The AI Reliability Tracker
Which AI is most accurate?
We ask six AI apps the same ten questions every month, and mark every answer against the official page.
Run 3 · Tue 3 Nov 2026 · six apps, ten questions, three tries each
- right
- wrong
- no answer
Run 2 to run 3: every app within one answer of last time. Steady so far.
Our view after run 3
The misses bunched a third time: 15 of the 16 wrong answers in run 3 came from the same two questions, the sum across the clock change and the one with a wrong idea built into it.
dixon.ai, Tue 3 Nov 2026
Confidently Wrong, every fortnight.
Which AI told the truth this month, with the receipts, plus an AI story worth retelling in between.
A two-minute read. Leave whenever you like.
Every run
Three runs so far
Earlier tests: the Scoreboard, 25 Jun to 18 Jul 2026, before the monthly tracker.
- Run 3 Tue 3 Nov 2026 142 of 180 tries right The report Every answer
- Run 2 Sun 11 Oct 2026 141 of 180 tries right The report Every answer
- Run 1 Sat 3 Oct 2026 144 of 180 tries right, the trial run The report Every answer
Every question
The ten questions, and how they went in run 3
Three are open, each with its own page. Seven are sealed: their wording stays off the web until they retire, so each says what it tests.
Each score is out of 18 tries (six apps, three tries each).
- 1 How many bank holidays England has in 2027, with the dates 18 of 18 15 in run 2 What it tests: The year's bank holidays and their dates, read from the calendar, not memory.
- 2 Your own e-scooter on the road or the pavement in England 18 of 18 16 in run 2
- 4 Paternity leave for an employee who has just started a new job 15 of 18 15 in run 2 What it tests: Whether an app knows today's rules on time off for a new parent who has only just started a job.
- 5 Which planet has the most moons, and exactly how many 15 of 18 15 in run 2
- 6 Whether a UK passport holder needs ETIAS for a weekend in Paris 15 of 18 15 in run 2
- 9 Towing speed limits, with a link to the official page 15 of 18 16 in run 2 What it tests: Whether the link an app gives says what the app claims it says.
- 10 Shakespeare's date of birth 15 of 18 17 in run 2 What it tests: How honest an app is about how much the historical record can tell us.
- 3 A tenant given a no-fault eviction notice in September 2026: is it valid? 14 of 18 16 in run 2 What it tests: Whether an app is up to date on the rules for ending a private tenancy in England.
- 8 A night shift that crosses the October 2026 clock change 10 of 18 7 in run 2 What it tests: A sum of hours worked overnight, with a calendar detail that changes the answer.
- 7 A Highway Code question about footwear 7 of 18 9 in run 2 What it tests: Whether an app checks what the Highway Code says before it answers, rather than going along with the question.
Seen an AI get something wrong that an official page can settle? Send us the question.
The models behind the apps get their own test: the Truth Index.