
Reviews AI: Automating Requests and Replies Without Sounding Automated
Send the review request when the job is marked complete, not when an invoice happens to be paid. Use SMS first and email as a separate follow-up, configure replies by sentiment and source, and require human review for every 1-star response. The goal is not to make every review interaction automatic; it is to automate the predictable work while keeping complaints and substantive praise out of generic templates.
Start with job completion, not payment
A paid invoice is a billing event. It does not reliably tell you when the customer has enough experience to review the work. Payment can happen before service, days after service, or in installments; a completed-job status is the cleaner operational signal.
HighLevel supports a review-request workflow triggered by an appointment status change. Its local-service playbook uses an Appointment Completed status, waits 1–2 hours, and then sends personalized SMS and email requests. That gives the customer time to leave the property or return to the rest of their day without disconnecting the request from the completed service.
Build the workflow around the status your team already uses to mean the work is actually finished. The essential sequence is:
- Change the appointment or pipeline status to Completed.
- Wait 1 hour.
- Run the Send Review Request action.
- Let the SMS and email request settings control their respective delivery schedules.
Do not let technicians mark a job complete when they are merely en route, waiting for a part, or returning later. The workflow will do exactly what the status tells it to do.
Remove invoice-based duplicates
HighLevel’s QuickBooks integration has a separate automation that sends a request when a contact’s first invoice is paid in full and reaches a $0 balance. It is independent of the normal workflow action. If job completion is your trigger, go to Subaccount → Settings → Integrations → Quickbook Integration and deselect Send Review Requests for contacts.
This matters because running both systems can produce two requests tied to different events. HighLevel also notes that its Invoice Paid workflow trigger does not react to QuickBooks payments; it reacts only to invoices paid inside HighLevel.
Give SMS and email separate jobs
SMS and email are not one combined campaign in HighLevel. Configure SMS at Reputation → Settings → SMS Requests and email at Reputation → Settings → Email Requests. Each channel has its own enable switch, initial delay, repeat interval, retry limit, sender, live template, and retry templates.
Use SMS for the first short request. Set its initial delay to 1 hour after the Send Review Request action. Keep the message focused on the completed service and include {{reputation.review_link}}. Do not ask for a five-star rating, tell the customer what to write, or imply that only happy customers should use the link.
Use email as the second channel rather than sending it at the same instant. Set the email initial delay to 24 hours. The email can identify the completed service, explain that honest feedback is useful, and present the review link through HighLevel’s Review Link element or a CTA connected to that link.
HighLevel supports SMS timing choices such as immediate, 1 hour, 2 hours, 4 hours, or a custom delay. It also supports custom retry intervals and one through five SMS retries with different templates. The fact that five retries are available does not mean five reminders are appropriate.
Set a maximum of two SMS retries: the first after 3 days and the second after 1 week. Write each retry differently. The first can be a brief reminder; the second should clearly read as the last request. Stop the sequence when the review link is clicked, using the channel’s built-in repeat behavior.
| Channel | Timing | Purpose |
|---|---|---|
| SMS | 1 hour after the request action | Short first request tied closely to the completed job |
| 24 hours after the request action | More context and a visible review CTA | |
| SMS retry 1 | 3 days later | Brief reminder with different wording |
| SMS retry 2 | 1 week later | Final request, explicitly framed as the last reminder |
A business with 11 reviews does need consistent requests. It does not need a burst of identical messages or a retry sequence that treats every non-response as a workflow failure.
Choose the response mode before tuning the writing
At Reputation → Settings → Reviews AI, HighLevel provides three response modes:
- Off: AI responses are disabled.
- Suggestive: HighLevel prepares a draft for a person to review and send.
- Auto Responses: configured Reviews AI Agents can publish replies automatically.
Start in Suggestive mode. Review drafts for positive, neutral, and negative feedback until the business name, perspective, response length, tone, and sign-off are stable. Move only the low-risk categories to automatic publication after the previews repeatedly produce acceptable replies.
Agencies can manage these modes at Agency → AI Suite → AI Setup → three-dot menu for the sub-account → Manage AI access. The agency and sub-account settings synchronize, so changing access centrally affects the same underlying configuration.
Do not confuse review replies with inbox conversations. Reviews AI is for public review responses; broader messaging behavior is a separate configuration described in Conversation AI in GoHighLevel: How It Actually Decides to Reply.
