GoNow · Method

The facts we refused to publish.

Travel sites show you what survived their checks, if they check at all. This page shows the other pile: what our verification caught, cut or corrected, with the source that settled each one. The blanks are here too, counted, because a blank we admit beats a number we invented.

785 venues2,355 fact slots1,451 confirmed904 left blank

The pipeline, in plain words

1

Discover

Candidate venues and facts are gathered from real pages, each with the URL it came from.

2

Extract, with citations

Every fact keeps three things or it does not exist: the source URL, a verbatim snippet containing the fact, and the date it was read.

3

Verify, adversarially

A separate pass re-fetches every cited page and checks each snippet against it word for word. It is built to distrust the pass before it.

4

Publish, or refuse

A fact that reproduces ships with its evidence. A fact that does not is corrected to what the page actually says, or set to null. Null beats a guess, always.

The rejections log

17 records

Everything below is real, dated, and was caught by the pipeline this page describes. Each record names the source that settled it, so you can check any of them yourself in about a minute.

Closed in the world, still open online

3 records

Three venues were discovered by the research pass and cut by the verifier before they ever reached the live map, because the physical place they described no longer exists.

R-01

Rejected

A Johor Bahru candidate record wrote up Danga Bay's fairground and its rides, drawn from a sourced description of the site.

Danga Bay's fairground closed in early 2024, and the sourced description was of attractions that no longer exist there. The venue was rejected outright rather than published with a caveat attached.

Search "Danga" on All places: zero rows.

R-02

Removed

Cinema Winehouse, Bangkok, was surfaced as a candidate venue in the second scale pass, ready to be written up.

Its official domain resolved to nothing and every signal read permanently closed, so it was dropped before publication rather than listed with a dead link.

The venue's own domain, dead on inspection.

R-03

Removed

The Museum of Floral Culture, Bangkok, was surfaced as a candidate venue in the same pass.

Trip.com marks the museum permanently closed, so it was cut rather than published on the strength of the candidate description alone.

Trip.com's own listing, marked permanently closed.

Evidence that failed the re-fetch

12 records

Every fact here cites a page, so the verifier re-opens the page. These are the ones whose quoted evidence did not survive that second look.

R-04

Corrected

National Museum of Singapore adult entry arrived from a listing page, quoted snippet attached, ready to ship as a sourced fact.

The quoted snippet was not on the cited page, and the figure it carried was stale twice over. The fact was re-sourced to the museum's own admissions page, which lists $20 standard adult with free entry to all permanent galleries for Singapore citizens and PRs. It ships at official tier now.

nationalmuseum.nhb.gov.sg, admissions

R-05

Removed

Mega Adventure Park carried a bundled ticket priced "from SGD 18", cited to a forevervacation.com writeup.

That figure does not appear anywhere on the cited page. Removed. The commonly quoted per-ride price stays out too, since it was never captured verbatim from any fetched page either.

forevervacation.com, the cited page (no SGD 18 figure on it)

R-06

Corrected

Universal Studios Singapore's free-admission age threshold was recorded as "under 3".

The source states the threshold as "below 4", not under 3. Corrected to match the page's own wording.

singapore-tickets.com, Universal Studios tickets

R-07

Corrected

Three Mandai wildlife park price facts (Singapore Zoo, River Wonders, Bird Paradise) carried snippets that in fact stated only the under-3 admission policy, no price at all.

A value with no price in its own snippet does not stand, so all three were pulled and re-sourced from Mandai's own ticket pages, which do carry adult, child and resident price tiers.

mandai.com, Singapore Zoo tickets

R-08

Corrected

Singapore Botanic Gardens, National Orchid Garden, carried a child-price snippet reading "Child below 12 years of age Free Free".

That snippet was invented. The page's actual admission table lists the line once, not twice.

The invented snippet, as it had been recorded

Child below 12 years of age Free Free

The page's actual line

Child below 12 years of age Free

sbg.nparks.gov.sg, National Orchid Garden

R-09

Corrected

OCBC Skyway's price fact stated that resident rates were not listed on the page.

The page does list a Singapore resident tier alongside the non-resident one. Corrected to state both.

gardensbythebay.com.sg, OCBC Skyway

R-10

Corrected

Flower Dome's price fact read a struck-through promotional figure on the page as a resident-versus-tourist price split.

The struck-through number is a crossed-out promotional price, not a residency tier. Corrected to the page's actual resident and non-resident admission figures.

gardensbythebay.com.sg, Flower Dome visitor information

R-11

Corrected

A sample AirAsia flight snippet was quoted with fields trimmed from inside its own quotation marks, and the source's em dash had been silently rewritten as an en dash.

