A season in numbers: what guests asked and how fast they got an answer

On a sample of 434 conversations from 6 to 13 September 2026, the Supalobby agent replied in a median 3.6 seconds, and 9 replies in 10 came within 8.1 seconds. It handled 46 percent of text conversations alone, drafted 25 percent for the host and handed 30 percent over. One AI message cost us about 1 US cent.

The title promises a season and we have one week at the end of summer. It is a small sample and we present it as one: what we measured, what we do not measure yet, and how the numbers sit next to outside research on reply times, on what guests want to know before arrival, and on how much AI closes by itself.

Where the numbers come from and what they do not tell you

These are Supalobby's own figures from one week, 6 to 13 September 2026: 434 conversations handled by the agent. Everything we measured fits in this table.

The three text-conversation shares add up to 101 because of rounding.

What these numbers do not tell you. A week in September is the tail of the summer season, not the whole season. The sample covers properties that happened to use Supalobby that week, so it is not a cross-section of the market. With 434 conversations, one percent is about four conversations, so a difference of a few points may be noise. We do not yet measure:

• which topics guests ask about (directions, check-in, parking, Wi-Fi),
• the split by channel and by language,
• the time of day guests write,
• how long the host took to send a draft or reply after a hand-over,
• how many conversations ended in a booking.

Where we would want our own number and do not have it, we give outside research and say whose data it is.

How fast guests got a reply

A median of 3.6 seconds, and within 8.1 seconds in 90 percent of cases. That is the time from the guest's message to the agent's reply. For drafts and hand-overs, the guest then waits for the host, and we do not measure that time.

For comparison, Airbnb data. Rezihost analysed Inside Airbnb data for 92,597 listings in New York, Los Angeles and Austin (updated 2 July 2026).

Airbnb tracks two things. Response rate is the share of new inquiries and reservation requests from the past 30 days that you answered within 24 hours. Response time is the average time it took you to respond to all new messages in the past 30 days. The best label Airbnb shows guests is "within an hour". Seconds do not get their own shelf.

How many conversations the agent closed alone, and how many it handed over

The agent sent the reply itself in 46 percent of text conversations. In 25 percent it prepared a draft that the host approved or edited. In 30 percent it decided it was not confident and handed the thread to the host.

Drafts come from two product rules. On Airbnb the agent only reads the thread and prepares the reply; the host sends it. In Copilot mode the agent drafts in every channel. We do not yet measure how many drafts in the sample came from Airbnb and how many from Copilot mode.

Market figures are higher and need careful reading:

• Hotel Technology News (21 July 2025, Jordan Hollander) writes that AI agents in leading guest messaging platforms handle 70 to 90 percent of routine inquiries without human intervention. That is hotels, and routine inquiries only.
• Conduit states on its site that at its client Cash Flow Street, 96 of every 100 guest messages are resolved without the operator touching them. That is a vendor's claim about one client.

Our 46 percent counts every text conversation in the week, routine or not, including channels where the agent never sends by design. It is not the same measure. We think it is fairer to show it as it came out than to pick a subset that looks better.

Hand-overs are 30 percent. For the host, each one is a message to answer personally. That beats an agent guessing whether a late check-in is fine, but it also marks the places where the property's knowledge base has gaps.

What guests ask about

Directions and check-in, most of all. That comes from outside data, because we do not yet break our own sample down by topic.

Touch Stay surveyed 1,204 guests worldwide in late 2025 and early 2026. Close to nine in ten said they want detailed check-in information before arrival, including access codes and parking. In Touch Stay's own guidebook network the order is: location and directions, check-in, food, amenities, booking questions, Wi-Fi, parking. Four of the seven are arrival questions.

Conduit, one of the largest AI guest messaging companies, had handled more than 50 million guest and operational messages across more than 100,000 properties, including over 10 million voice interactions, according to ShortTermRentalz (19 June 2026). We could not find a breakdown of those messages by question type on the pages we opened.

That order gives you a ready pre-arrival message. Send it the day before check-in and half the questions never come:

Hi [name], we're looking forward to seeing you tomorrow.
Address: [street, number, postcode], on Google Maps: [link].
Check-in from [time]. Door code: [code], key box: [where].
Parking: [where, free or paid, space number].
Wi-Fi: [network] / password: [password].
Nearest shop and restaurant: [name, distance].
Any questions, just reply here.

Does a fast reply turn into bookings

The data shows a link, not proof. IntelliHost (7 November 2023) analysed over 5,000 Airbnb properties:

Properties replying within an hour also had about 16 percent more daily impressions. IntelliHost itself notes these are correlations shaped by many variables.

Rezihost adds two observations. Hosts who reply within an hour are Superhosts 52.6 percent of the time, against 2.6 percent for hosts who take days. Average ratings are almost flat across the first three time buckets (4.774, 4.771 and 4.785) and drop to 4.644 only for hosts who take days. "Responsive" appears in 13.2 percent of reviews since January 2024, up from 9.4 percent all-time, and 77 percent of communication complaints in reviews involve unanswered messages.

What we take from this: the big penalty is for silence, not for a quarter-hour delay. Seconds will not lift your rating if you already reply within the hour. In our view they matter in one situation: a guest messaging three listings at once and booking whichever answers first. None of the studies above measures that, and neither do we.

Where Supalobby fits

The Supalobby agent reads Booking.com, Airbnb, WhatsApp, Instagram, e-mail, Slowhop, Aloha Camp and the website chat in one inbox. In Copilot mode it drafts every reply; in Autopilot it replies on its own, switched on per channel, and when it is not confident it asks the host and hands the thread over. On Airbnb it always drafts only.

The figures here come from one sample of 434 conversations in September 2026 and are not a promise for every property. We do not yet measure the split by topic or channel. The measured 0.037 PLN per message is our cost; the host's price on Autopilot is 49 PLN net, about 11 euro, per unit per month.