Where reviews do matter
of answers about a community by name involve AI searching for reviews.
Once a renter has 2 or 3 options that fit, it's time to weigh them.
Say the shortlist has three communities.
Before AI, renters went somewhere specific for reviews, like Yelp, Google or ApartmentRatings. Now there's no URL to type.
They ask the model instead. Of these three, which one has the best reviews?
This is where the model looks for reviews. Review, forum and social sites make up about half of what it cites here (48 to 57%).
Then it gives the renter its take on each place. 42% of the specific claims it makes are negative.
So the question becomes what AI is saying about you.
A star rating and a review count.
Operators track their average on Google, Yelp and ApartmentRatings and work to push it up. That made sense when renters read the reviews themselves.
Residents like the location and the layouts, but reviews keep mentioning slow maintenance, broken gates and limited parking.
MixedSame community. AI calls it mixed.
The renter sees the model's take, positive, mixed or negative, with the specific complaints it pulled from wherever it looked.
Simply monitoring your star rating on one source doesn't really do you anything. You have to monitor from the front door.
Austin Lo, Peek
That's why we're launching AI Sentiment (beta), a new part of Discover. It asks the models what renters ask about you and reads the answers. Here's what it found when we pointed it at real communities.
Two communities, equally visible. AI described them very differently.
Visibility gets you on the shortlist. Once a renter asks about you by name, what the model says decides the rest.
A renter comparing them would mostly hear good things about A and mostly mixed or negative takes on B. A visibility score alone wouldn't show that gap, which is why Discover tracks both.
Each model tells its own story.
Same community, same month.
One test community.
So check each model on its own. AI Sentiment scores ChatGPT, Gemini, Claude and Perplexity separately.
AI repeats the same complaints over and over.
Share of answers raising each objection, one test community. Across about 2,100 complaints about 25 communities, a list of 44 objections covered about 80% of them.
The good news is your property teams can fix these. AI Sentiment sorts what AI praises and faults into 13 topics, with the specific complaints under each, so you know where to start.
That's how far a community's sentiment score can move from one month to the next with nothing changed at the property.
So judge it by the trend over a few months.
of the star ratings AI quoted didn't match its other answers, usually because of an old listing or a sister property.
AI Sentiment shows where each rating came from.
How AI Sentiment works
answers per community, every month.
The goal is to hear what your renters hear, as closely as we can. So every month AI Sentiment asks the four major models the questions renters ask by name, and reads every answer.
What happens every month
- It asks what renters ask. It puts four types of question about your community, by name, to ChatGPT, Gemini, Claude and Perplexity. Each question is worded six different ways.
- It reads every answer. Each one is scored for tone: positive, mixed or negative. Those roll up into your overall score and a score for each model.
- It sorts what AI says. Praise and complaints go into 13 topics, like maintenance, staff and noise, with the specific complaints listed under each.
- It tracks where AI looked. It records the star rating AI quoted, where that rating came from, and every site each model cited.
- It reports it back to you. You see the overall score, each model's score, the topics and complaints, the sources, every answer word for word, and how it all moved since last month.
Every question, every model, six ways
One answer can be a fluke: AI's answers shift with wording and from month to month. So AI Sentiment asks each question six ways, on every model, every month.
Each dot is one answer.
Your AI Sentiment score is the tone of what AI tells renters about your community, with the specifics behind it. Your star rating is one input into that.
Your AI Sentiment report
Every answer is saved word for word, so after you fix something, like a stale listing or an unanswered complaint, you can see whether AI's take on you changes.
Don't wait for AI to show up in your web analytics. Most renters who hear about you from AI don't click a link. They come back later and search your name, and your analytics counts that visit as Google or direct traffic. AI's effect is real, but it's hidden inside your branded search. So measure what AI says about you, not just the clicks it sends.
AI Sentiment figures come from testing about 2,000 AI answers across 25 communities. Some come from a single test community, as noted.