Recipes on Instagram: where to find them, and how to actually cook one

Rose · 16 September 2026

Instagram is full of recipes, but a saved Reel is not a recipe until someone writes down the amounts. Many food Reels keep the quantities in the voiceover or on-screen text, not the caption, so saving does nothing. To cook from Instagram, the ingredients have to come out of the video, then become a list you can shop from.

Why do I save recipes on Instagram and never cook them?

You have four hundred saved posts in a folder called Food, and last Tuesday you cooked pasta from memory. Yumzi exists because of that gap: saving is a zero-effort action, and it produces a zero-effort result. The bookmark tap costs you nothing and it commits you to nothing, so the archive grows and the Tuesday menu does not change.

Almost every page that ranks for “recipes on Instagram” answers a different question: which accounts should you follow. That question is solved. You already follow enough accounts. If you save five or six food Reels a week on the tram home, your supply problem has been over-served for about two years.

The part nobody covers is what happens between the save and the stove. To cook a dish you need three things written down: what to buy, how much of it, and in what order to use it. A Reel reliably gives you the third one, sometimes the first, and often not the second. So at seven in the evening you are standing in the kitchen holding a phone, scrubbing back four times to catch whether she said one tablespoon of gochujang or two, and by the time you have it written on the back of an envelope the appetite for cooking it has gone.

That is not a discipline problem. It is a format problem, and it has a specific shape.

Is the recipe actually in the caption?

Usually not in full, and this is the single most useful thing to understand about recipes on Instagram. Yumzi was built around it. A Reel hides its recipe in four different places, and only one of them is text you can copy.

1. The caption. Sometimes the whole thing is there, ingredients with grams, method in steps. Sometimes it is a mood, three emoji and “recipe below”, where below means a comment that got buried.
2. The voiceover. The creator says the amounts out loud while her hands move. “Two chicken thighs, big spoon of gochujang, and honestly just a splash of the vinegar.” Nothing about that exists as text anywhere.
3. On-screen text. A card of ingredients that appears for roughly a second and a half, often under the caption bar, often mid-transition.
4. Link in bio. The full recipe does exist, on a blog, behind a homepage or a link aggregator where you then hunt for the post you just watched.

The format pushes towards the second and third. A benchmark across six million Reels found the 45 to 60 second band performs strongest for engagement and views, across all Reels rather than food specifically, and creator guidance for this genre recommends 20 to 60 seconds for food and recipes. In forty seconds you can show four steps beautifully. You cannot dictate nine ingredients with quantities at a speed a person can write down. So the amounts get compressed into one spoken line or one text overlay, and the caption carries the hook instead.

Here is our limit, stated plainly: nobody has published a count of how often Instagram food captions contain complete quantities. We have not coded a sample of fifty Reels either, so we are not going to put a percentage on it. What we can say is that the caption alone is often not enough, which is why an app that only reads captions will hand you back half a recipe and call it a success.

Where do saved Instagram recipes actually go?

They go into Saved, optionally sorted into Collections, and once they are there you can only scroll them. Yumzi treats that as the second half of the problem, because a recipe you cannot find again is the same as a recipe you never saved.

The bookmark puts a post into one flat, reverse-chronological Saved area. Collections let you file it into a named folder. You can also send a post to yourself in DMs, which many people use as a to-cook queue, and you can screenshot the ingredient card into Photos, which at least gets it into an image library.

The catch is retrieval. Third-party guides on this are consistent: there is no native search inside Instagram’s Saved area, including no search by caption text. Instagram’s own Help Centre pages on saving posts and managing collections do not state any maximum number of saved posts or collections, so nothing slows the pile down. An archive with no ceiling and no search is a pile, not a library. You cannot ask it “what did I save with aubergine”, you can only scroll until you recognise a thumbnail.

Screenshots are worse in a specific way: the ingredient card is now an image in Photos, mixed in with parking signs and dog photos, separated from the video that explained the method.

How do you save recipes on Instagram so you can find them again?

Save into named Collections rather than the flat Saved feed, and name them after decisions instead of cuisines. Yumzi still reads the video for the amounts, but this habit makes the shortlist stage faster, and it takes about four seconds per post.

