Reports

Is the system getting better, is it earning, and where is everything right now?

Can readers tell our AI writing from the real writer?2 tests

A reviewer sees short excerpts and guesses who wrote each one. Here, about 50% is the goal — it means they're guessing. A high score means the AI version is easy to spot, so that voice isn't ready.

Warm voice

40 excerpts · labelled by Sara · closed 4 d ago

Not ready

72.5%

guessed correctly (29 of 40)

0%50% · guessing100%

Readers could tell the AI version apart — this voice isn't ready to scale.

Witty voice

41 excerpts · labelled by Sara · closed 7 d ago

Passes

51.2%

guessed correctly (21 of 41)

0%50% · guessing100%

Readers couldn't tell — the voice carried over.

Is each voice drifting?weekly

Distance from each writer's real style, week by week. A steady rise means the voice is slipping and an alarm goes to the quality owner.

Warm voice Alarm sent to Imran

▾ Data table

Date	Warm style distance
2026-08-31	0.95
2026-09-07	0.94
2026-09-14	0.97
2026-09-21	0.94
2026-09-28	0.97

Witty voice

▾ Data table

Date	Witty style distance
2026-08-31	0.95
2026-09-07	0.96
2026-09-14	0.95
2026-09-21	0.97
2026-09-28	0.98

Words the AI is overusing6 watched

How much more often a word appears than expected. Words well above normal are put forward for the banned list.

Word or phrase	Times more common	Status
“delve”	25.1×	Put forward for ban
“showcasing”	9.2×	Put forward for ban
“underscores”	9.1×	Put forward for ban
“comprehensive”	3.5×	Watching
“crucial”	3.2×	Watching
“notably”	3.0×	Watching