The widgets below are the real components, using the extension's own stylesheet and class names. The percentages are real figures from a live united.com search and from the published fleet data. Nothing here is a drawing of the product.
The whole point of the release. It names one flight only when the data earns it, and otherwise tells you exactly why it declined. Click through all eight states.
Best WiFi choice
64%11 points higher historical odds than UA2265
12 tracked departures · Medium confidence · Historical tracker odds
Two or more scored flights, an 8-point-or-better lead, and the tracker grades the leader high or medium confidence. Rare and earned.
Next-gen odds lead, streaming-class ConnectScore follows, and both say their own name.
The old build showed a bare 🛰 48% that meant per-flight odds on one site and an airline-wide score on another. Eight row states, and not one of them prints a zero for something
unknown.
A real historical probability for this flight number. Only this state takes the colour ramp.
The confirmation is a separate token beside the odds, never folded into them.
The tracker knows this flight number and has no history for it. It is not zero, and not “no Starlink”.
The request failed. We could not measure it, which differs from measuring nothing.
No per-flight odds for this airline, so the row shows its fleet share instead, outlined rather than on the odds ramp, because it is a different kind of fact.
The airline has committed to a next-gen system that carries no passengers yet. That is a promise, not a percentage.
A current-fleet fact, written as words rather than a numeric zero.
The airline is not in the data set. This is unknown, which is not the same as absent.
On single-carrier sites every row is the same airline, so ordering by odds displaces nobody. Sorting is on by default there. Press the button to watch the page reorder, then take it back. "Keep site order" restores the booking site's own sequence exactly.
On mixed-carrier sites like Navan and Google Flights this is off by default. Moving a scored United flight up necessarily moves an unscored airline down, and an airline we have no data for is unknown, not worse. There you get an explicit action and a coverage line instead.
Coverage: United. Other airlines stay unscored and keep the booking site's order.
Star a flight and the extension re-checks its tail assignment from booking to boarding. Three honest states, each carrying its reason. An aircraft whose WiFi cannot be determined says so, and never gets downgraded to "No Starlink".
The airline published the aircraft and it has Starlink.
A known aircraft whose equipment we cannot verify. Never rendered as a negative.
This differs from "Update unavailable" (an outage) and "Flight not found" (a bad number).
These are two different questions, never merged into one leaderboard.
The chance this flight draws a Starlink or Amazon Leo aircraft. Per-flight history exists for United and Alaska only, 2 of 18 airlines. For the rest, the row shows fleet share and says that is what it is.
How good the airline's system is today, published as a floor across the whole fleet. United 42, Alaska 55, Southwest 16. Unresolved aircraft never count as zero.
Built 2026-08-01 from extension 3.0.0. The
components inline extension/content.css verbatim and use the class names the content
script emits, so this page cannot drift from the product without the build changing too. Flight
percentages were captured from a live united.com DEN→SFO search on 31 Jul 2026; airline scores come
from the extension's airlines.js, which is byte-generated from the site's model.