Study plan, registered on 11 October 2026. This study has not been run yet. I’m publishing the method before any results exist so it can’t be adjusted to fit them. Results will be added to this page, with the date they were collected.
On 10 October 2026 I asked four AI engines ten questions about current UK figures, such as the inflation rate and the energy price cap. 7 of the 40 answers didn’t say when the figure was from, and all 7 came from Google’s two engines (the study). This plan reruns the same ten questions every quarter, so the year-on call in that post isn’t the only data point.
What will be measured
The ten questions from the 10 October run, worded exactly as published there, on the same four engines with the same session settings:
- ChatGPT, default model, signed in, temporary chat set to unpersonalised.
- Google AI Mode, signed in.
- Claude, web search on, incognito chat.
- Gemini, signed in, temporary chat.
One fresh chat per question, no follow-up questions, run from the UK. Each answer is scored with the same definition of dated as the 10 October run, including the rule that a period appearing only in a cited source snippet doesn’t count. The answer key is re-read from ONS, the Bank of England, SMMT, GOV.UK and Ofgem on each run day, because the official figures get new releases between runs.
Perplexity stays out unless it can be run under the same conditions. Adding it would be a dated method change.
Schedule
10 January, 10 April and 10 July 2027, then 10 October 2027, which also scores P17 (Still open), the call in the original post.
Dated call
- Claim: the pattern from 10 October holds: on the 10 January 2027 rerun, Google AI Mode and Gemini together will give more undated answers than ChatGPT and Claude together.
- Called: 11 October 2026.
- Check date: 10 January 2027.
- Hit if: Google’s two engines give strictly more undated answers than the other two combined.
- Miss if: they give the same number or fewer, or the rerun hasn’t been run and published by the check date.
- Void if: one of the four engines no longer exists under that name, so the comparison can’t be made as written.
- Baseline: 7 undated answers from Google’s two engines and 0 from ChatGPT and Claude, on 10 October 2026.
This call is P20 (Still open) on the predictions page. If it misses, it stays up in red.
What I will not claim
- Why an engine leaves a figure undated.
- Anything about questions outside the ten.
- That one quarter’s change is a trend.
Data
Each run gets its own dated CSV, named undated-current-figures-YYYY-MM-DD.csv, with the same columns as the 10 October file. Past files are never edited.