VRMA is the big one. Every fall the vacation rental management industry gets in a room together, and this year the room was in Nashville. Our team spent three days there, in sessions, at dinners, and walking a vendor floor that gets longer every year. Most of the people in that building manage hundreds of units.
Which is what made one chart worth the trip.
AirDNA put up average review scores broken out by portfolio size. The line goes down as the portfolios get bigger. A host with one listing averages 4.86 stars. An operator with 250 or more listings averages 4.67. Even the top 10% of the large operators only reach 4.81, which is still below the typical person running a single property out of a spare bedroom and a phone.
If you host one, two or five properties, you walked into this year holding the thing the whole industry spent three days trying to buy. This article is what the data said, and what to do with it.
Those decimals are the business now
It is tempting to read 4.86 against 4.67 as a rounding error. It is not. Two numbers from the same session:
- Listings rated 4.9 or higher earn about 23% more RevPAR than listings under 4.6. RevPAR is revenue per available night, so that is the blended effect of both price and occupancy.
- Airbnb's Guest Favorite badge, the top 1% of listings, carries roughly 19% above market.
Then there is the number that explains why every manager in Nashville was taking a photo of that slide. Managers averaging under 4.7 lost 26.5% of their listings over the last twelve months. Managers at 4.9 and above lost 13.3%. Roughly double the churn, decided by two tenths of a star.
For a small host, churn is not the lesson. Pricing power is. You are already sitting at an average the national operators cannot reach at scale, and the gap between where you are and 4.9 is usually three or four fixable operational details rather than a strategy. The ones that show up most often are here, and if your scores have drifted down while you changed nothing, it is worth understanding how the standard moved under you.
Why quality matters more this year than last
Review scores have always paid. The reason they now decide the year is on the supply side.
AirDNA opened the conference with its economic outlook, and US listings grew 1.6% year over year, the slowest pace since the 2021 reopening. For context, supply grew more than 20% in both 2022 and 2023. The land rush is over.
The rest of the picture:
- 60 US markets are actually shrinking, and not all of them are small ones.
- Demand is growing 1.4% to 2.1%, so roughly in line with supply.
- Occupancy has held near 57% for two years. It found a floor and stayed there.
- RevPAR is forecast up about 3.3% for 2026, which is close enough to inflation to call it flat.
Read those four together and you get the single most useful sentence from the whole conference: there is no longer a rising tide. New guests are not arriving fast enough to lift every listing in a market. If your bookings grow in 2026, you took that guest from the property down the street.
That is not a bad position for a small operator. It is a very bad position for anyone whose plan was to do an average job in a growing market. The lever that still works is the quality of the stay, which is the lever you personally control and a 250-unit manager does not.
One bright spot in the supply data worth noting if you are shopping: properties with six or more bedrooms saw occupancy rise 1.5% and nightly rates rise 6.6% this year. Big groups are still underserved in most markets. We looked at how property type changes the math in this breakdown, and the large-home signal has only gotten stronger since.
Airbnb went all in on AI, and it changes how you get found
Channel mix first, because it frames everything else. Airbnb is 52% of US nights booked, Vrbo is 14%, and direct plus everything else is 33%. Only 28% of listings worldwide are on two or more channels. When Airbnb ships a product change, it is not one option among several, it moves the market.
On September 30, Airbnb launched AI search in the US. Guests can now search by text or by voice. The results pages show AI-written summaries of your listing and side-by-side comparisons against other homes. Host tools are promised later this fall, including better dynamic pricing and a multi-listing calendar, and Brian Chesky has publicly committed to a full AI travel agent in 2027.
Here is what that means in practice. For twenty years, the job was to get a human to look at your photos. Now a model reads your listing, writes a two-line summary of your property, and puts it next to two competitors for comparison. You do not write that summary. Your listing data does.
Two things to do this week, both free:
- Fill in every single amenity field. Not the ones you think matter, all of them. An amenity you never ticked cannot appear in a summary and cannot win a comparison. This is the cheapest work available right now and almost nobody has done it.
- Write plainly and specifically. Bed configuration, parking, workspace, distance to the thing people come for. Atmospheric copy that says "your perfect mountain escape awaits" gives a model nothing to extract. Facts do.
