Updated October 1, 2026

Corrections

Every time we found that something we had published was wrong, or said more than we could back, we changed it. This is the list, newest first: what we said, what was wrong, what we changed and when. It has 61 entries, from July 2026.

Found something wrong? Email hi@underbook.co. How the product works, and what we measured about its accuracy, is on How Underbook works.

What this log covers

Figures and claims we published: on this site and in our price study, in the extension's banner, in our store listings, in the emails we send to people who set a price watch, and in press emails we sent.
Not listed: mistakes we caught before anything went out, and figures we refreshed when new price snapshots arrived, unless the earlier figure was wrong.
Times are UTC. "About" means our records give only the hour. Extension changes reach each browser when it updates, which can take a few days after the store publishes a version.

Jump to: October · September · August · July

Fixed, waiting on store review

Corrected in our next release, not yet live in the stores.

Store listing

"See which site is cheapest" in our store summary

What we said: The extension's one-line summary in the stores says "See which site is cheapest."
What was wrong: Underbook shows the lowest price among the sites Google Hotels lists, not every site, and "cheapest" is a promise we cannot check.
What we changed: Version 0.5.30 says "See if room names match." instead, and our package check now refuses "cheapest". It is uploaded and waiting on store review.

October 1, 2026

Study pageWebsiteabout 23:10 UTC

Some study prices were listed for travelers in other countries

What we said: The price study's two October 1 runs (25 US cities and 25 international cities) count 578 hotels with a listing 5% or more below the going rate, and the study page and the hotel pages give each one's cheapest site and price as prices for a traveler in the US.
What was wrong: Every search asked Google Hotels for a US traveler, but 134 of 394 answers in the US-city run and 145 of 427 in the international run (34%) came back priced for a visitor in another country, as the booking sites' links show. For 179 of the 578 hotels (31%) the cheapest listing was one of those, and for 156 no listing priced for a US traveler was left. Such a price can differ for a US traveler: of 6 we could check, 2 (Agoda) were higher, by 5.6% and 11.7%. Earlier runs did not keep the links, so they cannot be checked; in a separate test of 300 hotels, 31% of answers came back this way.
What we changed: No total changed. The study page carries a dated caveat with these numbers (read it); the hotel pages are unchanged for now. From the next run on, each run records which country each answer was priced for, and the week-later repeat planned for October 2 reports its figures both for all answers and for answers priced for a US traveler.

Websiteabout 20:10 UTC

The Hoi An pages, relabeled

What we said: A city page and 8 hotel pages presented their hotels as hotels in Hoi An, Vietnam.
What was wrong: Our search for "hotels in Hoi An" had returned hotels in and around Myrtle Beach, USA (see the next entry).
What we changed: The city page is now titled "Hotels returned for a Hoi An search (Myrtle Beach area, USA)", each hotel page names the hotel's own town, and the page is listed with the US cities and kept out of search engines. Page addresses are unchanged.

Study pageWebsitePressabout 19:20 UTC

Hoi An and Naples: two searches returned hotels in another place

What we said: The price study counted 349 cities, among them Hoi An (Vietnam) and Naples (meant as Naples, Italy), and put both in its international figures: in those two runs, 312 of 432 international hotels (72.2%) and 293 of 408 (71.8%) had a listing 5% or more below the going rate.
What was wrong: The search "hotels in Hoi An" returned 15 US East Coast beach hotels and none in Hoi An. The search "hotels in Naples" returned 14 hotels in Naples, Florida.
What we changed: The study now counts 348 cities. Naples is labeled Naples, Florida (region US) everywhere. The two runs' international figures leave those hotels out: 300 of 417 (71.9%) and 286 of 394 (72.6%), with a dated correction note in each section of the study page. Figures computed by region moved on 9 pages (four comparison pages and five "is it legit" pages). The study's totals did not change: 6,254 hotel prices, 4,512 (72%) with a site 5% or more below the going rate. A press email we sent that afternoon said "349 cities"; every other figure in it still holds. Our study script now sends the country with every city outside the US and checks where the hotels are after each run.