Set tone by sentiment, then constrain the prompt
Reviews AI Agents can be assigned to Positive, Neutral, and Negative sentiment. Each agent has a name, tone, and custom prompt. HighLevel lists tones including Friendly, Professional, and Empathetic, and it can assign agents to specific Google Business Profile pages.
Use these settings as an operating policy:
| Sentiment | Tone | Publishing rule |
|---|---|---|
| Positive | Friendly | Automatic only after preview testing |
| Neutral | Professional | Suggestive until drafts consistently address the actual feedback |
| Negative | Empathetic | Human review required; 1-star reviews never publish automatically |
In Brand Voice, specify the response perspective, exact business name, preferred length, and sign-off. Keep replies short. Turn on Quality Guardrails to reduce repeated openings and potentially risky wording, and use sensitive-review holds wherever human oversight is required.
A useful positive-review prompt is: “Reply in two or three sentences. Acknowledge one specific detail from the review. Do not repeat the reviewer’s full wording. Do not use the same opening across replies. Do not add an offer, promotion, or request for another action. End with the configured personal sign-off.”
For neutral reviews, instruct the agent to acknowledge both the positive and critical parts without arguing. For negative reviews, instruct it not to disclose customer details, speculate about the event, assign blame, or promise an outcome before a person has investigated.
Use the live preview for positive, neutral, and negative examples, regenerating samples while you adjust the settings. Round Robin Assignment can rotate agents for variety, but multiple weak prompts only create varied weak replies. Establish one controlled agent per sentiment before adding rotation.
Never auto-reply to a 1-star review
A 1-star review is not ordinary reputation content. It can involve a disputed service, a safety claim, a refund request, an employee accusation, or information that should not be discussed publicly. An instant empathetic sentence can still be inaccurate or expose private details.
HighLevel does not document a preset literally named “Never auto-reply to 1-star.” Build the rule from its available controls:
- Create a workflow using the New Review Received trigger.
- Filter the rating to exactly 1. HighLevel supports exact rating filters from 1 through 5.
- Route the review to an internal notification or service-recovery step.
- Apply a sensitive-review hold and confirm that no automatic public-reply action is attached to the 1-star path.
- Have a manager inspect the customer record and the underlying job before approving any public response.
The public response should acknowledge the concern without litigating the job in the review thread. Google’s guidance says negative replies should avoid personal attacks and private information, acknowledge mistakes where appropriate, explain limitations honestly, apologize when appropriate, and move complicated resolution to phone or email.
That same handling rule does not require hiding criticism. Google says a mix of positive and negative feedback often appears more trustworthy. The task is to respond accurately and constructively, not to make every public record look uniformly positive.
Do not waste a detailed 5-star review on templated gratitude
A five-star review containing only “Great service” can receive a short reply. A detailed five-star review deserves a response based on what the person actually wrote.
Google recommends replies that are short, relevant, conversational, and non-promotional. It also says businesses should make replies count instead of posting the same thank-you beneath every review. Its personalization guidance includes addressing the reviewer by name, acknowledging specific feedback, and using a personal sign-off.
Configure the positive agent to extract one concrete detail from the review: the service performed, punctuality, communication, cleanliness, or another point the reviewer chose to mention. The reply should reflect that detail without copying the entire review back to the customer.
Bad response:
Thank you for your five-star review! We appreciate your business and look forward to serving you again!
Better response pattern:
Thank the reviewer by name, mention one detail they praised, connect that detail to the team’s work, and close with the configured sign-off.
The second pattern is still structured, but the substance changes with the review. Automation should supply the first draft or publish a controlled response; it should not flatten every customer’s experience into the same sentence.
Review gating is prohibited
Google permits asking genuine customers for reviews when no incentive is offered and the business does not attempt to influence the rating or wording. It prohibits discouraging or preventing negative reviews and selectively requesting reviews only from customers expected to be positive.
Do not send customers through a satisfaction question that directs happy customers to Google while routing unhappy customers to a private form. Do not ask for “a five-star review.” Do not tell customers what to mention, ask them to identify a staff member, set staff quotas for solicited reviews, or pressure customers to post while they are still on the premises.