A quote that edits its own punctuation is not a quote. Replaced with the true contiguous substring straight from the page's structured data block, dash included.

airasia.com, Bangkok to Chiang Mai flights

R-12

Corrected

Horizon Village & Resort, on the Bangkok to Chiang Mai route, was priced at US$113 for two nights, two adults.

Re-fetching the identical search (same dates, same occupancy) showed US$86, and the listing's own card labels the property Doi Saket, not Old City as the search had implied. Replaced with the re-verified rate, and a note that hotel rates are dated snapshots, not live prices.

booking.com, the same search re-run

R-13

Corrected

The Bangkok to Siem Reap flight time was recorded as about 1 hour 30 minutes.

The cited page's own sentence reads "1 hour and 20 minutes is the average flight time from Bangkok to Siem Reap." Corrected to the source's own figure.

flightconnections.com, Bangkok to Siem Reap

R-14

Corrected

Two snippets were dressed up as verbatim quotes when they were actually paraphrases: a Tourism Authority of Thailand railway advisory on the Bangkok to Siem Reap route, and a Transtar coach schedule on the Singapore to Kuala Lumpur route.

Both were restructured summaries, not the source's own sentences. Both were replaced with the pages' actual wording.

tatnews.org (rail advisory) and checkmybus.com (Transtar)

R-15

Removed

Five figures (US$208, US$24, US$35 to US$45, US$28, and a duration of 8.5 to 9.5 hours) sat inside prose notes on the Bangkok to Siem Reap route, reading like sourced facts.

None of the five had ever been fetched from any page. A note is still a surface, so a check on the value fields alone would have missed all five. All five were stripped from the notes, and the route page's own verification note records the removals.

The route's verification note, live on the page, listing exactly these removals

Where the published world disagrees with itself

2 records

Sometimes every source is confident and they still disagree. The record then shows all of them, and says so, instead of quietly picking one.

R-16

Conflict shown

The Bang Krachao ferry crossing at Wat Bang Na Nok pier needed a fare.

The posted fare board at the pier read 7 THB per passenger and 16 THB with a motorbike, read on site on 2026-07-25. Published articles give lower and conflicting figures, 3 to 4 THB per passenger and up to 10 THB with a motorbike. OpenStreetMap's own tags on the ferry way state 5 THB, from a survey dated 2025-01-27. Nothing agrees, so the record states the board, the articles and the map, and tells the reader to read the board on the day rather than picking one silently.

OpenStreetMap's own tags on the ferry way

amount "5 THB", source "survey", source:date "2025-01-27"

OpenStreetMap way 40346702

R-17

Withheld

A nearest-station transit feature was built and dry run across 531 venues in three cities, to answer "how do I get there" next to "is it open".

OpenStreetMap maps stations that are not yet in service and does not tag them consistently. The MRT Orange Line is under construction, yet its Democracy Monument station node carries its construction marker under a key the dry-run filter did not check, so the run reported Thipsamai as 207 metres from a station nobody can board. A name-based exclusion list was tried and rejected, because the first attempt swept in four open Blue Line stations along with the unopened ones. The feature ships nothing until an operator source exists that cannot contain an unopened station.

The station node's own note tag, its English half

Do not add railway=station and/or public_transport=station tags before operational service

OpenStreetMap node 13971907544, every tag visible on the page

Audit a random fact

No fact drawn yet.

The number, without the varnish

Evidence behind every fact slot

  • We went ourselves110.5%
  • Official source69929.7%
  • Booking site23810.1%
  • Single report50321.4%
  • Left blank90438.4%

Every one of the 2,355 fact slots sits in exactly one band. 1,451 carry a published fact and the tier of evidence behind it. The hatched band is the 904 we left blank, and it stays on the chart at full size because that is the honest shape of the dataset.

Coverage by field

Hours

649 of 785

82.7%

Price

514 of 785

65.5%

Age suitability

288 of 785

36.7%

1,451 of 2,355 facts confirmed. 61.6 percent.

The aggregate hides the shape, so the fields lead. Hours are the most publishable fact in the world and sit at 83 percent. Prices move and sit at 65. Age suitability sits at 37 because most venues never state one anywhere: a hawker stall has no age policy page, and a child ticket price band is a price, not a suitability claim. Those blanks are the product working. The alternative is inventing 904 answers.

Of the 904 blanks, 796 were investigated and the fact is genuinely unpublished. 108 have not been attempted yet. The database records which is which.

Verified place by place, and the failures kept on record.