ExtensionWeb checkerabout 19:05 UTC

Stricter "same room name" rules, after a test we failed

What we said: Results from the extension and the web checker said "same room name" when a cheaper listing's room name matched a room sold at the going rate.
What was wrong: In our 300-hotel test, written down before the first search, an AI reader that saw only the two room names judged 42 of 50 matches to be the same room (84%), under the 95% we had set as the bar. The ones it did not judge the same were accessible rooms matched with standard rooms or with a different kind of accessible room, and room words our rules did not know (Western against Japanese style, "Comfort" against "Grand", a "Design" room against a duplex).
What we changed: In the price service, accessibility, grade words, room style and layout must now agree. On 100 fresh pairs the new rules call the same room, the reader agreed with 92 (92%): better, still under the bar, so results keep telling you to check the room. The study's published room figures are unchanged and labeled as the October 1 check.

Websiteabout 18:40 UTC

City pages that disagreed with their own data files

What we said: Each city page says at how many of its hotels a site was 5% or more below the going rate.
What was wrong: On 213 of 349 city pages that count differed from the city's rows in the study's published CSV. The page counted all-in prices over the hotels it lists; the CSV counts before-tax prices over every priced hotel. Neither said so, so one city had two different counts.
What we changed: Every city page now states its CSV's own count (0 of 349 differ), and a check script compares every city page with its CSV.

Study pageWebsiteabout 17:20 UTC

Our original room-name figure, restated beside a stricter check

What we said: "515 of 5,830 hotels (8.8%)" had a cheaper listing whose room name matched the room at the going rate, quoted on 9 pages; one of them called it "rare". (Before October 1, with fewer snapshots, the same check was quoted as 433 of 4,996, 8.7%.)
What was wrong: That check looked only at the cheapest listing's room and counted loose matches, for example "Standard Room" with "Triple Room". 140 of the 515 do not hold on a stricter check.
What we changed: The 9 pages now add "on our original room check (952, or 16.3%, on the stricter October 1 check)", and the study page gained a box explaining both checks. "Rare" became "uncommon". The original figure stays, labeled, so nothing changed silently.

Website16:45 UTC

KAYAK, momondo and HotelsCombined

What we said: /is-momondo-legit said KAYAK and momondo show "overlapping results with slightly different ordering", with no figure.
What was wrong: Our own price data showed more than overlap: their prices were the same.
What we changed: The page, and /is-kayak-legit, now say that over 14 snapshots KAYAK.com and momondo.com listed the same before-tax price at all but one of 938 shared hotels, and KAYAK.com and Hotelscombined.com at all 1,045, as the signed-out Google Hotels price, not a room check.

Website09:32 UTC

Orbitz and Travelocity against Expedia

What we said: /is-orbitz-legit and /is-travelocity-legit said Orbitz and Travelocity are "rarely" cheaper than Expedia, and that "our study reports no Orbitz-versus-Expedia figure".
What was wrong: Our data had the figure: at every hotel where both were listed, Orbitz's and Travelocity's prices were identical to Expedia's.
What we changed: Both pages now say that over 14 snapshots Orbitz.com and Expedia.com listed the same before-tax price at all 3,317 shared hotels, and Travelocity.com at all 3,372, as the signed-out Google Hotels price, not a room or member-price check.

September 30, 2026

Websiteabout 20:00 UTC

Two comparison pages reached the wrong conclusion

What we said: /expedia-vs-booking-com and /hotels-com-vs-booking-com said the two sites were "neither reliably cheaper" and that one being cheaper "is mostly a myth", from the September 25 snapshot alone.
What was wrong: More data said otherwise. Over 12 snapshots, Expedia was 2% or more cheaper than Booking.com at 982 of 3,187 hotels (31%) and Booking.com cheaper at 478 (15%); Hotels.com at 1,172 of 3,565 (33%) and Booking.com at 507 (14%).
What we changed: Both pages now say the prices are often close and, when they differ, the Expedia Group site was cheaper about twice as often; neither is always cheaper.

Extensionabout 19:50 UTC

"Rebook and save"

What we said: When no site was lower, the extension's banner offered a price watch with "Booking now? We'll email you if the going rate is at least 5% lower on two daily checks in a row before your stay, so you can rebook and save."
What was wrong: It promised a saving we cannot promise.
What we changed: The offer now says what we send: "Emails about this stay only, and at most one drop alert. No account. Unsubscribe in one click." Nothing in it promises a saving. Live in extension 0.5.28 on Chrome at 20:28 UTC.

Websiteabout 18:30 UTC

Two city pages lost hotels we had told reporters about

What we said: We had told reporters that our Los Angeles page listed 16 hotels and our Raleigh page 15.
What was wrong: A rebuild that afternoon moved hotels found in two runs to a nearby city's page, so the pages showed 15 and 13.
What we changed: Restored within the hour to 16 and 15. A rule in the code now keeps every hotel on the city page that first published it.

