What Airbnb's AI Writes on Search Cards: A Study of 837 US Cards (October 2026)
By Jonathan Miksis · October 23, 2026 · 25 min read

Airbnb's AI now writes the line guests read on US search cards, in place of the host's title. To see what it actually says, we read 837 cards on airbnb.com across 12 US markets on October 23, 2026, coded every line with written rules checked line by line on a sample, and matched 183 of the lines against the hosts' own listing pages. The data is free to download.
I run AI Hustle Guy. Since 2020 I've done 100+ listing audits and partnerships with hosts and driven over 3 million page views to vacation rental content on my travel blog, Global Viewpoint. This study expands the 54-card check in Airbnb's 2026 AI changes for hosts and is part of my AI for Airbnb hosts hub.
The short answer (October 23, 2026): Airbnb's line never showed the host's title, came in a short or a long version depending on the visitor, and mostly repeated what hosts wrote in their description and ticked in their amenities. Airbnb's UK, Canadian, Australian and Irish sites still showed the host's title, often cut short.
Key findings
- No titles on US cards. On October 23, 2026, none of the 837 airbnb.com search cards we read in 12 US markets showed the host's title; every card showed an AI-written line under the home type and town.
- Two versions of the line. Airbnb was serving a short line (median 38 characters) to some logged-out visitors and a long one (median 74 characters) to others. In 15 fresh visits, 7 got the short version and 8 the long one, and the visitors we followed across repeat searches never switched.
- The long line is an amenity list. 96% of long lines named an amenity, 2.5 on average, led by pools (40% of lines), hot tubs (37%) and decks or balconies (36%); 18% mentioned free parking. 58% of short lines named an amenity.
- The short line leans on location. 72% of short lines named the home's setting or a nearby place, such as lakefront, downtown, 2 minutes from Acadia or across from UGA's Arch.
- It goes beyond your title. 86% of short lines and 97% of long lines named something the host's title didn't mention.
- Its words come from your description. Of 696 details named in 183 lines, 91% appeared in the host's description, 54% in the amenities list and 35% in the title. None named an amenity the listing marked as unavailable.
- Few sales words. 1% of short lines and 10% of long lines used words like luxury, epic or amazing, against 14% of the hosts' own titles.
- Other countries' sites still show titles. All 160 cards we read on Airbnb's UK, Canadian, Australian and Irish sites showed the host's title, and 45% of those titles were 48 characters or longer, long enough to be cut on every screen we've measured.

Airbnb's line on 837 US cards, October 23, 2026, next to the hosts' own titles from the same page data.
How we did the study
The sample
On October 23, 2026, between 11:31 and 11:41 a.m. Eastern, we ran a logged-out "homes" search on airbnb.com for each of 12 US markets, with no dates and no filters, and read the first two results pages (18 homes each). We ran every search twice, as two visitors whose cookies we kept like a browser does: one Airbnb was showing the short line and one it was showing the long line, so each market has up to 36 homes per version.
| Market | Type of market | Short-line cards | Long-line cards | Most common home types |
|---|---|---|---|---|
| Gatlinburg, TN | Mountain cabins (Smoky Mountains) | 33 | 33 | Cabin, condo |
| Destin, FL | Gulf beach town | 34 | 34 | Condo, home |
| Chicago, IL | Big city | 33 | 36 | Apartment, condo |
| Joshua Tree, CA | Desert | 35 | 36 | Home, cabin |
| Lake of the Ozarks, MO | Lake | 36 | 36 | Home, condo |
| Breckenridge, CO | Ski town | 35 | 31 | Condo, cabin |
| Athens, GA | College town | 36 | 36 | Condo, apartment |
| Bar Harbor, ME | National park gateway (Acadia) | 36 | 36 | Home, apartment |
| Lancaster County, PA | Rural farmland | 36 | 36 | Home, cottage |
| Kissimmee, FL | Theme parks, family pool homes | 35 | 36 | Home, apartment |
| Nashville, TN | Music city, groups | 36 | 30 | Apartment, guesthouse |
| Kihei, HI | Tropical beach condos | 36 | 36 | Condo, apartment |
| Total | 12 markets | 421 | 416 | 713 different listings |
- Homes only. We left out 39 hotel cards, which show the hotel's name instead of an AI line. In Chicago the long-line visitor's first two pages were mostly hotels, so we read two more pages there.
