Save a recipe from a blog without the ads

Paste the link, get the recipe. Cookfolio reads the page, keeps the title, ingredients, steps, times and photo, and leaves behind the childhood memories, the pop-ups and the 40 comments. Two seconds on a well-built site, a few more when the AI has to read the page itself. You check the result before anything is saved, and the link to the original stays on the card.

What lands on the card, and what gets left behind

The import pulls out the title, description, servings, prep time, cook time, and calories per serving when the page states them. Then the full ingredient list, with quantities and units separated rather than kept as one string, and the numbered steps, including sub-steps when the page splits the recipe into “for the dough” and “for the filling” sections. Plus the main photo.

What does not come across: the personal story, the ads, the navigation menus, the sidebars, the comments, the share buttons, the newsletter form. That is the point: you end up with a recipe card, not a copy of a web page.

Ingredient parsing is the part most people underestimate. A line like “2 tablespoons of olive oil” becomes a quantity of 2, a unit of tablespoon, and a name of olive oil. The unit table covers the six languages of the app, so “cuillère à soupe”, “cucchiaio”, “Esslöffel”, “EL”, “łyżka” and “cucharada” all resolve to that same unit. Fractions such as 1/2 and decimal commas such as 0,3 are understood. That normalisation is not cosmetic: it is what later lets you adjust the number of servings and have quantities recalculate, and lets duplicate ingredients merge in the shopping list.

The photo is downloaded and re-hosted on your own account, resized to 1200 px at most and re-encoded, so it does not break if the blog goes offline. You can crop or replace it. Source images heavier than 5 MB are skipped and the recipe imports without one, which is a size guard, not a failure.

Two ways to read a page: structured data first, AI second

The first path is structured data. Most serious recipe sites publish an invisible, machine readable copy of the recipe alongside the visible page, in schema.org Recipe format, because search engines need it. Cookfolio reads that block directly: no AI, close to instant, and the quantities come out exactly as the author typed them.

The second is the fallback, used when the page has no such block or when the block is incomplete. The page is cleaned up first: scripts, styles, navigation, header, footer, sidebars, iframes, forms, ad blocks, comments and share widgets are stripped out. It then zooms in on the portion that looks like a recipe, and only that cleaned text is sent to the AI, capped at around 15,000 characters. This matters for the exact complaint most people have: the AI never sees the ads or the comment thread, and the personal story is largely gone before the reading begins. The AI import in general works on the same principle.

The switch has a precise condition: the structured block is only trusted when it has at least a title, one ingredient and one step. A site declaring a Recipe block with an empty ingredient list falls through to the AI instead of handing you an empty card. The character cap has one consequence: on a very long page the tail can be cut off, another reason to glance at the final steps.

Nothing is saved until you check it

No import writes straight into your collection. After extraction, a review screen shows the whole card with every field editable: title, description, servings, times, each ingredient line, each step, the tags, the photo, and the source URL. Nothing is stored until you confirm.

Both paths fail in predictable places, so a checklist beats a general warning. Look for an ingredient listed twice because it appeared in two sections, an oven temperature read as a quantity, a servings count the page expressed as “4 to 6” (only the first number is kept), and a step that merged two instructions into one line. Read the last steps if the page was long. Thirty seconds of proofreading: the alternative is retyping the recipe by hand.

When it does not work, and what to do instead

Some pages cannot be read from the outside. Each case has a workaround.

If the site requires a login or a subscription, what gets fetched is the paywall, not the recipe: open it in your browser where you are logged in, select the text, and use the paste text import instead. A site that blocks automated reading takes the same workaround, or a screenshot imported as a photo. If the recipe exists only as an image or infographic, there is nothing textual to read, so use the photo import. If the link points at a home page, a category or a search result, copy the direct link to the recipe itself. Pages over the 2 MB cap are refused, and a page holding several recipes gives back one. Threads and X (Twitter) links are refused up front.

Two rules of thumb: if the page opens in a private browsing window without asking you to log in, the import will almost certainly work; if it does not, go straight to paste text. One more guardrail: fetching the page gives up after 10 seconds, so a slow site sometimes succeeds if you retry a minute later.

One clarification, because it causes confusion. Instagram, Facebook, TikTok, YouTube and Pinterest links do not belong in this list. They have their own import paths and work. A Pinterest pin, for instance, is handled by following the pin through to the original recipe.

What a web import costs

An import from a blog or website consumes one credit out of the 5 lifetime free imports, exactly like a photo, a PDF, a YouTube video or a Pinterest pin. There is no separate or cheaper allowance for web links, even though the structured data path uses no AI. The credit is only taken on success, so a failed import does not burn one.

A second free plan limit catches people out: the collection is also capped at 5 recipes, so a free account can hit the recipe ceiling before it ever reaches the import ceiling.

Two throttles apply to everyone, free or paid, as infrastructure protection rather than a commercial limit: five seconds between imports, and 100 imports per day.

Pro at 4.99 euros a month or 39.99 a year lifts both caps. Family at 6.99 a month or 59.99 a year covers up to 4 accounts sharing recipes, meal plan and shopping list in real time, each person keeping their own login. These are the web and Android prices; on iOS the App Store applies its own local pricing.

Archiving a recipe without erasing the person who wrote it

The address of the original page is prefilled on the review screen and saved with the recipe, then shown in a Source section on the recipe detail as a tappable link. The credit is not a promise, it is a working link on the card.

Three principles are enough. This is a personal archive: importing a recipe for your own kitchen is not the same thing as republishing it. When you share a recipe from Cookfolio you share a web link, and the source stays attached. And go back to the source when it matters: the author’s notes, the variants, and the comments where readers report what worked never make it into any structured extraction. If a blog is a regular source for your household, the most useful thing you can do is visit it, not only import from it.