The steps, precisely:

1. Tap and hold the bookmark icon under the Reel instead of tapping it once. Holding brings up your Collections.
2. Pick a collection, or tap the plus to create one.
3. To move something later, open Saved, open the post, tap and hold the bookmark again and choose a different collection.

Naming is where most people lose it. “Food” and “Recipes” and “Yum” are not decisions. These are: 20 minutes after work, feeds four, have the ingredients already, weekend, longer, chicken thighs. When you are deciding what to cook tonight, the constraint you are working under is time and who is eating, not whether the dish is Korean or Italian.

Be honest about what this fixes, though. Better folders solve the finding problem and do nothing at all for the cooking problem. Opening “20 minutes after work” still leaves you with eleven videos, none of which has a shoppable ingredient list attached, and you still do not know whether the gochujang jar in the fridge door is enough for any of them.

What kinds of Instagram food accounts are worth following?

Sort them by how recipe-complete they are, not by how good the food looks. Yumzi works with all three types below, but it is worth knowing which kind you are collecting, because it decides how much work is left after the save.

Full-caption cooks. The whole recipe is typed into the caption, ingredients with amounts, method in numbered steps. These are the accounts food editors quietly rely on. Follow more of them than you think you need. If you copy a caption into a note, you have a real recipe.

Voiceover cooks. Often the best cooking on the platform, and the worst documentation. She tells you the amounts as she goes, the caption says “so good, recipe in the video”. The dish is real, the recipe is real, and it exists only as sound.

Link-in-bio bloggers. A complete, tested, written recipe exists on their site. Between you and it sit a bio, a link page and a search box. Fine on a Sunday. Not fine at seven on a Tuesday.

Vibes accounts. Beautifully shot, no quantities, no intention of giving you any. Nothing wrong with them, they are television. Just do not file them next to dinner.

This is the one thing the roundup articles do well, so take their lists. Just notice that a list of accounts hands you more supply, and supply was never your bottleneck. The ranking pages stop exactly where your problem starts.

Are healthy recipes on Instagram actually balanced?

A single Reel is one dish, and one dish cannot be balanced or unbalanced in any useful sense. Yumzi looks at the week instead, because that is the unit Dutch dietary guidance uses and it is also the unit you actually eat in.

A saved folder skews towards whatever performs on video, which is melted cheese, crispy chilli oil and anything involving a dramatic pull. That is not a moral failing of the algorithm, it is what looks good in forty seconds. Seven dinners chosen purely from that folder tend to land in the same place: heavy, similar, low on vegetables.

The Schijf van Vijf gives a reference point that is easy to check against a menu. For adults it sets 250 grams of vegetables and 200 grams of fruit per day, and across a week it recommends fish once a week and legumes at least once a week.pdf). Those are weekly and daily targets, which means no individual Reel can meet or fail them. A weekly dinner plan can.

What we do with that: when Yumzi builds a week out of your saved videos, it checks the shape of the week against those reference points, so a fish dish and a legume dish get into the plan and the vegetable weight is visible rather than assumed. What we do not do: count calories, score you, or claim a health outcome. It is a balance check on a menu, nothing more.

Can an app read the recipe out of an Instagram video?

Yes, and the distinction that matters is between scraping the caption and reading the video itself. Yumzi reads the video: the spoken audio and the text that appears on screen, not just the caption text under it.

Yumzi is a Dutch meal-planning app that reads the recipe out of TikTok and Instagram videos you saved, plans a balanced week of dinners, and builds one aisle-sorted grocery list you confirm at your own supermarket.

Caption scraping is the common approach because it is the easy one. Text is already text. It works perfectly on full-caption cooks and returns almost nothing on the voiceover cooks, which is a large part of what you have saved. Reading the video means transcribing what the creator says, reading the overlay cards, and then doing the harder part: deciding that “big spoon of gochujang” is about one tablespoon, that “a couple of thighs” is two, and that the “splash” of vinegar belongs in the sauce and not the marinade.

We are not the only people who noticed the audio problem. At least one other recipe app’s own help documentation states that when an Instagram post has no written recipe, it detects the recipe from the video’s audio. We are pointing at that on purpose: it is the only published acknowledgement in this category that the amounts are frequently spoken, and it tells you the problem is real rather than something we invented to sell against.