- One card per listing per version. 124 listings turned up in both versions, which lets us compare the two lines for the same home.
- From the page data, not screenshots. Each search page carries the card's home type and town, Airbnb's line, the host's title (still there, just not displayed), the rating and the badges. We read those fields directly.
- Listing pages. For 102 listings, 12 in each of the first five markets in the table and 6 in each of the other seven, weighted toward homes seen in both versions, we also read the public listing page: title, description, amenities, photo captions and Airbnb's own highlights. Airbnb rate-limited us once while we did this, so we paused for about 20 minutes and finished at one page every 30 to 40 seconds.
- Repeat visits. We loaded some searches again as the same visitor and as new visitors, 73 US page loads in all, to see whether a listing's line changes.
- Outside the US. The same afternoon we read the Gatlinburg and Nashville searches on airbnb.co.uk, airbnb.ca, airbnb.com.au and airbnb.ie, plus London on airbnb.co.uk and airbnb.com: 160 cards on the four non-US sites.
How we coded each line
Each line was coded with written rules: lists of words for each kind of detail, such as pool, hot tub or jacuzzi for amenities, and lakefront, downtown or minutes to a named place for location. A line can get several codes.
| Code | What counts |
|---|---|
| Amenity | Something at the home guests use: pool, hot tub, sauna, game room, theater, fire pit, fireplace, kitchen, deck, balcony, yard, dock, kayaks, parking, EV charger, workspace, king bed |
| Setting or nearby place | Where the home sits (waterfront, wooded, desert, acres, downtown, a resort) or what it's near (a distance, a walk, a named place) |
| Building | Type beyond the card's label (A-frame, log cabin, loft), age (new, renovated, 1880), design (modern, Art Deco), size or floor |
| View | Any view, panorama, sunset or "overlooking" |
| Sales word | Luxury, epic, amazing, gorgeous, incredible, spectacular, magnificent and similar words |
| Beyond the title | Names at least one amenity, setting, place, building detail or view the host's title doesn't, counting abbreviations (Mtn, HT) and synonyms (jacuzzi for hot tub) |
Checked line by line. We also coded 160 lines one at a time against the codebook, without the word lists, and compared the two. On a first sample of 100, the rules agreed with the line-by-line codes on 88% to 100% of lines, depending on the code.
We fixed the patterns behind the misses, then coded a fresh sample of 60 before looking at the rules' output: agreement was 93% for building details, 95% for location and 100% for amenities, views, sales words and beyond-the-title. We fixed those few misses too, and every number below uses the final rules.
Results
Two versions of the line, depending on the visitor
Airbnb showed one of two versions to each logged-out visitor, and kept showing it:
- Short line: 21 to 55 characters, median 38, about six words. "Octagon cabin with mountain views and hot tub."
- Long line: 48 to 89 characters, median 74, about 13 words, almost always "a place with a list": "Octagon cabin with sweeping mountain views, a hot tub, a firepit, and a fireplace." 413 of 416 long lines were longer than the 50 characters Airbnb allows in a title.
Both examples are the same Gatlinburg cabin. In 15 fresh logged-out visits, 7 landed on the short version and 8 on the long one. The four visitors we followed across repeat searches, 2 to 31 page loads each, never switched.
The note above the results, "Descriptions are based on details hosts provide," appeared in both versions. Airbnb hasn't announced the long version; it looks like a test.
The short line didn't change. Every time a listing showed up again in the short version, for the same visitor or a different one, the line was identical (91 of 91 pairs of sightings). The long line mostly didn't: 133 of 135 pairs of sightings by the same visitor were identical, but different visitors saw different wording for 16 of 61 listings, such as "a fire pit, a gas grill, and free parking" for one visitor and "a grill, fire pit, & free parking" for another.
