Reply soon and keep your tone calm and solution-focused — never defensive. Thank the customer for taking the time, acknowledge and empathise with what went wrong, and apologise for the experience itself rather than admitting legal fault. The single most important move is to take the conversation offline: give a direct email, phone number or “ask for the manager” so the details are sorted privately instead of in a public back-and-forth. Briefly say what you’ll do to put it right, invite them back, and never argue or reveal private customer information. The generator’s negative band follows exactly this shape and drops in your contact for the offline step; you then edit and personalise it before posting.

Positive replies are easy to get right and easy to get wrong. Thank the reviewer by name, reference at least one specific detail they mentioned (a dish, a staff member, the speed, the atmosphere) so it doesn’t read like a template, show genuine appreciation, and warmly invite them back. Replying quickly — ideally within 24 hours — and keeping each reply individual is what builds trust and encourages more reviews. The generator’s positive band thanks by name, reinforces the specific point you enter, and invites them back, and gives you a few varied drafts so they don’t all sound identical.

The tool is 100% client-side. The review text and the details you type are turned into reply drafts locally in the page; nothing is sent to a server, logged or stored, and it works offline once cached. It is not an AI or LLM: each draft is assembled from hand-written, best-practice parts (greeting, thanks, an acknowledgement, an action/offline step for negatives, a sign-off) with your details slotted in, and the phrasing is varied with the browser’s crypto-secure randomness so you get genuinely different drafts. Because it is a template assistant and not legal or PR advice, you should always read and personalise the draft, reference real specifics, avoid admitting legal fault or sharing private details, and take serious complaints offline.

The star rating is the authoritative signal: 4–5 stars is treated as a happy customer (positive), 1–2 stars as an unhappy one (negative), and 3 stars as a mixed review (neutral). The tool also scans the pasted review for a small set of clearly positive and negative words to refine the on-screen explanation and to nudge an ambiguous 3-star review toward the side its wording leans to. It then shows you the detected band, a short reason and any words it noticed, so you can sanity-check it. This is a deliberate, transparent heuristic — not sentiment AI — so you stay in control and can pick a different tone or edit the draft if you disagree.