Introduction
Responding to every Google review builds trust and improves rankings — but it takes time. AI-powered review replies can draft responses in seconds, which you review and post. This post shows you how to automate review replies safely, without sounding like a robot.
Why respond to reviews?
- Google favors active profiles — responding signals engagement
- 89% of consumers read business responses
- 45% of consumers are more likely to visit a business that responds to negative reviews
- Unanswered negative reviews look like you don’t care
The safe automation approach
Never fully automate review replies. AI should draft responses for you to review and approve. This ensures:
- Accuracy (AI can misunderstand context)
- Personal touch (AI can sound generic)
- Compliance (no policy violations)
- Quality control (catch errors before posting)
Step 1: Set up the AI reply generator
Using OpenAI API
import openai, os
from openai import OpenAI
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
def generate_reply(review_text, rating, business_name):
prompt = f"""You are a customer service representative for {business_name}.
Write a professional, empathetic reply to this Google review.
Review ({rating}/5 stars): "{review_text}"
Rules:
- Thank the reviewer by name if mentioned
- For positive reviews: express gratitude, mention specific details from the review
- For negative reviews: apologize, take responsibility, offer to resolve offline
- Keep it under 75 words
- Don't use generic phrases like "We value your feedback"
- Don't ask for anything (no "please visit again" or "tell your friends")
- Sound human, not corporate
- Sign off with the business name, not a person's name"""
response = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": prompt}],
temperature=0.7,
max_tokens=150,
)
return response.choices[0].message.content
Using local Ollama (free, private)
import ollama
def generate_reply(review_text, rating, business_name):
prompt = f"""Write a professional reply to this Google review for {business_name}.
Review ({rating}/5): "{review_text}"
Rules: Thank them, be specific, keep under 75 words, sound human."""
response = ollama.chat(model="llama3.2", messages=[{"role": "user", "content": prompt}])
return response["message"]["content"]
Step 2: Fetch reviews (Google Business Profile API)
from google.oauth2 import service_account
from googleapiclient.discovery import build
# Authenticate
creds = service_account.Credentials.from_service_account_file("service_account.json")
service = build("mybusiness", "v4", credentials=creds)
# Fetch reviews
account_id = "accounts/YOUR_ACCOUNT_ID"
location_id = "locations/YOUR_LOCATION_ID"
reviews = service.accounts().locations().reviews().list(
parent=f"{account_id}/{location_id}"
).execute()
for review in reviews.get("reviews", []):
print(f"Rating: {review['starRating']}")
print(f"Text: {review.get('comment', 'No comment')}")
print(f"Reply: {review.get('reviewReply', {}).get('comment', 'No reply')}")
print("---")
Step 3: Generate and review replies
import streamlit as st
st.title("🤖 AI Review Reply Generator")
# Fetch reviews
reviews = fetch_reviews()
for review in reviews:
with st.container(border=True):
st.write(f"**Rating:** {'⭐' * int(review['starRating'])}")
st.write(f"**Review:** {review.get('comment', 'No comment')}")
if st.button("Generate Reply", key=review['reviewId']):
reply = generate_reply(
review.get('comment', ''),
int(review['starRating']),
"Your Business Name"
)
st.session_state[f"reply_{review['reviewId']}"] = reply
if f"reply_{review['reviewId']}" in st.session_state:
reply = st.text_area("Edit reply", value=st.session_state[f"reply_{review['reviewId']}"], key=f"edit_{review['reviewId']}")
if st.button("Post Reply", key=f"post_{review['reviewId']}"):
post_reply(review['reviewId'], reply)
st.success("Reply posted!")