WebsiteStore listing08:38 UTC

What the extension sends from pages that are not hotel pages

What we said: The privacy policy and our Chrome and Edge store privacy answers said that on a page that is not a hotel page, the only request the extension makes is the once-a-day anonymous usage ping.
What was wrong: Clicking the Underbook toolbar button on any page also sends that page's title or main heading for a price check, as a possible hotel name.
What we changed: All of them now say so. The privacy policy changed that morning; the Chrome answers went live after review that evening.

Website05:26 UTC

Hotwire: "not tested"

What we said: /is-hotwire-legit said "We have not tested it on Hotwire's own pages."
What was wrong: Out of date: extension 0.5.26, live that day, runs on Hotwire's named-hotel pages.
What we changed: The page now says it runs there, and that a Hot Rate hides the hotel's name until you pay, so on a Hot Rate it has nothing to check.

September 29, 2026

Study pageabout 18:50 UTC

"The discount percentages are close"

What we said: The study page's notes said the discounts in the CSVs (before tax) and on the hotel pages (all-in) "are close".
What was wrong: Not always: comparing all 2,448 hotel pages of that day with their CSV rows, 87 (3.6%) fell on different sides of the 5% line and 200 (8.2%) differed by 5 points or more.
What we changed: All five notes now say the percentages usually agree within a few points but not always, and that a hotel near the 5% line can land on either side.

Website17:30 UTC

Traveluro page totals

What we said: /is-traveluro-legit said "2,116 hotel prices" and "At 1,638 of them some site was 5% or more below".
What was wrong: 8 of the 2,116 rows had no price, so 2,108 were priced, and the five snapshots add up to 1,631.
What we changed: The page says 2,108 and 1,631 in its title, description and text. Every other figure on it was checked again and held.

Study pageabout 00:20 UTC

Two prices for one hotel, unexplained

What we said: The study's CSVs and our hotel pages gave different dollar figures for the same hotel (ARIA: a $288 going rate in the CSV, $326 on its page), and neither said why.
What was wrong: The CSVs hold before-tax prices and count the listings the going rate was computed from; the hotel pages show all-in prices and every site found.
What we changed: A note beside each CSV link on the study page now says so.

September 28, 2026

EmailExtension20:46 UTC

A price watch on a booking site instead of a hotel

What we said: The first price watch a traveler set was accepted for "Agoda Official Site": the extension had read a booking site's homepage title as a hotel name.
What was wrong: It is not a hotel. Our bug.
What we changed: Booking-site names are now refused as hotel names, in the price service and in the extension. We emailed the traveler to say what happened and how to watch the right hotel, and removed the mistaken watch.

Store listing18:14 UTC

Our Chrome Web Store description and privacy answers

What we said: The store description said there was "no Underbook server collecting your activity". The privacy answers said the extension stores "two small values" and that its page scripts send nothing.
What was wrong: Every check goes to our price service. The extension stores more local values than two, all listed in our privacy policy, and its scripts send an anonymous once-a-day ping and banner-event totals.
What we changed: The description and the privacy answers were rewritten to match the privacy policy, and "User activity" was ticked. Live after review on September 29.

ExtensionEmail16:29 UTC

"You've got the best rate", about a price we had not seen

What we said: When no listing was lower, the banner said "You've got the best rate", even when the booking site you were on was not among the listings. The price-drop email's summary line said the same.
What was wrong: The extension does not read the price on your page, so it could not know.
What we changed: The line now appears only when your site is in the listings. Otherwise the banner says "If this page's price is about the same or lower, you've got the best rate." The email's line now ends at the lowest price listed. Extension 0.5.24, live at 17:41 UTC.

ExtensionEmail15:30 UTC

"Confirmed room" on the no-lower-price result

What we said: "No site with a confirmed room lists below $X."
What was wrong: The code only checks that a site names a room. Nothing confirms a room.
What we changed: "No site that names a room lists below $X." In the banner (0.5.24) and in the price-drop email.

Extension10:27 UTC

Price-watch promises without the rule

What we said: Six lines in the extension promised an email "if it drops".
What was wrong: We email only when the going rate is at least 5% lower on two daily checks in a row.
What we changed: The lines now state that rule. Extension 0.5.21, live the same day.