Where we differ is what happens next. Reading is step one. A read recipe that sits in a library is still not dinner.

What is the difference between an app that saves recipes and one that plans dinners?

An organizer stores what you give it. A meal planner decides what you eat on which night and what you have to buy. Yumzi is the second kind, and this is the difference that decides whether Wednesday is different from last Wednesday.

An organizer’s job is finished when your recipe is filed, tagged and searchable. That is a genuine improvement on a folder you can only scroll. But it hands the hard decision straight back to you: on Wednesday at seven you still have to pick something, check what is in the fridge, and work out what to buy for it.

A planner’s job starts there. It picks which of your saved dishes go on which evenings, looks at the week as a whole rather than the dish, merges the shopping across all of them, and gives you one list instead of seven recipes. Three of your saved Reels want spring onion. A planner buys spring onion once.

This matters more in the Netherlands than in the messaging of most of these apps, because the decision is genuinely last-minute here. A Dutch survey write-up reports that 41% of respondents decide what to eat on the day itself, with 15% deciding and buying the meal on the same day. With 8.92 million Instagram users in the Netherlands as of January 2025, a lot of those same-day decisions are being made by someone holding a phone full of food videos they cannot cook from.

How do you go from a saved Reel to a shopping list?

You pick the Reel, Yumzi reads it, and you get ingredients with amounts that merge into one aisle-sorted grocery list. Here is the whole path with a real example, so you can see exactly where the work goes.

Take a gochujang chicken Reel, 38 seconds, caption reads “the only thing I want to eat right now”. The creator says the amounts once, at speed, over a pan.

What comes back:

• 2 chicken thighs, boneless
• 1 tbsp gochujang
• 1 tbsp honey
• 2 tsp rice vinegar
• 2 cloves garlic
• 2 spring onions (AH Bosui, 1 bunch)

Six lines, each with a number. That is the difference between a video you liked and a dish you can buy for. Add two more saved Reels for the same week and the lists merge: one bunch of spring onion covers all three, the garlic consolidates, and the list reorders itself the way the shop is laid out so you walk it once, top to bottom, instead of going back for the vinegar.

In the Netherlands there is one more convenience at the end, once you have the list. One tap sends your shopping list to your Albert Heijn basket. You review, confirm and pay at AH yourself. Nothing is bought without you looking at it, and you can ignore that step entirely and shop from the list on your phone.

What if the creator never wrote anything down at all?

Then Yumzi reads the audio and the on-screen text, and when there is genuinely no recipe in the video, it tells you so instead of making one up. We never invent a recipe. That is the load-bearing part of this whole piece, so here is what it means in practice.

Cases we handle well: a caption-less Reel where she talks through the dish. An overlay ingredient card with no caption at all. A creator who says “a good glug” and “a knob of butter”, where we convert to a workable amount and show you that we did, so you can change it.

Cases we do not handle, stated plainly:

• Silent montages set to music, with no text. There is nothing to read. A fast cut of hands and a finished bowl is not a recipe, and no amount of processing makes it one.
• Compilation Reels. “Five lunches for the week” in thirty seconds gives fragments of five dishes and complete quantities for none.
• Recipes that exist only behind a link. If the video shows the dish and the caption sends you to a homepage, the recipe is on a website, not in the video.
• Posts that have been deleted or made private since you saved them.

When one of those comes up, you get a clear “we could not read a recipe from this one”, not a plausible-looking ingredient list we guessed at. A wrong list is worse than no list, because you only find out at the hob with the pan already hot. Saying no costs us a save. It is still the right answer.

What to do with the four hundred you already saved

You do not need to start collecting again. Yumzi is built to work backwards through the archive you already have, which is the only part of this problem where months of saving turns into an asset rather than a reproach.

Open the folder tonight. Do not scroll it all, that is how the last eight months went. Scroll until you find three dishes you want to eat this week, which takes about two minutes because you saved them for a reason and you will recognise them instantly.

Three Reels is one shopping trip, three dinners, and one list. That is the smallest version of this that changes anything, and it is a much better first move than reorganising four hundred posts into perfect collections you will also not cook from.

Open your saved folder, pick three, and let Yumzi read them into one list you can shop from tonight. #reeltodinner