Why a page cannot be read · low_confidence
The extraction did not clear the trust gate
Typed data was there and something was extracted, but not well enough to rely on, so nothing is returned rather than something uncertain.
The page was read, but the extraction did not clear the 0.7 trust gate, so it is not safe to rely on.
§01 · what is actually happening
What is actually happening.
The extraction produced a result whose calibrated confidence landed below 0.7, which is the gate at which a read is treated as trustworthy.
The usual causes are specific and fixable: a price with no currency attached, a product name that differs between the markup and the visible page, or several structured blocks on one page that disagree with each other.
Below the gate the answer is withheld rather than returned with a caveat. A number that is probably right is worse than no number, because it gets used.
§02 · what it costs
What it costs.
You have done most of the work and are getting none of the benefit. The markup exists, and the disagreement inside it is enough to make the whole page unusable to anything that checks before it answers.
Anything that does not check before it answers will use it anyway, which is the worse outcome of the two.
§03 · what would change it
The fix.
Structured data was present but incomplete or inconsistent enough that the extraction could not be relied on. The usual causes are a price without a currency, a name that differs between the markup and the visible page, or several conflicting blocks on one page.
Who does it. Whoever owns the product template. Usually a small correction rather than new work, because the markup is already there.
Check your own site.
One request, one verdict, no account. If this is the finding you get back, the page above is the one that applies.
See what a machine seesThe other findings