Discounts, payments, free goods, and free services cannot be offered for posting, revising, or removing a review. Google also treats unusual review volumes or patterns indicating rating manipulation as prohibited activity. Its review-content policy requires genuine, unbiased experiences.
This is not a minor copywriting issue. Google can remove violating reviews and restrict an entire Business Profile by stopping new reviews for a period, temporarily unpublishing existing reviews, or displaying a public warning that fake reviews were removed.
Review Balancing is different from gating
HighLevel’s Review Balancing distributes request traffic between connected platforms by percentage. When Google and Facebook are connected, the documented initial split is 50%/50%, and a ratio such as 70%/30% can be configured. It does not use predicted satisfaction or star rating to decide who sees the Google link.
That neutral platform distribution is not sentiment-based gating. Keep the distinction clear: distributing all genuine customers between connected platforms is a routing setting; filtering out likely critics is prohibited selection.
A controlled rollout for a business with 11 reviews
The competitor’s 300 reviews do not change the operating rules. Build a repeatable completion process, send compliant requests to genuine customers, and protect the public reply channel from careless automation.
- Week 1: Create the Completed trigger, wait 1 hour, and enable the SMS request. Disable the QuickBooks request automation if it is active.
- Week 2: Add the email request at 24 hours and two SMS retries at 3 days and 1 week.
- Week 3: Configure Brand Voice and one agent for each sentiment. Keep Reviews AI in Suggestive mode while checking every draft.
- Week 4: Enable automatic publication only for tested low-risk positive replies. Keep neutral and negative drafts under human review, with the exact 1-star workflow branch always held.
Track whether completed jobs actually enter the request workflow, whether links are clicked, which channel generated the click, and whether replies required edits. Those checks reveal workflow failures and weak prompt rules without turning review collection into a campaign for manufactured praise.
What to do
- Trigger the review workflow only when the appointment or pipeline status changes to Completed, then wait 1 hour before running Send Review Request.
- At Subaccount → Settings → Integrations → Quickbook Integration, deselect Send Review Requests for contacts when job completion is the chosen trigger.
- Set the initial SMS request delay to 1 hour, the email request delay to 24 hours, and the maximum SMS retries to 2.
- Schedule SMS retry 1 for 3 days after the initial message and SMS retry 2 for 1 week later; assign a different template to each stage.
- Start Reviews AI in Suggestive mode and preview positive, neutral, and negative replies before allowing any automatic publication.
- Set the positive agent to Friendly, the neutral agent to Professional, and the negative agent to Empathetic; require human review for neutral and negative replies.
- Create a New Review Received workflow filtered to rating equals 1, send an internal notification, apply a sensitive-review hold, and attach no automatic public-reply action to that path.
- Configure the positive-agent prompt to use one specific detail from the review, respond in two or three sentences, omit promotions, and end with the configured personal sign-off.
- If Review Balancing is enabled for Google and Facebook, use a fixed 70% Google and 30% Facebook distribution without using satisfaction or predicted rating as a routing condition.
Questions people ask
Should the request go out immediately when the job is completed?
Use a 1-hour delay after the Completed status. This keeps the request close to the service while avoiding an immediate message as the customer is leaving.
Should SMS and email be sent at the same time?
No. Set SMS for 1 hour after the workflow action and email for 24 hours later. Their delays, retries, senders, and templates are configured independently.
Can Reviews AI automatically answer every positive review?
Auto Responses can publish through configured Reviews AI Agents, but first test the positive agent in Suggestive mode. Detailed five-star reviews should receive replies that acknowledge a specific part of the feedback rather than generic gratitude.
Does HighLevel have a built-in switch called Never auto-reply to 1-star?
No preset with that exact name is documented. Use a New Review Received workflow filtered to exactly 1 star, route it to internal service recovery, apply human oversight, and ensure that path has no automatic public-reply action.
Is sending unhappy customers to a private feedback form allowed?
Not when the purpose is to prevent them from receiving the public review opportunity offered to happy customers. Google prohibits discouraging negative reviews and selectively soliciting positive reviews.
The request workflow, channel cadence, response agents, and 1-star hold all need to agree with one another; otherwise, the system automates the wrong event or publishes a reply that should have waited. We build voice and conversation AI inside GoHighLevel. Book a call and we'll tell you if it's the right fit — including when it isn't.