Interaction DesignTYPENORMLabs6 minSeptember 18, 2026

Airbnb UX Teardown: A Header That Forgets the Trip

A UX teardown of Airbnb's signed-out search flow: a filter button that counts homes before you commit, one all-fees price carried from histogram to Reserve, hotels at the top of a Homes search, and a header that forgets the trip on the listing page.

Airbnb's search starts as one rounded bar with three fields: Where, When, Who, and a pink Search button at the right end. Each field opens its own panel directly under itself. Destination suggestions appear under Where, a two-month calendar under When, four counters under Who.

This capture follows one sample trip through that bar and past it: San Francisco, 2 to 4 October 2026, two adults. It sets three filters, opens one listing, reads its amenities, and then visits the Experiences tab. That is eleven states, recorded on 7 September 2026, signed out, in a 1440×900 Chrome window. The interface language is US English, but prices are in Brazilian reais and the recommendations are for southern Brazil, because the session ran from there. Nothing was reserved.

1 / 11

Choose a destination

The destination picker keeps matching places beside the active search field, letting the guest resolve the city before choosing dates.

View full flow (11 screens)

Where, When, Who

Typing "San Francisco" into Where returns five suggestions. The first is San Francisco, California. The others are Union Square, São Francisco do Sul in Santa Catarina, the San Francisco Bay Area, and San Francisco in Córdoba, Argentina. The Brazilian town ranks third, above the Bay Area, which fits the session's location and would look odd to a guest in California.

Once a suggestion is picked, the field shows it truncated ("San Francisco, California, Unite…") and focus moves to When. The calendar draws 2 to 4 October as two black circles joined by a grey band, and the field above updates to Oct 2 - 4 while the panel is still open. Under the calendar a row of pills offers exact dates or ±1, 2, 3, 7 or 14 days. A Flexible tab sits beside Dates for a guest who has no dates yet.

The Who panel defines each age band in grey under its label: Adults are 13 or above, Children 2 to 12, Infants under 2. Pets gets a link instead of an age: Bringing a service animal? The field reads 2 guests as soon as the second adult is added.

Hotels at the top of a Homes search

The results page shrinks the bar into a summary, Homes in San Francisco · Oct 2 – 4 · 2 guests, and heads the list with Over 1,000 homes in San Francisco. The four cards that follow are Hotel Julian, Hotel Nikko, Hotel Zelos San Francisco and Harbor Court Hotel. Each carries a Featured hotel badge, and the map pins beside them use a small hotel icon.

Compare homes on the map
Compare homes on the mapSearch results pair listing photos and total stay prices with a persistent map. Destination, dates, and guest count stay visible above the results.

The Homes tab was selected before the search, and the summary still says Homes.

The first row of results is hotels anyway.

The quick-filter chips under the bar begin with Hotel, followed by Free parking, Self check-in, 1+ bathrooms and Kitchen. The filter dialog opens on a Recommended for you row where Hotel comes first as well.

A guest who picked Homes has already said which of those they want. This capture cannot show whether the hotels are paid placement or ranking. Either way, "Featured" is the only word on the card that marks them as different from what was asked for, and nothing above the cards explains it.

The count on the button

The filter dialog is long and scrolls inside a fixed frame, with Clear all and a black results button pinned to its footer. The Type of place control is a four-way switch: Any, Home, Room, Hotel. Under it the price range is a histogram with the caption Trip price, includes all fees. The bars inside the selected range turn pink and the bars outside it turn grey.

With Home chosen and the maximum dragged to R$3,000, the footer button reads Show 666 homes. Scrolling down to Amenities and switching on Wifi changes it to Show 657 homes before anything is applied.

The count shows what is left: Wifi removes only nine places, and the price cap does most of the narrowing. The guest learns that without closing the dialog. The results page then opens on 657 homes in San Francisco, the number the button promised.

The applied filters come back as chips: Filters with a badge of 3, then Wifi, Up to R$3,000 and Entire home. The dialog called the choice Home; the chip calls it Entire home. Same setting, two names, a few seconds apart.

See the filtered results
See the filtered resultsApplied filters become editable chips above the listing grid. The result count falls to 657 homes, and the map updates alongside the matching places.

The map also zooms out after the filters apply. Before, it showed downtown at a 500 m scale. Afterwards it shows the peninsula at 2 km, with price pins running south past Daly City toward Millbrae and the airport, and the third card in the grid is a Guesthouse in Pacifica. The heading still says San Francisco. It doesn't say the search area grew.

R$2,534 on three screens

The price is the one thing that stays put. The histogram labels its axis as a trip price including all fees. The first filtered card shows R$2,780 struck through and R$2,534 total, with 6x R$422 in a small grey tag underneath. The listing page repeats the same figures in its reservation card, adds Or 6x R$422 without fees, and puts a banner above it: Save R$246 on these select nights. R$2,780 minus R$2,534 is R$246, so the banner and the struck-through price agree.