If you run a direct booking site, the same shift is happening off-platform. Tools like the Semrush AI Visibility Toolkit now show which AI prompts surface your site and which surface your competitors. That is one more argument for owning the guest relationship rather than renting it, particularly when 52% of your nights come from a company rewriting its own front door.
The AI pricing exercise you can run on one listing
The most practical session came from Jasper Ribbers of Freewyld, who showed how to build an AI revenue analyst out of four PriceLabs exports and one prompt.
The framing is what makes it work. He is not asking the model for the perfect price, because there isn't one. He is asking it to find the obviously wrong ones, and he grades them the way chess engines grade moves:
- Blunder: 30% or more outside your range
- Mistake: 20% to 30% outside
- Inaccuracy: 10% to 20% outside
The arithmetic behind it: 25 listings across 365 days is 9,125 individual prices. Nobody audits that by hand, so in practice nobody audits it at all. The model flags the outliers and you make the call, because you are the one who knows about the festival in June, the road closure in March and the floor you promised the owner.
And if you have one listing, you still have 365 prices, which is more than you have ever actually reviewed. Run the same prompt on a smaller export. The reason this stood out at a conference full of enterprise software is that it costs nothing and scales down perfectly. If you have never pulled your pricing apart by segment, start with the four prices on your listing and then let the model look for blunders.
A couple of good filters came out of other sessions on where AI belongs. Guesty suggested four questions: is the task repetitive, can you define what good looks like, what is the damage if it gets it wrong, and can you measure the result. Renjoy and PriceLabs offered a blunter version that I liked more. If a person still has to touch it to finish, it is not automated. Most "automations" fail that test.
Six things from the hallway
The best ideas at VRMA are rarely on the slides. These came from conversations between sessions, and all of them are small enough to do this month.
- Thank your cleaner by name. Set an automatic text to your cleaner whenever a guest review mentions how clean the place was. It costs nothing and it protects the single most important person in your operation, which matters because a bad housekeeper is a revenue problem, not a staffing one.
- Read the thread before you ask for a review. If a guest sent three messages about the wifi, do not send the cheerful review request. Fix it, then ask.
- Find your missing SOPs. Run a month of guest messages through an AI and ask what questions keep coming up. Every recurring question is a gap in your check-in instructions or your listing.
- Log the searches that did not book. On a direct booking site, record the date ranges people searched and abandoned. A pile of them on the same dates usually means your price is wrong, not your property.
- Launch in the off season. A new listing that opens before peak collects its first reviews when stakes are low, so it arrives at high season with a score that can carry a higher rate.
- Explain the why. With owners, co-hosting clients or cleaners, a rule gets compliance and a reason gets buy-in. Different outcomes entirely.
If you are buying rather than hosting
Two of these numbers belong in an underwriting model, not a to-do list.
Supply growth at 1.6% with 60 shrinking markets means the market-selection question has changed shape. For three years, almost any decent market absorbed new inventory. Now you are buying into a fixed pie in a lot of places, and the question is not "is this market growing" but "can this property beat the specific listings it will sit next to." Pull the comp set and look at review scores, not just rates. A street full of 4.6s is an opportunity. A street full of 4.9s is a fight.
The six-plus bedroom data point is the other one. Occupancy up 1.5% and rates up 6.6% in a flat year is the clearest demand signal in the outlook, and large homes are harder to operate, which is exactly why fewer people compete there. If you are screening deals this winter, run them through this first and pay attention to group capacity as a differentiator rather than a nice-to-have.
Pick one thing
It is easy to assume conference data is for people with 300 doors. This year, most of it pointed the other way. Single-listing hosts out-rate every category of large operator in the country. The AI pricing prompt works just as well on one calendar as on twenty-five. Filling in your amenity fields takes an afternoon and costs nothing.
So choose one before the week is out:
- Audit your amenity fields, all of them, so Airbnb's AI has something to work with.
- Pull your last twenty reviews and find the one recurring complaint standing between you and 4.9.
- Run a pricing export through the blunder test and fix the worst ten nights.
Supply has stopped growing, quality is what gets paid, and the search box is now a model instead of a filter. All three favor the host who sweats the details.
Holding 4.9 across a lot of properties is genuinely hard, which is the whole reason that chart slopes downward. It is what our team does every day across the portfolio we manage. If you want a second set of eyes on your scores, your pricing or a deal you are screening this winter, book a strategy call and we will go through it with you.
