Full guide: Clairic field types: text, number, date, list, estimate and forecast
Every single-value answer from Clairic carries p_null, the probability that the value is not in the input. The null_threshold setting decides what happens next. The Workers AI schema puts it as “Probability of null above which the value is null (default 0.5).”
The schema lists it for every field type. Set it lower on fields where a wrong value costs more than a missing one, such as an amount that releases a payment: the field comes back null more often and your code asks for the value instead. Set it higher when you would rather review a doubtful value than lose it.
Either way the answer keeps guess, which the Cloudflare blog says “shows what the model would otherwise have picked.” So the threshold changes what lands in value, not what the model saw. It is the extraction-side cousin of confidence gating. The Clairic field types guide shows it beside the other settings.