The short version
When a new review comes in, Ominvo sends it to Claude (Anthropic's AI) along with context about your business — name, industry, your tone preferences, and any custom instructions you've set. Claude returns a draft reply. You review it, edit if needed, then post. One-click posting to Google is pending API approval (see when is Google integration going live); for now you copy and paste.
This is available on Chad and GigaChad. Noob accounts can see reviews but don't get AI drafts.
What the model sees
Every draft is generated with:
- The review text — full content, rating, and reviewer name
- Your business name — used to personalize the reply ("Thanks for visiting [Business]!")
- Industry — helps Claude pick vocabulary and context that fits (a dental practice reply sounds different from a gym reply)
- Reply tone — you set this in account settings: Professional, Friendly, or Casual
- Custom instructions — a free-text field where you can tell Claude things like "always mention our loyalty program" or "don't use exclamation marks" or "we're family-owned, lean into that"
- Review metadata — star rating influences the reply's tone and content automatically; a 5-star gets a warmer, shorter reply than a 3-star
The model does not have access to your customer's purchase history, prior interactions, or any data outside what's listed above.
What makes a reply "good"
Claude is specifically prompted to produce replies that:
- Acknowledge the specific thing the reviewer mentioned (not a generic "thanks for your feedback")
- Use the reviewer's first name when it's visible
- Match the star rating in tone — celebratory for 5-stars, measured and solution-focused for 1-2 stars
- Stay within Google's response guidelines (no promotions, no arguing, no disclosing personal information about the reviewer)
- Stay concise — most replies are 3-6 sentences. Google buries long replies behind a "read more" link and reviewers rarely read them
One reply per review, on demand
Ominvo generates one draft per review when the review first appears in your dashboard. It doesn't auto-generate multiple options or keep re-drafting. If you don't like the draft, you can regenerate it (button in the reply editor) or edit it directly.
Drafts aren't sent automatically. Nothing posts to Google without your explicit action — either clicking post yourself or using one-click posting once the API integration is live.
Tone settings
Three options under Account Settings → Reply tone:
- Professional — formal language, business-appropriate, no contractions. Best for medical, legal, or high-end services.
- Friendly — warm and conversational, uses contractions, moderately upbeat. Works for most businesses.
- Casual — relaxed, informal tone. Works well for gyms, coffee shops, bars, salons with a younger clientele.
You can override tone per-reply in the editor if a specific review warrants a different approach.
Custom instructions
The custom instructions field is where most of the personalization lives. Claude treats it as standing context for every reply. Examples that work well:
- "We're a family-owned business run by two brothers since 1987. Mention this when it's natural."
- "Do not use the phrase 'we strive.' I hate it."
- "Always end by inviting them to come back and ask for [specific staff member] if they want consistency."
- "If someone mentions parking, acknowledge that it's a known issue and mention the lot on Oak Street."
There's a 500-character limit on custom instructions. If you need more nuance than that, the per-reply editor is the right tool.
Ready to post your first reply? See editing and approving replies for the full flow.