Site Reviews stores each review as a site-review custom post type with rating, author, title, and content fields plus the assigned post ID as postmeta. Categories and types map to taxonomy terms. That structure makes it straightforward for SleekAI's data-source wizard to expose the data: aggregate scores per product, the latest reviews on the current page, and category-level averages can all become named variables in the chatbot's system message.

The bot can use that to answer real visitor questions. Instead of saying the product is well-reviewed, it can quote the current average across a real review count, mention the most common themes from recent feedback, and confirm that a specific concern (size, durability, shipping speed) has been raised before. On a single product page, the bot reads the reviews assigned to that post; on a category archive it can summarize across all products in the term.

Display conditions keep the reviews-aware bot tied to the post types where reviews matter, multibot lets a public review-aware bot run alongside an admin moderation bot, and every conversation is logged inside WordPress with the page URL, model, and token usage attached.

Reviews are a deciding factor for many purchases, but a star rating and a list of comments only go so far. Most visitors scan a handful of reviews, miss the long tail of detail, and form an impression based on whatever happened to be on screen. They then ask the same questions over and over in chat, by email, or in the product Q&A widget: does anyone mention the fit, has anyone had a bad shipping experience, what is the most common complaint.

A generic chatbot trained on the open web cannot answer those questions, and a vague reassurance is worse than no answer because it teaches visitors that the bot does not actually know the product. Site Reviews already stores everything the assistant needs: each review as a structured post with a rating and a relationship to the reviewed item. SleekAI's data-source wizard maps that into the chatbot's system message at request time, so the bot can quote real averages, summarize themes from recent feedback, and acknowledge legitimate concerns rather than dodge them.

Pair that with display conditions and the bot only loads where reviews are relevant, multibot lets a separate moderation bot run for admins, and conversation logs feed back into both product copy and the system prompt over time. The result is a chat layer that turns existing review data into a useful pre-sales tool, instead of forcing visitors to read every single review before they decide to buy.