WebsiteStudy page07:27 and 08:27 UTC

"Is the same room", "can't confirm" and "really the lowest"

What we said: The homepage said Underbook tells you "whether that cheaper listing is the same room", and the study page's call to action said the same. 17 sentences on 16 pages said it speaks up "when it can't confirm two rooms match". Five lines said it makes sure the price you pick "is really the lowest one", and two said it "makes sure the number is the real one". The homepage's structured data said "taxes included" with no condition.
What was wrong: The extension compares room names and never confirms a room, and "can't confirm" implies that it sometimes can. It sees only what Google Hotels lists, signed out. Taxes are included only when the sites report them.
What we changed: The pages now say whether the cheaper listing names the same room, the other phrasings are gone, and the structured data says "taxes included when the sites report them".

Web checkerEmailExtension05:26 and 06:27 UTC

How the price watch works, stated as it is

What we said: The web checker said "We recheck it daily and email you if it drops 5% or more". The watch emails said "We'll email you the moment it drops 5% or more". The extension called the watched price "this page's current rate".
What was wrong: We check once a day and email only when the going rate is at least 5% lower on two daily checks in a row. The watched price is the going rate, not your page's rate.
What we changed: The checker, the three watch emails and the extension now state the rule and say "going rate". No traveler had set a watch yet when the emails were changed.

September 27, 2026

Website22:28 UTC

Rate parity, overstated

What we said: /booking-com-vs-hotel-direct said that under rate-parity clauses the public rate "is contractually kept the same across channels". /expedia-vs-booking-com said the base price "usually matches".
What was wrong: Unmeasured and overstated. At 262 hotels listing both, Booking.com was more than 1% above the hotel's own site at 106. /expedia-vs-booking-com itself showed the two within $2 at only 54% of 245 hotels.
What we changed: The pages say the clauses discourage one channel from undercutting another, and give the measured figures.

Website21:29 UTC

The Genius page

What we said: /genius-discount-worth-it said Underbook "compares the Genius rate against the same room on other sites, all-in, fees included"; that Genius Level 1 comes after a first stay; that discounts are "10-15% off"; that rate parity keeps prices "usually close"; and that we choose "not to take a cut of your booking".
What was wrong: The extension compares Google Hotels' signed-out listings, so it never sees a Genius price. Booking.com's own Genius page says Level 1 starts when you create a free account and discounts run from 10% to 20%. Our study found a site 5% or more below the going rate at 331 of 424 hotels. And our booking links can earn a commission on Booking.com.
What we changed: All five corrected. The page says the extension does not see your Genius price and suggests holding your Genius total against the lowest price it finds.

Website20:27 UTC

"Compares the all-in price"

What we said: 48 self-descriptions on 40 pages said Underbook "compares the all-in price". llms.txt, our summary for AI assistants, still quoted the retired label "same room confirmed".
What was wrong: At about 3 in 10 hotels no site reports a before-tax price and the extension shows the listed rate instead. "Same room confirmed" had been retired the day before.
What we changed: The pages say it "compares the price for your hotel and dates, all-in where the sites report taxes and fees". llms.txt no longer quotes the old label.

Web checker19:27 UTC

The web checker's "all in"

What we said: The checker said "The lowest listed price is about $X a night, all in" and headed its list "All-in price per night, by site", even when no site reported a before-tax price. Its description said it "shows the all-in price on each".
What was wrong: When no site reports a before-tax price, every figure is just the listed rate.
What we changed: In that case the checker drops "all in", and the list reads "Listed price per night, by site (fees may be missing)". The descriptions say all-in when the sites report taxes and fees.

Website16:43 UTC

How to start a price watch

What we said: Five pages said to start a watch by pasting a link ("Copy the booking link for the hotel, dates, and room"). /best-hotel-price-tracker said the watch follows "the all-in total, over time, on the same room". /genius-discount-worth-it said it alerts "if the rate drops below the Genius price you booked".
What was wrong: No form on the site took a link. The watch followed the going rate for a hotel and dates, not a room or an all-in total, and it never saw the price you booked.
What we changed: The pages say to type the hotel and dates into the price checker and state what the watch follows and when it emails.

Website16:27 UTC

"Taxes and mandatory fees included"

What we said: 34 places on 31 pages said the extension shows "the all-in price side by side", 8 of them "taxes and mandatory fees included" with no condition. /resort-fees-explained added that "a fat resort fee can't disguise itself as the best deal".
What was wrong: When no site reports a before-tax price, the banner shows the listed rate instead. That was the case at 130 of 424 hotels in our first snapshot and 138 of 434 in the second.
What we changed: The pages say all-in "when the booking sites report them (at about 7 in 10 hotels in our price data), otherwise the listed rate, labeled as such".

