CoBlocks adds Gutenberg blocks that small business, restaurant, and event sites use heavily: pricing tables, accordions, food and drinks menus, click-to-tweet, hero, and gallery layouts. The block markup ends up in wp_posts.post_content and resolves to real HTML when the page renders. SleekAI reads the rendered page, so a food menu becomes a list of dishes with prices, a pricing table becomes a list of plans, and an accordion becomes Q&A pairs, all of which land in the chatbot's context.

The data-source wizard pulls ACF fields, custom post meta, and taxonomies referenced from a block, so a restaurant CPT with allergens stored in ACF can surface those values when the food menu block is on the page. Display conditions scope different bots to different templates or post types, so a menu-page bot and a contact-page bot can run side-by-side under multibot without code.

Conversations are logged in WordPress with model name, token usage, and the page URL. The JS API and PHP embed open the bot from a custom button or a specific block click, with the current CoBlocks page already loaded as context.

CoBlocks is popular with restaurants, cafes, small agencies, and local services because its blocks cover the layouts those sites need most: food menus, pricing tables, hero sections, galleries, accordions. The downside of a generic chatbot on those sites is that the menu changes every few weeks. New seasonal dishes appear, prices nudge, allergens get re-evaluated, and the bot's training data has to keep pace.

In practice it doesn't, and within a month the bot is recommending a dish the kitchen stopped serving. Wrong answers on a restaurant page are worse than no answers, because they lead to a guest arriving expecting something that isn't on the menu. The structural fix is to ground the bot in the rendered page so the menu the marketing team published in Gutenberg is also the menu the bot sees.

With CoBlocks specifically, the food and drinks block renders to structured HTML with dish title, description, and price, which is exactly what the model needs to answer "do you have anything vegan" or "what's the priciest starter" correctly. Combined with display conditions that put a menu-aware bot on the menu template and a booking-focused bot on the reservations page, the chat layer matches the structure of the site itself.