AI keeps
the index clean.

When Weeknight Dinners crawls a URL, it needs to make a series of calls about it. Is this page actually a recipe, or is it a blog post, a restaurant review, a product listing? What kind of dish is it? What other recipes are similar enough that they belong grouped together? These judgments are what makes a search index worth using, and they’re hard to get right at scale with rules alone.

We use AI models to help make those calls — filtering out the non-recipes, identifying the dish or technique a page is about, and associating related content together so the index is navigable. The output is metadata that lets us index a page well; the recipe content itself stays where it was published.

For partner creators who have an agreement with us, we may also use AI to assist with extracting structured data from their pages — identifying ingredient lists or step boundaries in unstructured text. This is still identification and extraction, not generation. We expect the specifics of where AI helps with the index to evolve over time; what doesn’t evolve are the commitments below.

Weeknight Dinners will never use AI to generate content. That means no AI-written recipe suggestions, no generated meal plans, no synthesized reviews, no chatbot assistants, no “based on what you cooked” anything. If text appears in Weeknight Dinners, a human wrote it — either you, another user, or a content creator whose work you saved with their permission.

This isn’t a temporary policy we expect to revisit. The value Weeknight Dinners offers is in organization, scheduling, and the shared record of what your household actually cooks — not in having a model invent recipes for you. Adding generation would change what Weeknight Dinners fundamentally is, and we don’t want to build that product.

Weeknight Dinners has a social feed. People in your household and friend group post what they cooked, share recipes, and log how meals went. That feed is chronological — newest first, oldest last, no exceptions.

We do not rank posts by engagement. We do not surface content because it got more likes or comments. We do not learn from your behavior to decide what to show you next. There is no recommendation engine, no “you might also like,” no “trending in your area.” AI plays no role in the feed whatsoever.

This is a deliberate choice, not a technical limitation. The whole problem with algorithmically ranked feeds is that optimizing for engagement rewards content that provokes reaction rather than content that’s useful or true. A cooking app doesn’t need that dynamic. The people you follow cooked something — you should see it, in order, and decide for yourself whether it matters to you.

The legal framing that reflects this is Section 230 of the Communications Decency Act: we are a platform that hosts and displays user content, not a publisher that curates or editorializes it. Choosing a chronological feed is how we stay on the right side of that line in practice, not just in law.

Weeknight Dinners will never use your data to train AI models — ours or anyone else’s. Your saved recipes, your meal calendar, your posts, your household’s cooking history: none of it is fed into a training pipeline.

We will never share your personal data with third-party companies for any purpose other than what’s necessary for the fundamental operation of the service — AI-related or otherwise.

For the full picture of what data we collect, how it’s stored, and what rights you have over it, see our Data Policy.

Weeknight Dinners does not sell user data. Weeknight Dinners does not share user data with advertisers, data brokers, analytics companies, or AI vendors for purposes beyond operating the service. The AI classification we use is run against page content — not against user profiles — and we select vendors with strict data processing agreements that prohibit training on inputs.

When we use third-party services to operate Weeknight Dinners (hosting, authentication, payments), those relationships are governed by contracts that restrict how those services can use the data they handle on our behalf. See the Data Policy for the complete list.