Review responses are the worst kind of marketing work: high stakes, low creativity, and impossible to delegate to someone who doesn't know that the pool at Ridgeline has been closed since March.
So they pile up. Forty of them across six communities, some three weeks old, and Google is very publicly counting how long you took. Meanwhile the responses that do go out sound like they came from the same person, because they did, and that person was tired.
Here's the workflow that fixes it. Batch, not one at a time. And the whole trick is what you give the model before you give it a single review.
Step one: build the six-community brief (do this once, use it for a year)
The reason AI review responses read like a form letter is that people paste in one review and ask for a reply. The model has nothing to work with, so it writes the average of every apartment review response ever published. "We're so sorry to hear about your experience. Please reach out to the office at your convenience." Congratulations, you've automated the thing residents already ignore.
Fix it upstream. Open a doc and write a short block for each community. Not a brand book. Ten lines, max:
- Community name, and how staff actually refer to it out loud ("Ridgeline", not "The Residences at Ridgeline Park")
- Manager's first name and whether they sign responses personally
- The office phone and email you want people directed to
- Known open issues, dated: "package room camera down since 10/3, vendor scheduled 11/1", "valet trash inconsistent Tuesdays", "elevator 2 modernization runs through December"
- Recent wins you're allowed to name: new gym equipment, dog park resurfaced, gate motor replaced
- Voice notes: does this community's audience skew renewal-heavy and long-tenured, or is it a lease-up where half the reviewers moved in six weeks ago?
- Anything legal or PR you never put in writing: pending litigation, an eviction dispute, an injury, anything involving another resident's behavior
That last one matters more than the rest combined. We'll come back to it.
Save this as your standing brief. You paste it at the top of every review session. When Ridgeline's package room gets fixed, you update one line.
Step two: dump the reviews raw
Don't clean them up. Copy them out of Google, Yelp, ApartmentRatings, whatever your reputation tool exports, and paste them in a numbered list with three fields: community, star rating, review text. Include the reviewer's first name if you have it. Typos, all caps, the person who wrote a 400-word review about a parking spot. Leave it all in.
Forty reviews in one paste is fine. That's the point.
Step three: the batch prompt
The prompt block at the end of this lesson is the one to use. But understand what each constraint is doing, because you'll want to adjust it:
Word cap by rating. Five-star reviews get 25 to 40 words. One and two stars get 60 to 90. Three and four stars get 50 to 70. This single rule kills 80% of the form-letter feel, because form letters are all the same length. A gushing five-star review that gets a three-sentence corporate response looks worse than no response.
No apology in the first sentence of a positive response. Models love to hedge. Ban it.
Name one specific thing from the review. If someone mentions the maintenance tech by name, the response uses that name. If someone mentions the commute, the response mentions the commute. If the review is genuinely contentless ("Good place"), the response is short and warm and doesn't invent detail.
Vary the openers across the batch. Tell it explicitly: no two responses in this batch may start with the same three words. Otherwise you get forty responses beginning "Thank you for" and residents scrolling your Google profile will notice, because that profile is a single scrollable page.
Route, don't resolve. For negative reviews, the response acknowledges, states what's actually happening if you have a dated fact from the brief, and moves the conversation offline to a named person. It never promises a fix you haven't scheduled and never argues.
Step four: the ten-minute pass that makes this legitimate
You get forty drafts back. Now read them.
Sort by rating, worst first. The one and two stars get real attention: is the fact you cited still true, is the person you're routing them to actually the right person, does it accidentally confirm something you shouldn't confirm in public. The five stars you skim for weirdness. Models occasionally thank someone for feedback they did not give.
Two things to catch specifically. First, invented specifics. If the brief didn't mention a fitness center renovation, and the response mentions a fitness center renovation, delete it. The model filled a gap. Second, confirmations. A response saying "we're sorry about the ongoing issues with your neighbor" has now published, under your brand, a statement about another resident. Rewrite to acknowledge the reviewer's experience without characterizing anyone else.
Then paste and post.
One thing to be careful about
Review responses are public housing-related communications, and models drift toward describing people. A response that says "this community is perfect for young families like yours" or "we love having students here" describes who lives there, and that is a fair housing problem regardless of intent. The fix is the same one that applies to your listing copy: instruct the model to describe the unit, the community and the service, never the resident or the resident's demographic. Add that line to the standing brief and check for it on the read-through, particularly on positive responses where the model gets warm and chatty.
What this actually buys you
Not just speed, though the speed is real: the writing goes from the bottleneck to the fast part, and your time moves to the ten-minute review pass where your judgment matters.
The bigger win is that response rate and response time both become achievable numbers instead of aspirational ones. Google's local ranking factors reward engagement signals, and unanswered reviews are the most visible sign that nobody's home. Prospects read responses. They read them specifically to find out whether management answers when something goes wrong.
Run this monthly per portfolio. Update the brief when something changes. And keep the negative responses in a separate doc, because in six months, when asset management asks what residents are actually complaining about at Ridgeline, you'll have the answer in one place and won't have to reconstruct it from Google.
Where it breaks
Three places. Long reviews with multiple unrelated complaints get flattened into one vague acknowledgment, so pull those out and handle them by hand. Reviews that reference something legal or an injury should never go into the batch at all, full stop, they go to your legal or ops contact. And if your standing brief goes stale, the model will confidently cite a fix that never happened, which is worse than saying nothing. Date every line in the brief so stale facts are obvious.