We read the lines from the page data, so we can't tell you how much of a 74-character line fits on a card before it's cut. Look at your own card in a logged-out window to see.

Share of lines naming each kind of detail. A line can name several.
What the line names
| What we counted | Short line | Long line |
|---|---|---|
| Cards | 421 | 416 |
| Showed the host's title | 0 | 0 |
| Length in characters: median (middle half) | 38 (34 to 41) | 74 (62 to 79) |
| Shortest and longest | 21 and 55 | 48 and 89 |
| Words: median | 6 | 13 |
| Named an amenity | 58% (244) | 96% (401) |
| Named the setting or a nearby place | 72% (305) | 66% (275) |
| Named the building's type, age, design or size | 50% (211) | 54% (224) |
| Named a view | 26% (111) | 31% (127) |
| Included a number | 25% (104) | 46% (190) |
| Said who it suits (couples, families) | 0% (1) | 3% (11) |
| Used a sales word (luxury, epic, amazing) | 1% (5) | 10% (42) |
| Used pipes, bullets, asterisks, tildes or ! | 0% (0) | 0% (0) |
| Named something the host's title didn't | 86% (364) | 97% (405) |
| Amenities named per line, on average | 0.8 | 2.5 |
| Details named per line, on average | 2.9 | 4.5 |
Amenities vs views. Views came well behind amenities and location: 26% of short lines and 31% of long lines named one, and almost always next to an amenity in the long version (125 of the 127). Amenities without a view were the most common pattern in both versions: 42% of short lines and 66% of long lines.
The same home, both versions. For the 124 listings we saw in both versions, 92% of the long lines and 56% of the short ones named an amenity, 2.4 against 0.7 on average. The long line kept 75% of the short line's details, but only about half of the long lines (52%) kept all of them. The extra words went to amenity lists, not to sales talk: 6 of the 124 long lines used a sales word, against 2 of the short ones.
Most-used words. "View" or "views" was the most common word in short lines (106 of 421). In long lines it was "private" (225 of 416, as in "a private hot tub" or "a private balcony"), and "free parking" appeared in 74.
Most-named details, by version
| Rank | Short line | Long line |
|---|---|---|
| 1 | A distance or walk to somewhere: 37% | Pool: 40% |
| 2 | A named place (Dollywood, UGA's Arch, Acadia): 36% | Hot tub or jacuzzi: 37% |
| 3 | A view: 26% | Deck, balcony, patio, porch or lanai: 36% |
| 4 | The building type (A-frame, log cabin, loft): 24% | A view: 31% |
| 5 | Hot tub or jacuzzi: 17% | Waterfront, lakefront, beachfront or creekside: 26% |
| 6 | New, renovated or historic: 17% | The building type (A-frame, log cabin, loft): 24% |
| 7 | Pool: 16% | New, renovated or historic: 19% |
| 8 | Deck, balcony, patio, porch or lanai: 15% | Fire pit: 19% |
| 9 | Waterfront, lakefront, beachfront or creekside: 15% | Parking (usually free parking): 19% |
| 10 | Design (modern, rustic, Art Deco): 11% | Design (modern, rustic, Art Deco): 16% |
By market
The line follows what each market sells. Mountain and desert lines led with hot tubs, lake lines with docks and the water, city and college-town short lines with a named place or a walk to one.
