About
A content score tells you nothing.
Semantic Checker exists to answer the question a score never does: which paragraph is pulling this article off topic, why, and what should change.
The problem
Most content tools hand you a number. You are told a draft scores 68, and then you are on your own — the number does not say which of your nine sections earned it, whether the weak part is the intro or a tangent halfway down, or what to actually edit. So the number gets gamed: keywords get sprinkled until it goes up, and the article gets worse.
The useful unit of analysis is not the document. It is the section.
What we do instead
You give Semantic Checker one target query and one draft. It compares every section against that query and returns a Semantic Checker Map — each section marked Aligned, Mixed, or Off-topic, with the evidence that produced the verdict and a prioritized list of what to fix first.
The result is something you can act on in the editor: this section wandered into industry history when the query wants a buying decision; cut it or tie it back.
What we believe
- —Evidence before recommendations. Every verdict shows its working. If you disagree with the diagnosis you can see exactly what it was based on.
- —Every run ends with an edit direction. Short or context-dependent sections receive a Mixed verdict with a limited-evidence note; unavailable external data is labelled as not included in the run rather than fabricated.
- —No ranking promises. Nobody can sell you a position in search results. We diagnose the draft in front of you and stop there.
- —You should not pay for a broken result. An analysis costs 1 credit; if it fails, the credit comes back automatically.
Where the product is today
Semantic Checker is an early, working product rather than a finished one. Paid analysis uses a recent US-English Google organic snapshot from DataForSEO, collected live or reused for up to six hours, reads accessible competitor pages, and compares their recurring topic language with the draft. The draft itself is not sent to DataForSEO or a language model.
The scope is deliberately narrow while we learn: English drafts, one target query at a time. A coverage score requires at least five readable competitor pages; otherwise the local report is saved and the credit is refunded.
Try it on a real draft
Eligible new users get 2 analyses, no card required. Bring something you have already published and see whether the section map matches your own instinct about the piece.