Website15:27 UTC

"Only when it's the cheaper one"

What we said: 21 pages said we show a site's rate "when it's the cheaper one, and only then".
What was wrong: The extension also names a site within 2% of the lowest price when that site lists free cancellation or names the room.
What we changed: The pages say "We lead with the lowest price we found, whoever lists it", and when we also name another site.

Website10:28 to 14:26 UTC

"Often cheaper than booking direct" on seven "is it legit" pages

What we said: Our pages on Hotels.com, Expedia, Priceline, Agoda and Trip.com said they are "often genuinely" or "often meaningfully cheaper" than booking direct, and "The savings are real". /is-prestigia-legit said Prestigia "often undercuts the big sites". /is-hotwire-legit said "the savings are real". The structured data on /hotel-price-comparison still said prices are "sometimes lower for the identical room".
What was wrong: Against the hotel's own site, more than 1% cheaper at only 44 of 225 hotels (Hotels.com), 29 of 178 (Expedia), 41 of 242 (Priceline), 60 of 216 (Agoda) and 39 of 171 (Trip.com). Prestigia was the cheapest listing at 0 of the 45 hotels where it appeared. Hotwire is not in our price data at all.
What we changed: Each page says "sometimes cheaper, though usually not" and gives the measured split, or says the discount is unmeasured (Hotwire, Priceline Express Deals, Trip.com app prices). The structured data matches the visible answer.

Website09:27 UTC

Algotels

What we said: /is-algotels-legit said "We surface their rate when it's the cheaper one" and that Algotels is "frequently the cheapest listed price on metasearch".
What was wrong: Algotels is absent from all 497 sellers in our price data, and the extension reads Google Hotels' listings, so it never shows Algotels' rate.
What we changed: The page says Underbook does not show Algotels' rate. On September 28 its verdict became "A real discounter whose prices we could not measure", and its owner was corrected: it had said Algotels is "operated by Holisto"; it is a Holisto brand now under trivago DEALS.

Website07:27 and 08:27 UTC

"The same room", "your exact room" and the watch

What we said: About 30 pages said Underbook checks "the same room" and shows whether "the cheaper number is actually the same booking". About 20 lines said it "compares the all-in price for your exact room". The price-watch pages said the watch follows "that exact room" and emails "the moment the rate drops".
What was wrong: The extension checks the hotel and dates and compares room names; it never confirms a room. The watch stores no room, runs once a day and alerts only at 5% below on two daily checks in a row.
What we changed: The pages say "for your hotel and dates" and "whether the cheaper listing names the same room", and describe the watch as it works.

Website06:29 UTC

A closing line on 13 pages, and Hotwire figures

What we said: 13 pages ended with a line saying Underbook "compares the all-in price for your exact room across booking sites, so you can see whether the cheaper rate really is the same booking". /is-hotwire-legit said Hot Rates can be "a room 30% under the hotel's own price" and are "often the lowest around".
What was wrong: The cheapest listing named no room at 67.5% of hotels in our first snapshot, so no one can see that. The Hotwire figures had no source, and Hotwire is not in our data.
What we changed: The closing line says it checks your hotel and dates and whether a cheaper listing names the same room or needs a check. The Hotwire page drops the figures.

Website05:27 UTC

"The hotel price parity myth"

What we said: 25 links on 19 pages called our study write-up "the hotel price parity myth".
What was wrong: That page had been retitled, and its data shows that exact price parity mostly does not hold.
What we changed: The links name the question the page answers: "Is the same room cheaper elsewhere?"

Website03:28 UTC

A Las Vegas "fee gap" that was a tax

What we said: /do-hotel-resort-fees-show-in-price cited "$821 a night versus $931 all-in" four times as a resort-fee gap, "typical of the class".
What was wrong: $821 plus the Las Vegas Strip's 13.38% room tax is $931: the check measured tax, not a resort fee. And one check is not typical of anything.
What we changed: The page keeps the numbers, says the gap is the room tax, and says one check cannot tell whether the resort fee was included.

