or vegan, needs flagging plus attention to hidden ingredients like gelatine and alcohol in extracts. A medical allergy needs both of those plus label reading on every product, every time, because reformulation happens without notice.

On apps, ReciBites vs Intent covers a diet-led approach and ReciBites vs Lifesum covers filtering a catalogue against rewriting your own recipes.

Filtering

Hiding recipes that break a dietary rule, so they never appear in search or suggestions.

Flagging

Marking the specific ingredient lines that break a rule, leaving the recipe visible and usually offering a replacement.

Hidden ingredient

An ingredient whose name does not reveal the problem, such as gelatine, whey, malt or fish sauce.

Functional swap

A replacement chosen for what the original ingredient was doing in the dish, rather than for belonging to the same category.