If you run a WordPress multisite network, you already know that most third-party chatbot widgets ship as if every install were a single site. They store config in a single options row, ignore the blog ID, and happily serve answers about subsite A while the user is reading subsite B. Network admins end up either disabling the widget per subsite or rebuilding the integration from scratch with custom JavaScript.

SleekAI is built on top of switch_to_blog() and the wp_BLOGID_ table prefix. Each subsite gets its own chatbot configuration, its own display conditions, and its own model key if you want, while the plugin code lives at the network level. The bot on the marketing subsite reads wp_1_posts, the bot on the docs subsite reads wp_2_posts, and the conversations log into per-blog tables so nothing leaks across subsites.

That subsite isolation matters for regulated networks like university departments, agency white-label setups, and franchise WordPress installs where each location runs as its own subsite. Generic bots cannot enforce that boundary, which is why most network admins block them entirely. SleekAI keeps the boundary intact while letting the network team monitor token spend in one place.

WordPress multisite powers a long tail of high-value setups: universities with one subsite per department, franchise chains with one subsite per location, agencies hosting dozens of small business clients on a single install. In all of these the boundary between subsites is the whole point. Mixing data across subsites is not a feature, it is a leak.

Most off-the-shelf chatbot widgets are written as if WordPress were single-site only. They store config in a single options row, ignore the blog ID in cache keys, and run queries against wp_posts without ever calling switch_to_blog. On a network of 50 subsites that means either disabling the widget or accepting that the marketing subsite bot might answer with docs subsite data, depending on which page was rendered first.

SleekAI was built with multisite as a first class deployment target. The plugin is network-activated, every chatbot record lives in blog-prefixed tables, the conversation log is per subsite, and the network admin sees aggregate token usage in one place. That combination is what lets agencies sell a chatbot per client, franchises run per-location promos, and universities keep departmental data inside the right walls.

It also matters for cost control. With per-subsite token tracking the network owner can spot a runaway integration in hours instead of weeks, set soft caps, and bill clients accurately if the bot is part of a hosted service. Multisite without this kind of awareness is one bad widget away from a support fire.