Website01:26 UTC

"Tier arbitrage" and a Maui example

What we said: /same-room-different-price presented "a real exception: tier arbitrage within a single provider ... in the case we measured", and told a Maui resort example ($318 on eight sites against $188 from a wholesaler) as if it came from our data.
What was wrong: Our study did not measure member or referral rates, and no data we published backs the Maui example.
What we changed: The page says our study did not measure those rates and cites one that did (Hannak et al., 2014). The example is labeled "an illustration, not a hotel from our data".

Website00:30 UTC

"No affiliate relationship"

What we said: 10 pages said "We have no affiliate relationship with" a booking site, or "no affiliate relationships with any booking site".
What was wrong: Our privacy policy says we may earn an affiliate commission, and our booking links can earn one on those sites.
What we changed: The pages say "If you book [the site] through Underbook we may earn a commission, which never changes which rate we show."

September 26, 2026

Website22:30 and 23:32 UTC

"No affiliate pitch", and a cookies claim without its source

What we said: Five guide pages said "No affiliate pitch." The cookies page said clearing cookies is "mostly a myth", resting on unnamed tests by "journalists and consumer groups".
What was wrong: Our booking links can earn a commission. And the page did not name the test behind its claim; we have run no cookie test of our own.
What we changed: The five pages carry the commission line. The cookies page names the study it rests on (Hannak et al., ACM IMC 2014), says it is from 2014, and says we ran no cookie test ourselves. Its search snippet now leads with what we measured.

Website21:29 UTC

"Confirmed rate" and "We earn nothing"

What we said: 34 pages promised "the cheaper confirmed rate for your room" or similar. Seven lines said "We earn nothing from" Booking.com, Agoda, Expedia, Hotels.com, Priceline, Trip.com or "the booking sites".
What was wrong: The extension matches room names; a name match is strong evidence, not proof. And our booking links can earn a commission on those six sites, as the privacy policy says.
What we changed: The pages say Underbook shows the cheaper rate "and whether it names your room", and that we may earn a commission, which never changes which rate we show.

Website20:34 UTC

An old 24-hotel test, retired

What we said: 12 pages cited a 24-hotel test from June and August ("1 of 36 (~3%)", "86% unverifiable", mainstream sites "at parity"). /hotel-price-parity-myth argued that the same room costs the same everywhere. Other pages said "same room is usually the same price", that momondo "shows up all the time" and that Travelocity is "almost always identical".
What was wrong: The "1 of 36" counted a named room, not a matched one, and our September study contradicted the rest: cheaper listings were common, and the ones with matching room names were mostly on Agoda, Booking.com and Priceline. The momondo and Travelocity claims were never measured.
What we changed: 88 replacements on 12 pages, every number taken from our published data. The parity page is now titled "Is the Same Hotel Room Really Cheaper Elsewhere?". momondo: "114 of 434 hotels". Travelocity: marked unmeasured (measured on October 1, see above).

ExtensionWeb checkerWebsite20:19 UTC

"Same room confirmed" became "same room name"

What we said: The extension's banner, the web checker, the homepage and guides said "same room confirmed".
What was wrong: A matching room name is strong evidence, not proof, as our own study says.
What we changed: "Same room name" (or "Room name matches") everywhere. Extension 0.5.18, live at 20:57 UTC.

WebsiteStudy page20:13 UTC

51 same-room overclaims on 16 pages

What we said: Lines such as "the identical room genuinely sells for different totals" on the hotel's own site, Booking and Expedia; Expedia and Booking.com "within a dollar or two on the identical room"; "Chains held rate parity almost perfectly"; "Chain hotels leaked least"; a table of 618 comparisons from before September 23 ("confirmed cheaper on a named room 9%"); and, on the study page, a "true same-room figure".
What was wrong: A listing 5% or more cheaper with a matching room name turned up at only 25 of 434 hotels on September 25. Expedia and Booking.com's lowest prices were within $2 at 54% of hotels. Chains had a cheaper listing 82% of the time. The table came from the check that never compared room names.
What we changed: The pages quote the study's published sentences and figures. The study page says "matched room names" and calls its chain explanation "one possible reason, which this data cannot confirm".

Website20:03 UTC

Algotels "constantly" near the top

What we said: /is-algotels-legit said "You'll see it near the top of Google Hotels price lists constantly."
What was wrong: It rested on a June check of 24 hotels, and Algotels was not among the 497 sellers Google Hotels listed for 434 hotels on September 25.
What we changed: The page attributes the old figures to the June check and says Algotels did not show up in September.

