No. The system prompt holds a firm boundary against recommending supplements for sleep, anxiety, energy, immunity, or any specific health condition. The bot can describe what the manufacturer prints on the label, including general use language, but it never goes beyond that into therapeutic recommendations. Condition-specific questions get referred to pharmacists, naturopaths, registered dietitians, or doctors, depending on the question.

Only what the label says. The bot quotes manufacturer suggested-use language verbatim from your product data, including dose ranges (one to four capsules, take with meals, divided doses). It does not adjust dosages for body weight, age, or condition, and it does not suggest 'starting low and building up' or any other personalization. Dosage tuning is a pharmacist or doctor's job.

Yes, when you tag products with testing fields (NSF Sport, NSF Contents Certified, USP Verified, ConsumerLab Tested, GMP-only). A customer asking 'what third-party tested protein powders do you carry' gets a precise shortlist by seal. Customers who care about supplement-industry rigor are often shopping precisely on this axis, and the bot meets them with a real answer.

Yes. Form (capsule, tablet, powder, liquid, gummy), bioavailability claims (methylated B vitamins, picolinate vs gluconate zinc, glycinate vs oxide magnesium), and active dose all live as fields the bot reads. A customer comparing two B-complexes gets a side-by-side from your data, not a generic 'both are good' deflection that does not help the purchase decision.

Yes. Multibot plus display conditions lets a sports-nutrition bot live on /sports/ with focus on protein, creatine, and electrolytes. An herbs bot on /herbs/ can describe traditional uses from the manufacturer's labels (the herb's traditional name, plant part, extraction method) without making therapeutic claims. Each scope is set with URL and category conditions.

Yours. SleekAI is bring-your-own-key for OpenAI, Anthropic, Google, or OpenRouter. Vitamin shops often prefer a model with strong refusal behaviour for the no-advice line. The system prompt and the model behaviour together hold the boundary. Conversation logs let you audit interactions periodically to verify the bot stays in scope, which is the kind of compliance-friendly visibility supplement retailers need.

It refuses. Drug interactions are pharmacist territory. The bot describes the supplement and refers the customer to their pharmacist or doctor with their full medication list. This refusal is consistent and explicit. For a vitamin shop, deferring drug-interaction questions is not just good practice, it is the only legal stance, and the bot reinforces it rather than improvising plausible-sounding answers.

Yes. Manufacturer-stated expiry dates and storage conditions (refrigerate after opening, store away from light) live as product fields the bot reads. A customer asking 'how long is the fish oil good for after opening' gets a real answer from the label. Storage and expiry are practical questions that fall well inside the product-description boundary, and the bot handles them confidently.