Step 4: Post the reply
def post_reply(review_id, reply_text):
"""Post a reply to a Google review."""
service.accounts().locations().reviews().updateReply(
name=f"{account_id}/{location_id}/reviews/{review_id}",
body={"comment": reply_text},
).execute()
Step 5: Automate the flow (with human review)
import schedule, time
def daily_review_check():
reviews = fetch_reviews()
unreplied = [r for r in reviews if not r.get("reviewReply")]
for review in unreplied:
reply = generate_reply(
review.get("comment", ""),
int(review["starRating"]),
"Your Business Name"
)
# Save to a queue for human review (don't auto-post)
save_to_review_queue(review, reply)
print(f"Drafted reply for {review['reviewer']['displayName']}")
print(f"Drafted {len(unreplied)} replies. Review and approve in the dashboard.")
# Run daily at 9 AM
schedule.every().day.at("09:00").do(daily_review_check)
while True:
schedule.run_pending()
time.sleep(60)
Key: The AI drafts replies; a human reviews and approves before posting. Never auto-post.
Template-based replies (no AI needed)
For simple cases, templates work fine:
Positive review templates
- “Thank you, [Name]! We’re thrilled you enjoyed [specific detail]. We look forward to serving you again!”
- “Thanks for the kind words, [Name]! [Specific detail they mentioned] is one of our favorites too. See you next time!”
- “We appreciate your feedback, [Name]! It’s customers like you that make our work rewarding.”
Negative review templates
- “Thank you for your feedback, [Name]. We’re sorry your experience didn’t meet expectations. We’d love to make it right — please contact us at [email/phone].”
- “Hi [Name], I’m sorry about [specific issue]. We’ve addressed this with our team. Please reach out so we can make it up to you.”
- “Thank you for bringing this to our attention, [Name]. We take feedback seriously and are working to improve. We’d appreciate the chance to make it right.”
Neutral review templates
- “Thank you for your feedback, [Name]. We appreciate you taking the time to share your experience. We’re always working to improve!”
- “Thanks, [Name]! We’re glad you visited and appreciate your honest feedback. We hope to see you again soon.”
AI vs human replies
| AI-drafted | Human-written | |
|---|---|---|
| Speed | Seconds | 2-5 minutes |
| Quality | Good (needs review) | Best |
| Personalization | Medium | High |
| Scalability | Unlimited | Limited by time |
| Cost | $0.001/reply (API) | Your time |
| Risk | Can hallucinate or sound generic | Low (you control it) |
Best approach: AI drafts, human reviews and approves. Best of both worlds.
Common mistakes
- Auto-posting without review — AI can misunderstand context or sound generic. Always review.
- Generic replies — “Thank you for your feedback” for every review looks automated and insincere.
- Arguing with negative reviews — never. Acknowledge, apologize, offer to resolve.
- Ignoring negative reviews — unanswered negative reviews look worse than the review itself.
- Not personalizing — mention specific details from the review to show you read it.
- Over-promising — “We’ll fix this immediately” when you can’t. Be realistic.
Key Concepts
- AI reply generation = using LLMs to draft review responses
- Human-in-the-loop = AI drafts, human reviews and approves
- Google Business Profile API = programmatic access to reviews
- Template-based replies = pre-written responses for common scenarios
- Review velocity = number of reviews per month
What to Try Next
- Automate Google review requests to get more reviews
- Reply to negative reviews with templates
- Get more Google reviews with proven strategies
- Build a review management dashboard in Streamlit
FAQ
Is it against Google’s policy to use AI for review replies?
No — Google doesn’t prohibit AI-assisted replies. However, the replies must be genuine and relevant. Auto-posting generic AI replies without review can lead to policy violations if they’re misleading or spammy.
How much does AI review reply automation cost?
Using GPT-4o-mini: ~$0.001 per reply (150 tokens). For 100 reviews/month, that’s $0.10/month. Using local Ollama: free. The cost is negligible.
Can AI handle all types of reviews?
AI handles positive and neutral reviews well. For negative reviews, AI can draft a response, but a human should always review and personalize it — especially for sensitive complaints or legal issues.
What if the AI generates an inappropriate reply?
That’s why human review is essential. Always review AI-drafted replies before posting. Set up a review queue (dashboard) where you approve/reject/edit drafts. Never auto-post. See Automate Google Reviews for more strategies.