Building AI That Cooks: How Chefadora Created a Context-Aware Recipe Assistant

You're halfway through a recipe when you realize you're out of soy sauce. Panic sets in. You frantically Google "soy sauce substitute" and get bombarded with generic listicles that don't account for what you're actually making. Should you use Worcestershire sauce? Tamari? Will it ruin the dish?

This is the moment where most cooking apps fail you. They give you the recipe, but when things go sideways—and they always do—you're on your own.

Chefadora decided to fix this by building what they call "the world's first AI cooking assistant that is built right into the recipes." On the Levels Podcast, co-founder and CEO Sanjam Kohli explained how they created an AI that doesn't just answer cooking questions—it understands exactly where you are in the recipe and gives you answers specific to that moment.

The Context Problem

The difference between a helpful AI assistant and an annoying one comes down to context. Generic AI can tell you that soy sauce substitutes exist. Context-aware AI knows you're making fried rice, you're at step three, and based on what other ingredients you've already added, it can tell you exactly what to use and how much.

"With Chefadora AI, it knows what recipe you're following, what step you're on, and it will give you very recipe specific instant answers."

This seems obvious in hindsight, but it's surprisingly rare. Most recipe sites treat AI as a separate tool—a chatbot you can access if you navigate away from the recipe. Chefadora built the AI directly into the cooking experience, accessible from what they call "cooking mode" where users follow recipes step by step.

The AI doesn't just have access to general cooking knowledge. It understands the specific recipe in front of you, the techniques being used, the cuisine style, and can adapt its recommendations accordingly.

Beyond Text: The Future of Kitchen AI

The current version of the AI is text-based, but Chefadora has ambitious plans for where this technology goes next. Within the next few months, they're rolling out features that feel straight out of science fiction.

Image recognition is coming first. Soon, you'll be able to send the AI a photo and ask questions like "Are my onions brown enough?" or "Is the sauce thick enough?" The AI will analyze the image in the context of your specific recipe and tell you whether to keep cooking or move to the next step.

"Soon you'll also be able to talk to it because, you know, nobody likes to touch their phone if their hands are messy."

Voice interaction solves one of cooking's most annoying friction points. When your hands are covered in flour or raw chicken, fumbling with your phone to type a question is the last thing you want to do. Voice makes the AI genuinely hands-free, turning it from a tool you use when desperate into a natural cooking companion.

The multilingual capabilities extend to voice as well. You won't need to speak English to use the AI—you can talk to it in any of the 80 languages Chefadora plans to support.

"You don't need to know English to use the AI capabilities. You can just talk to it in any language."

Voice Cloning: Cooking With Your Favorite Creator

Perhaps the most innovative feature on Chefadora's roadmap is creator voice cloning. Imagine following a recipe from your favorite food blogger, but instead of just reading their instructions, the AI assistant speaks to you in their actual voice, adapted to answer your specific questions.

"They will be able to clone their voice. And when someone is following their recipe and asking the question, it would feel like, the creator themselves are talking to the person. So it enhances the experience."

This isn't just a gimmick. It deepens the connection between creators and their audience in a way that static content never could. A creator might post a recipe for Thai green curry, but when you're making it and ask "Can I use chicken thighs instead of breast?" the AI—speaking in the creator's voice—can answer based on their cooking philosophy and preferences.

As one of the podcast hosts noted, it's like "literally having your mum in the kitchen with you." Which, depending on your relationship with your mother, might be exactly what you want when attempting a complicated dish.

The feature is still in what Sanjam describes as a "pre-pilot stage" with lots of nuances to work out, but creator reaction has been positive. For creators building an audience, it's a powerful way to scale their presence beyond videos and written recipes.

Learning From Every Question

One of the AI's most powerful features is its ability to learn not just from the recipe itself, but from how other users have cooked it.

"It learns on its own as well. Like you said, it sees what other people would have if they have asked the same question in the same recipe."

This means the AI gets smarter over time. If dozens of people making the same chocolate chip cookie recipe ask about whether they can substitute brown sugar for white sugar, the AI starts to understand this is a common question for that specific recipe. It can then provide more nuanced answers that account for what actually works in practice, not just what works in theory.

The system combines recipe-specific context with crowdsourced cooking experience and general culinary knowledge to give answers that are simultaneously specific and reliable.

The Free-to-Use Model (With Limits)

Despite AI being expensive to run, Chefadora has made the assistant free to use with a daily message limit of 20 questions. For most home cooks making one or two dishes, that's more than enough. Power users who hit the limit can sign up for future premium features.

The AI is supported by ads shown in cooking mode, with revenue shared with creators. This aligns incentives nicely—creators benefit when people use the AI while following their recipes, so they're motivated to promote both the recipes and the AI features.

Future premium plans will expand beyond just more AI messages to include personalized meal planning, recipe recommendations based on your preferences and pantry, and even weather-based notifications suggesting recipes.

"The idea for the subscription not only includes the recipe AI help, but we plan to expand into personalized recommendations, personalized meal prep, like I mentioned, notifications based on weather or what's in your pantry and stuff like that."

The Challenge of Building AI That Actually Helps

Building useful AI in the kitchen is harder than it looks. The stakes are real—bad advice ruins dinner and wastes food and money. The AI needs to understand not just cooking techniques, but also food safety, flavor compatibility, dietary restrictions, and regional ingredient availability.

Chefadora spent six to seven months developing the AI before launching it on the web, and it's still being refined for their mobile app. They're taking a measured approach, rolling out features as they become reliable rather than rushing to market with half-baked capabilities.

This patience mirrors their broader philosophy seen in how they built relationships with creators—don't force growth or features before they're ready to deliver real value.

The cooking mode where the AI lives also includes step-by-step recipe guidance and integrated videos from creators, making it a comprehensive cooking experience rather than just a chatbot bolted onto a recipe site. This integration is key to making the AI feel natural rather than gimmicky.

Where Cooking AI Goes Next

Chefadora's vision extends beyond answering questions about substitutions. They see the AI becoming a true cooking companion that understands your skill level, dietary preferences, available ingredients, and even the weather to suggest what you should make.

The technology is still early, but the direction is clear: AI in the kitchen should understand context, speak your language (literally), and feel like cooking with a knowledgeable friend rather than consulting a database.

For product teams thinking about where to add AI features, Chefadora's approach offers lessons. They didn't just add AI because it was trendy—they identified a specific pain point (needing help mid-recipe) and built AI that genuinely solves it. The context-awareness isn't a feature; it's the entire point.

As more platforms experiment with AI, the ones that win will be those that make the AI feel like a natural extension of the experience rather than a separate tool you have to remember to use. Chefadora is betting that when you're standing in your kitchen with messy hands wondering if your sauce is thick enough, the AI that can actually see your sauce and give you a straight answer is the one you'll keep coming back to.

Key Points

• Chefadora's AI assistant is context-aware, knowing which recipe you're following and what step you're on to give specific answers
• Image recognition coming soon will let users ask questions like "Are my onions brown enough?" by sending photos
• Voice interaction will make the AI hands-free, solving the problem of touching your phone with messy hands while cooking
• Creator voice cloning will let the AI speak in a creator's actual voice, deepening the connection with their audience
• The AI learns from questions other users have asked about the same recipe, getting smarter over time
• The assistant is free with a 20-message daily limit, supported by ads in cooking mode with revenue shared with creators

Listen to the full conversation with Sanjam Kohli on the Levels Podcast.