| Market | Short line: most named | Long line: most named |
|---|---|---|
| Gatlinburg, TN | hot tub 70%, view 49%, building type 24% | hot tub 91%, view 55%, pool 36% |
| Destin, FL | view 65%, deck or balcony 47%, floor 44% | waterfront 85%, pool 68%, view 62% |
| Chicago, IL | named place 79%, distance 58%, building type 46% | building type 36%, deck or balcony 33%, new or historic 33% |
| Joshua Tree, CA | hot tub 46%, building type 46%, acres or farm 43% | hot tub 83%, pool 75%, desert 53% |
| Lake of the Ozarks, MO | waterfront 61%, view 42%, deck or balcony 33% | waterfront 86%, dock 67%, kayaks or boats 42% |
| Breckenridge, CO | distance 80%, named place 63%, view 34% | hot tub 74%, fireplace 65%, distance 52% |
| Athens, GA | distance 58%, named place 53%, downtown 25% | new or historic 42%, parking 42%, distance 39% |
| Bar Harbor, ME | distance 50%, named place 50%, building type 33% | parking 56%, fire pit 42%, new or historic 31% |
| Lancaster County, PA | new or historic 31%, downtown 31%, building type 25% | fire pit 53%, deck or balcony 42%, building type 39% |
| Kissimmee, FL | pool 54%, distance 54%, named place 51% | pool 92%, game room 50%, theater 50% |
| Nashville, TN | named place 61%, distance 39%, pool 36% | deck or balcony 60%, pool 60%, design 43% |
| Kihei, HI | view 50%, distance 50%, deck or balcony 36% | deck or balcony 86%, pool 75%, waterfront 64% |
| Market | Amenity: short / long | Setting or nearby place: short / long | View: short / long |
|---|---|---|---|
| Gatlinburg, TN | 91% / 100% | 30% / 24% | 49% / 55% |
| Destin, FL | 71% / 100% | 59% / 94% | 65% / 62% |
| Chicago, IL | 30% / 81% | 82% / 33% | 9% / 11% |
| Joshua Tree, CA | 63% / 100% | 77% / 83% | 20% / 36% |
| Lake of the Ozarks, MO | 86% / 100% | 69% / 92% | 42% / 28% |
| Breckenridge, CO | 54% / 100% | 91% / 74% | 34% / 32% |
| Athens, GA | 17% / 81% | 83% / 56% | 8% / 6% |
| Bar Harbor, ME | 44% / 100% | 83% / 72% | 14% / 25% |
| Lancaster County, PA | 47% / 97% | 75% / 61% | 11% / 28% |
| Kissimmee, FL | 74% / 100% | 77% / 64% | 0% / 8% |
| Nashville, TN | 61% / 100% | 75% / 57% | 17% / 33% |
| Kihei, HI | 58% / 100% | 64% / 81% | 50% / 47% |
Each market has 30 to 36 cards per version, so a gap of 15 points or so between two markets can be chance.
How often the line goes beyond your title
86% of short lines and 97% of long lines named at least one amenity, place, building detail or view the host's title didn't. Counted detail by detail, only 43% of what the short lines named and 34% of what the long lines named also appeared in the title. If your title is your only pitch, most of what US guests now read about your home comes from somewhere else.
Where the line's words come from
For the 102 listings whose pages we read (183 of their lines, 696 details), we looked for each detail in the host's title, description, amenities list and photo captions.
| Where the detail appeared | Short lines | Long lines | All |
|---|---|---|---|
| The host's description | 95% | 88% | 91% |
| The amenities list | 42% | 62% | 54% |
| The host's title | 41% | 31% | 35% |
| Photo captions | 30% | 30% | 30% |
| Airbnb's own listing highlights | 73% | 63% | 67% |
| Not found in anything the host wrote or ticked | 2 of 275 | 3 of 421 | 5 of 696 |
A detail can appear in more than one place, so the columns add up to more than 100%. 58% of details were in the description but not the title, and the long line leaned harder on the amenities list (62% of its details, against 42% for the short line). Airbnb doesn't say which fields its AI reads, so this shows where the same facts sit in a listing, not proof of which field the AI used.

102 listing pages read on October 23, 2026. Highlighted: the host's description.
Mismatches: what we couldn't find in the listing
We checked each detail the rules couldn't find against the full listing text. Of the 5:
- 3 named a nearby place the host never mentioned, such as "near the Belmont L" for a Chicago studio and "next to Kamaole Beach" for a Kihei apartment. Airbnb's own listing highlights named the same spots, so they most likely came from the home's location, not from the host.
- 1 was a misread by our rules: the description called the sauna "cedarwood," which our word list missed.
- 1 was a word choice: an apartment three blocks from the Magnificent Mile became a "River North hideaway."
