Recipe sites are often huge, well-organised, and almost unusable for the most common question their visitors actually have: what should I cook tonight with these things in my fridge. A search bar wants exact ingredient names. A category page expects the user to already know they want pasta. Neither matches the casual, half-decided way people pick dinner.

SleekAI reads your recipes custom post type, the ingredients taxonomy, the prep_time and total_time postmeta, the serves field, and any diet flags like vegetarian, vegan, gluten-free, dairy-free. The user types 'I have chicken thighs, half a lemon, and rice, vegetarian husband is out tonight' and the bot proposes two or three recipes that genuinely match, with a sentence about each.

For sites that monetize via affiliate grocery links, the bot can also suggest a single missing ingredient that would unlock a fourth recipe and link to the store. Display conditions scope the widget across the recipes archive, individual recipe pages, and the home page. Logs capture every query (the ingredients listed, the constraints, the recipe chosen) which is a goldmine for content planning.

Recipe sites have a long-tail content problem. The best ones have hundreds or thousands of recipes built up over a decade, and most of those recipes never see a second visit because nobody can find them. Search bars are useless when the user does not know the exact ingredient name.

Category pages assume the user has already decided they want pasta or curry. Pinterest and Instagram do better at the discovery moment than the site itself does. A chatbot changes the dynamic because it can ask the only question that matters: what do you have, what do you feel like, how long until dinner.

The bot translates a vague intent into a concrete recommendation, with a link to a real recipe page on the site. The reader spends time on a page they would never have found otherwise, and the site finally gets value out of its archive. The second benefit is editorial.

Conversation logs reveal which ingredients combinations come up most often and which queries return nothing. That data is more honest than analytics because it captures intent rather than just clicks. A pattern of 'fast vegan dinners with one pot' is a brief, not a search query, and over a few months the editorial calendar starts writing itself.

The chatbot is not a gimmick. It is the recommendation layer the recipe site was always missing.