The filter sets the unit, and the card and the reservation card quote the same total in it: a two-night trip for two adults, fees included. Under the Reserve button a single line reads You won't be charged yet.

The 6x figures are an instalment price. They appear on the result cards in view and on the listing in this capture, and they are probably shown because the session was in Brazil, where paying in instalments is common. Read that as an inference from the capture; a US session may never show them.

The header on the listing page

The listing opens with a five-photo grid, the title New suite in Victorian home, and a Guest favorite panel with a 4.98 rating from 189 reviews. Scrolling down, the reservation card on the right holds CHECK-IN 10/2/2026, CHECKOUT 10/4/2026 and 2 guests, the same trip, now written as numeric US dates rather than Oct 2 – 4.

The header search bar at the top of the same page reads Anywhere · Anytime · Add guests.

On the results page that bar was the summary of the trip. On the listing it resets to its empty state while the reservation card further down the same page still knows the dates and the party. A guest who wants to go back to the search and change one thing has two places that disagree about what the search was. Nothing in the capture shows the bar being used from the listing, so what happens when it is clicked is outside this evidence.

Experiences starts over

The last screen comes from a different branch. The capture went back to the homepage and selected Experiences, the tab next to Homes.

Browse experiences
Browse experiencesThe Experiences tab reuses the search shell while replacing stay cards with activities, start times, and per-guest prices. Local recommendations reflect the session’s Brazil context.

The tab keeps the same Where, When, Who bar, but all three fields are empty: Search by city or landmark, Add dates, Add guests. The rows below are Tomorrow in São José and Experiences this weekend, with sailing lessons, a dune-surfing session and trail hikes around Florianópolis, the session's own region, not San Francisco.

The price unit changes too.

Homes were priced per trip with all fees included. Experiences read From R$350 / guest, per person and with a From. A guest planning October in San Francisco has to type the city, the dates and the party a second time, then multiply by two to compare an activity with the stay they just looked at. Because the capture returned to the homepage first, this doesn't prove that the tab drops the trip when opened from the results page; it shows that the homepage's Experiences tab doesn't start from the stay search that was just run.

What this means for your product

Put the live result count on the button that applies the filters. Airbnb's guest sees what Wifi costs, nine homes, while the choice can still be undone.

If your product asks users to build a query one field at a time, list every screen where that query should still be visible, and check that it is. Airbnb's search bar holds the trip on the results page and drops it on the listing.

Take it further

The lens behind this teardown is the UX Clarity framework, the same one we apply in a Full UX Audit. Here it asks whether a guest can tell, on every screen, which trip they are looking at and what it will cost. For how that scoring turns into prioritized fixes, read what a real UX audit looks like.

Scope and limits: eleven states from one supervised session on 7 September 2026, signed out, Chrome at 1440×900, US English interface with BRL prices and Brazil-local recommendations. Sample trip only: San Francisco, 2 to 4 October 2026, two adults. No account, no login, no reservation submitted. Result counts, prices, the listing and the featured hotels belong to that session and will differ by date, location and currency. The amenities overlay was captured at its first page only; the homepage landing was captured separately and isn't analyzed here.

Sources: NN/g — Visibility of System Status · NN/g — Recognition vs Recall · NN/g — Consistency and Standards.

Ready to find where your product forgets what your users already told it? Apply for a Full UX Audit →

Free UX Snapshot for 50 Product Teams

Apply now and get a complimentary UX Snapshot — our rapid clarity audit delivered in 48 hours. Limited to the first 50 products.

Apply for Free UX Snapshot

Related

Interaction Design

TechCrunch UX Teardown: The Reading Mode That Drops the Date

A UX teardown of TechCrunch's read flow: one switch turns the feed into a dense scan queue and carries the preference into the next section — then removes the timestamp and the content type the returning reader came for.

TYPENORMLabs · 5 min · September 10, 2026

Web
Media & Entertainment
Interaction Design

UX Writing

Sentence Case vs Title Case in UI: What the Major Design Systems Actually Do

Sentence case has become the default across Microsoft, GOV.UK and Material 3, while Apple still capitalises its buttons. A look at what each convention specifies, why title case is expensive to enforce, and how to write the rule so a team stops re-deciding it.

TYPENORMLabs · 7 min · September 10, 2026

UX Writing
Web

Information Architecture

Reuters UX Teardown: A Search That Sorts Like a Wire

A UX teardown of Reuters' signed-out read flow: dependable section fronts, an article page that gives the reader its figures early, and a search that ranks 14,934 matches by the clock, then leaves a Business story at the top of a Technology filter.

TYPENORMLabs · 5 min · September 11, 2026

Web
Media & Entertainment
Information Architecture