None of the lines we checked named an amenity the listing marked as unavailable. We also checked 40 random details the rules had found in a description against the text itself: 39 were there, and the other was in the amenities list. That doesn't make every line right, though: matching shows the same thing appears in the listing, not that it's still true, and we didn't compare lines with photos.
Sales words and symbols: titles vs Airbnb's lines
- Sales words: 14% of the 713 hosts' titles used words like luxury, epic or amazing, against 1% of short lines and 10% of long lines. In 13 of the 42 long lines with a sales word, the host's title used the same word; "luxury" appeared in 23 of the 42.
- Symbols: 35% of titles used pipes, bullets, asterisks, tildes or exclamation marks. No Airbnb line did.
- Title length: hosts use the full allowance. 42% of titles ran 48 characters or more, and the median was 46.
Outside the US: your title, cut to fit
On airbnb.co.uk, airbnb.ca, airbnb.com.au and airbnb.ie, all 160 cards showed the host's title as the line under the home type, and none of the 9 pages carried Airbnb's note about AI descriptions. 36 of those listings were also in our US sample, so the same homes showed an AI line on airbnb.com and the host's title elsewhere. The guest's site decides: a London search on airbnb.com showed AI lines on all 17 homes.

Title length on 160 non-US cards, October 23, 2026. Zones from our airbnb.co.uk cut measurements.
How much of the title fits comes from our earlier measurements on airbnb.co.uk at four screen widths (Airbnb title ideas): cards in the main results grid showed 32 to 47 characters, a median of 39 on a 1,440-pixel desktop with the map open and 44 on an iPhone-sized phone. No title of 32 characters or fewer was cut.
| Title length | Cards on the four non-US sites | What guests saw, by our measurements |
|---|---|---|
| 32 characters or fewer | 13 (8%) | The whole title, on every screen |
| 33 to 47 characters | 75 (47%) | Cut on some screens, depending on width |
| 48 characters or more | 72 (45%) | Cut on every screen we measured |
By those medians, 82% of the 160 titles were longer than what a typical desktop card showed and 64% longer than a typical phone card.
The line is built from what you write and tick, so the fixes are in your listing, not your title.
What hosts should do
- Write your description for the line. 91% of the details Airbnb's lines named were in the host's description. Put the three things you'd want on your card in its first lines, in plain words: "private hot tub," "fire pit," "free parking," "5-minute walk to the lifts."
- Tick every amenity you really have, and only those. The long version names two or three amenities per card. An amenity you never ticked or wrote down is unlikely to appear, and one you ticked by mistake can. Airbnb requires amenities to be accurate, with restrictions disclosed (Airbnb Help).
- Say where you are, with a number. Short lines named a setting or a nearby place 72% of the time. A line like "a 4-minute walk to Main Street" or "on the creek, 10 minutes from the park entrance" in your description gives the AI something specific to repeat.
- Check your card more than once. Search your town on airbnb.com in a private window and read your line, then try a second browser or profile. You may get either version, and a long line can be worded differently for different visitors.
- Fix the source when the line is wrong. A listing's short line didn't change between visits, so don't expect a mistake to fix itself. Correct the description, amenity or caption it most likely came from; Airbnb doesn't say how quickly lines update, so recheck in a week or two and report it to Airbnb if it doesn't change.
- Keep the title for the listing page and other countries' sites. On Airbnb's non-US sites your title is still on the card, so put the reason to book in the first 32 characters, and stay under Airbnb's 50 (Airbnb Resource Center). More in my Airbnb title ideas and description examples.
If you'd rather see your listing the way guests do, next to the listings around yours, that's what our listing audit does.
Limitations
- One day, logged out, website only. Everything comes from October 23, 2026, on airbnb.com and four non-US Airbnb sites, from a US connection, without logging in, without dates and without filters. We didn't check the apps or AI search mode.
- Two visitors. The main sample comes from one short-version visitor and one long-version visitor, each searching all 12 markets. The two saw different homes as well as different lines (only 124 listings overlapped), so differences between versions can partly reflect the homes; the 124 matched listings show the same pattern.