Website19:53 UTC

The privacy policy's "completely idle", and "non-refundable"

What we said: The privacy policy said the extension "stays completely idle" away from hotel pages. /is-free-cancellation-worth-it used "non-refundable" without saying what it meant.
What was wrong: The once-a-day anonymous usage ping can be sent from any page. And our data records only whether Google Hotels showed a free-cancellation label, not a refund policy.
What we changed: The privacy policy says so, and now also describes the click pass-through for booking links, the web checker's daily limit and other details it had left out. The page defines "non-refundable" as a listing without the free-cancellation label.

September 25, 2026

Extension20:28 and 20:56 UTC

The banner's "best rate" and "This page" row

What we said: The banner said "You've got the best rate" whenever no listing was 5% or more under the going rate. On lower-price results, its first row was labeled "This page" but showed the going rate.
What was wrong: In our September 25 data, at hotels where the banner said this, the traveler's booking site (Booking.com, Expedia, Hotels.com or Agoda) listed the stay 5% or more above the lowest listing in 134 of 241 cases. And on lower-price results, the going rate was within 5% of the traveler's site's own price only 73.7% of the time, so about a quarter of "This page" figures, and the savings worked out from them, were off by more than 5%.
What we changed: Extension 0.5.15 says which site lists the stay for less and by how much. 0.5.16 shows the traveler's own site's price (or labels the row "Going rate") and says so when that site already lists the lowest price. Both live that evening.

September 23, 2026

ExtensionWeb checker20:36 UTC

"Checked 12 booking sites"

What we said: The extension said "Checked 12 booking sites" whether a hotel had 9 sites or 22. The web checker's first live result showed a going rate higher than every price in its list.
What was wrong: It counted the 12 cheapest entries, duplicates included, so the number was not the number of sites, and the sites at the going rate could be cut off.
What we changed: The price service returns the number of distinct sites and every site. The checker lists every site; extension 0.5.12, live that evening, shows the true count.

Study pageWebsiteExtension17:10 UTC

Retraction: "same room confirmed" never compared rooms

What we said: Our price study of September 22 reported "a confirmed same-room saving at 10.6%" of hotels and called 45 listings "confirmed same-room savings". The extension's banner said "Save ~$X/night, same room confirmed". We also reported "independents 13.5% vs chains 2.7%" confirmed same-room, and that when Booking.com or Agoda were cheapest "the room was confirmed about half the time".
What was wrong: Our check only saw that the cheaper site named a room. It never compared that room with the room the other sites were pricing, and among the 45 were beds in shared dorms priced against private rooms. The shares quoted were shares of listings that named a room.
What we changed: The price service now compares room names; it went live that day, so every banner changed at once. The study page carries a dated correction box and describes the 45 as listings that named a room; its share card, charts and CSV column, the other pages that repeated the claim, and llms.txt were corrected. On September 25 a second snapshot gave the study's first figure for matching room names: 25 of 434 hotels (5.8%).

Website17:00 UTC

Site copy that described a different product

What we said: The homepage and several guides said the check runs "using your own browser", that it works on "Google Hotels", that it "stays silent when there's no real saving", and that it shows "the room type and cancellation terms side by side".
What was wrong: The check calls our price service. The extension deliberately stays off Google's travel and maps pages. When no site is lower it shows a one-line best-rate note. It did not show cancellation terms.
What we changed: Rewritten on the homepage (including its search and social descriptions), /hotel-price-comparison, /is-trip-com-legit and /is-super-com-legit.

Website16:45 UTC

A homepage that said the extension did not exist

What we said: Every "Add to Chrome" button on the homepage opened a panel saying "Underbook isn't in the Chrome store yet", with a waitlist form, and the homepage's questions said the extension was not live yet. Its "Is my data safe?" answer said "the check runs in your own browser".
What was wrong: The extension was live on the Chrome Web Store. The buttons were left over from a pre-launch test in June. The check sends the hotel name, dates and party size to our price service.
What we changed: All five buttons link to the store; the panel and the waitlist form are gone. Both "Is my data safe?" answers (homepage and /hotel-price-comparison) say what is sent.

September 22, 2026

Email18:38 UTC

"Taxes & fees included" in watch emails

What we said: Price-watch emails printed "taxes & fees included" under the price.
What was wrong: They did so even when no site reported a before-tax price, when the figure is only the listed rate (for example $470 listed at the Bellagio, without its resort fee).
What we changed: In that case the emails say "listed rate", "resort/property fees may not be included, verify at checkout". No traveler had set a watch yet.

