There is an enormous amount of recipe content online and very little quality control. Some of it has been cooked repeatedly and tested by people who know what they are doing. Some of it has been generated, scraped, rewritten, or copied from a source that was itself unreliable, and it looks identical on the page.

You cannot tell the difference from the photograph or the layout. You can tell a surprising amount from the ingredient list and the method, which is what this article is about — and since we run a recipe site, it seems fair to apply the same test to ourselves at the end.

Read the ingredient list against the method

The fastest reliability check is whether every ingredient appears in the method and every method step refers to an ingredient in the list. Mismatches are the signature of a recipe that has been rewritten rather than cooked.

An ingredient listed but never used, or a step that suddenly mentions something absent from the list, means nobody followed the instructions from top to bottom. If nobody followed them, nobody verified them.

• Tick off each ingredient as it appears in the steps. Anything left over is a warning.
• Watch for quantities that make no sense at the stated serving size — a tablespoon of salt for two portions, or two cloves of garlic for a dish claiming to serve eight.
• Be suspicious of recipes with no quantities for key seasonings at all.

Look for the things only a cook would mention

Recipes written by someone who actually made the dish contain sensory checkpoints: what it should look like, smell like, or feel like at a given stage. "Cook until the onions are soft and translucent, about eight minutes" is written by someone who watched it happen. "Cook the onions" is not.

The same applies to warnings. Notes about what goes wrong, why a step matters, or what to do if the sauce splits are the strongest positive signal available, because they can only come from experience of it going wrong.

• Good signs: visual cues, time ranges rather than exact times, notes on substitutions, explanations of why a step exists, pan sizes specified.
• Bad signs: uniformly round timings, no cues at all, an implausibly short total time, a method that never mentions heat levels.
• Neutral: a long personal introduction. Irritating, perhaps, but unrelated to reliability.

Check the timings for physical plausibility

Certain steps have floors that no technique can go below, and recipes that violate them have not been cooked. Onions do not caramelise in five minutes; they need thirty to forty-five. Dried beans do not soften in twenty. A large joint of meat does not go from fridge to cooked in half an hour. Bread dough does not rise properly in ten minutes.

Total times are the most frequently inflated number in online recipes, usually because they exclude preparation and preheating. That alone is not dishonest. Times that are physically impossible for the steps described are a different matter.

Treat nutrition blocks with particular care

The nutrition panel is the least reliable part of most recipe pages. It is almost always calculated by software from the ingredient list, not measured, and it inherits every ambiguity in that list: what size a medium onion is, how much of the frying oil ends up in the food, what "one serving" means.

That does not make it worthless for rough comparison. It does mean the precision implied by a figure like 412 calories is fictional, and that it should never be the basis of a medical or clinical decision.

Dietary and allergen labels are the highest-risk field

Tags like vegetarian, vegan, or gluten-free are frequently applied automatically from a category or a keyword rather than verified ingredient by ingredient. Automated tagging cannot see fish sauce in a curry, chicken stock in a risotto, or wheat flour in a stock cube.

For a preference, a wrong tag is an annoyance. For coeliac disease or a food allergy it is a safety issue. The only reliable procedure is to read the full ingredient list yourself and check the packaging of anything processed, on any site, including this one.

Where our own recipes come from

Recipe Finder does not write recipes. It searches TheMealDB, a public, community-contributed recipe database, and presents the results in a consistent format. That has clear implications for how much confidence the data deserves, and it seems better to state them plainly than to let the tidy layout imply more rigour than exists.

• Ingredients, quantities, and method text come from the source database and are not tested by us.
• Cooking time is derived from the structure of the method, not measured. Treat it as a rough band.
• Serving counts are derived from the number of ingredients, not portioned by a cook.
• Nutrition figures are estimates calculated from the ingredient list, because the source provides no nutrition data at all.
• Dietary tags are inferred from the source's broad category, so they are useful for filtering and unsuitable for allergen decisions.

How to use a site like this well

None of the above makes a community recipe database a bad tool. It makes it a discovery tool rather than an authority. It is very good at answering "what could I make with what I have" and at showing you dishes and cuisines you would not have searched for by name.

Use it for that, then apply your own judgement to the specifics. Read the ingredient list and the method before you shop. Sanity-check the quantities against the serving count. Trust your own tasting over any published time or temperature. If a step looks wrong, it may well be wrong.