- Search position not controlled. We read the first two pages of a default search. Airbnb's ranking, which differs by visitor, decided which homes made it in; hotels are excluded.
- Markets chosen by us. The 12 markets cover common US vacation-rental types; they are not a random sample of US listings.
- Rules, not judgment. Codes come from word lists, checked line by line on 160 lines. They count whether a line names a pool, not whether the pool exists.
- Listing pages now, lines then. We read listing pages the same day, but a host may have edited the listing after Airbnb wrote the line. The 102 listings lean toward the first five markets in the table and toward homes seen in both versions.
- What we can't see. Airbnb doesn't say which fields its AI reads, how often lines are rewritten or how much of a long line a card displays. The non-US cut zones come from our earlier airbnb.co.uk measurements, not from this sample.
- Some visitors still see titles. In our separate layout measurements the same day, 3 of 26 browser sessions on airbnb.com got an older card layout that showed hosts' titles, with nearly square photos and no AI note (how we measure Airbnb). None of the 837 cards in this study did.
- It will change. Airbnb is testing at least two versions, so treat this as a dated snapshot.
How to cite this study
Miksis, Jonathan. "What Airbnb's AI Writes on Search Cards: A Study of 837 US Cards (October 2026)." AI Hustle Guy, October 23, 2026. https://www.aihustleguy.com/blog/airbnb-ai-search-card-study
- Data: download the CSV (1,014 rows: 837 US cards, 160 non-US cards and the 17-card London check on airbnb.com). Columns: market, site, line version, property type, host title, Airbnb's line, codes, line length, rating band and collection date. Codes follow the table above (amenity, setting, nearby_place, style, view, sales_word, number, guest_fit, beyond_title); non-US rows carry host_title_shown and a title-length band. No listing IDs, links, host names or photos.
- License: CC BY 4.0. Free to use, chart or quote, with credit to AI Hustle Guy and a link to this page.
- Next update: January 2027, with the same markets and method, to see whether the long version replaced the short one.
FAQ
What does Airbnb's AI write on search cards?
A short summary of the home, one or two lines, that replaces the host's title for most visitors on Airbnb's US site. In our October 23, 2026 study of 837 airbnb.com cards, short lines most often named the setting or a nearby place, such as lakefront, while long lines listed two or three amenities, such as a pool, a hot tub and free parking.
Why do two guests see different AI lines for the same Airbnb?
Airbnb was showing two versions on October 23, 2026: a short line (median 38 characters) and a long one (median 74). In 15 fresh logged-out visits, 7 got the short version and 8 the long one, and visitors we followed kept their version. Different visitors sometimes saw the long line worded differently, too.
Where does Airbnb's AI get the words for the search card line?
Airbnb says only that the lines are based on details hosts provide. When we matched 696 details from 183 lines to the hosts' listing pages, 91% appeared in the host's description, 54% in the amenities list and 35% in the title. Only 5 weren't in anything the host wrote or ticked, mostly nearby places.
Does Airbnb still show my title in search results?
Not on airbnb.com: none of 837 US cards we read on October 23, 2026 showed the host's title. On Airbnb's UK, Canadian, Australian and Irish sites, all 160 cards we read showed it, and 45% of those titles were long enough to be cut on every screen size we've measured.
Can I use this Airbnb data in my own article?
Yes. The study and the CSV are licensed CC BY 4.0: you can reuse, chart or quote them with credit to AI Hustle Guy and a link to this page. The CSV has every card's market, site, version, home type, host title, Airbnb line, codes, length and rating band, but no listing IDs, links, host names or photos.

Written by Jon Miksis
Founder of AI Hustle Guy, Make the Leap, and Europe Flight Club, and the travel blogger behind Global Viewpoint (9+ years, 73 countries). He brings 9+ years of coaching and mentorship experience and has attended 23 transformation retreats and facilitated 6 across four continents. Every AI tool and prompt on this site gets tested on his own businesses first.
Get my free AI prompt pack
Join 300+ founders, creators, and professionals. I’ll send you my go-to AI tools and copy-paste prompts that save hours every week. No spam.


