No, but more structure means cleaner answers. A WordPress custom post type with a version slug, a release-date field, and a category taxonomy (added, fixed, security, deprecated) gives the bot enough metadata to answer scoped questions like 'security fixes this quarter'. For sites running entries as regular posts under a 'changelog' category, the data-source wizard maps whatever fields exist.

Yes. If your entries are tagged 'added', 'fixed', 'security', and so on, the bot can answer category-scoped questions without confusion. 'What security fixes shipped in v4' returns only entries with the security tag, not the full release. For teams that publish a separate security advisories feed, scope a dedicated bot to that feed via multibot.

Display conditions and post status control this. Draft and private posts are invisible to a public chatbot by default. For internal usage, scope a separate bot to logged-in admins or a specific role, with its own system prompt that allows roadmap and unreleased notes. The two bots can run on the same site without leaking context between them.

Yes. The data-source wizard exposes the publish date for each entry, so the bot can reason about 'last 30 days' or 'since March 2026'. For very long histories, pair with OpenAI Files as a vector store, so retrieval pulls only the entries in the asked window into context, instead of stuffing the full changelog into the prompt.

Yes. SleekAI is a WordPress plugin, so it lives where the changelog already lives. Use display conditions to attach the bot to /changelog/ URLs, and the guideline filter to keep off-topic chat in check. For a public-facing page, the system prompt should defer politely on unannounced features rather than speculate, the bot is for what already shipped, not what might.

SleekAI is a one-time WordPress plugin license. You bring your own OpenAI, Anthropic, Google, or OpenRouter API key and pay the provider directly at standard API rates, with no token markup or resolved-conversation fee. For a low-traffic changelog page, monthly model spend is usually a few dollars even with a busy week.

Yes. The system prompt instructs the model to end every reply with a link to the changelog entry that informed it. For a public page that doubles as a sales artefact, the citation is what makes the answer credible, prospects need to verify the claim, and CSMs want the exact entry to forward in follow-up emails.

Two levers. First, the system prompt explicitly forbids invented entries and tells the model to defer when context is missing. Second, retrieval grounds answers in real posts, hallucination drops sharply when the right entry is in context. Conversation logs let you spot any drift, the rare invented feature is usually a missing changelog post, write it and the next answer is correct.