September 16, 2026

Website17:34 UTC

What the privacy policy said we send

What we said: The privacy policy said the extension sends "only the hotel name and those dates".
What was wrong: Since versions 0.4.7 and 0.4.9 it had also sent the number of adults, the booking site's market and currency, and its own version number, and the price service kept priced results for 30 minutes and a short list of lookups that found no rates.
What we changed: The privacy policy says exactly that.

September 14, 2026

Extension19:35 to 21:50 UTC

"The hotel may be sold out"

What we said: When no booking site had rates, the banner said "The hotel may be sold out or nearly sold out. Try different dates."
What was wrong: Every time we could check it, it was false: four hotels, two travelers, about thirty banners. A party of six was told this about a hotel that the same search priced for two adults, and the Vienna hotels it was said about had prices when we checked them again.
What we changed: The price service stopped sending the claim on September 14, first for parties of more than two, then for long stays, then for everyone. Travelers now get a plain line that we could not confirm rates. The sentence was removed from the extension itself in 0.5.7, live September 18.

September 13, 2026

ExtensionEmail

A price no room-naming site offered

What we said: When no listing was lower, the price-drop email said "Every site with confirmed rooms is ~$X/nt", and the banner quoted the same going rate.
What was wrong: The figure was the going rate, which could be a price that no site naming a room offered, and the line claimed confirmed rooms even when no site showed any.
What we changed: The line now quotes a price a site actually lists, the lowest among sites that name a room, or says that none showed room details. In the extension from 0.5.6 (live September 13), in the email from September 14.

September 11, 2026

Website

A browser extension called a website

What we said: The homepage title called Underbook a "Hotel Comparison Website", and so did the /hotel-price-comparison page.
What was wrong: Underbook is a browser extension (the web price checker came later, on September 23).
What we changed: Both pages describe it as a hotel price comparison extension.

September 6, 2026

Extension

Prices in one currency on a page in another

What we said: On a booking site showing US dollars, the banner could say "This page ~£101/night", with no note of the currency.
What was wrong: The note meant to name the currency never appeared: its condition compared the currency with itself.
What we changed: Every banner with prices now says "All figures in" and the currency. Extension 0.5.5, live September 6.

September 2, 2026

EmailExtension

"Unsubscribe anytime", without a way to

What we said: The banner's price-watch offer promised "Price-drop alerts only. No spam, unsubscribe anytime".
What was wrong: The emails only offered "Reply to stop the watch": no unsubscribe link and no one-click unsubscribe header.
What we changed: Every watch email now has a one-click unsubscribe link and the standard unsubscribe headers, and the link stops the watch at once.

August 21, 2026

Website

A June savings figure, measured again

What we said: Three pages cited a June measurement: "~6%" of hotels with a confirmed same-room saving, "~89%" unverifiable, "1 in 18".
What was wrong: Measured again in August on the same 24 hotels, the figures were lower: about 3% (1 of 36 over both rounds) and 86%.
What we changed: The three pages quoted both rounds. (On September 23 we found that the "same room" in this test only meant a room was named, and on September 26 the test was retired from our pages; see above.)

August 19, 2026

Store listing

"Find cheaper same-room rates" in our store listing

What we said: Our Chrome Web Store summary said: "Find cheaper same-room rates and deals, taxes included."
What was wrong: By our own June measurement that held at about 1 hotel in 18.
What we changed: The summary became "Lower listed rates flagged honestly", in extension 0.4.9.

August 12, 2026

Press

A press pitch with a figure our data did not support

What we said: On July 15 we emailed five travel writers offering data on a "$100/nt same-room spread".
What was wrong: Our own June measurement had found the same room mostly priced the same, with real savings at about 6% of hotels.
What we changed: Our next pitches, from August 12, dropped it. We have no record of writing to the first five writers again.

July 13, 2026

Extension

"All-in" without the fees, and dollars without a label

What we said: The banner called a price "all-in" ("taxes & fees incl.") for every hotel, and showed US-dollar prices on pages in euros with no currency label.
What was wrong: When no site reports a before-tax price, the "all-in" figure is only the listed rate, and resort fees may be missing (we saw it at the Bellagio). The unlabeled dollar figure read as the page's own price.
What we changed: In that case the banner says "listed rate" and that resort or property fees may not be included. The first price on each card carries its currency. Extension 0.4.5, live July 13 (the currency note was finished in 0.5.5, above).