Give it one target query and your draft. Every section comes back Aligned, Mixed, Off-topic or Unknown — with the wording behind each verdict and the section worth fixing first. No single score to chase, and no language model reads your writing.
2 free analyses. No card required.
Semantic Checker report
semantic content editor
Summary
Mostly aligned, with one section pulling the article away from search intent.
Semantic Checker Map
4 sections
Introduction
Aligned
Feature overview
Mixed
Industry history
Off-topic
How to choose
Aligned
Why it is small
Content optimizers keep growing — writers, audits, rank tracking, AI visibility. If you need all of it, buy all of it. This is for the check you actually run before publishing.
The text analysis is deterministic and runs on our own server. Nothing is sent to a third-party AI provider.
There is no sampling step, so a draft and a query always produce the identical verdict. You can re-check an edit and trust the difference.
A pasted draft is analysed in-process, so the report is there immediately. Importing from a URL adds a few seconds.
A content suite
Priced as one subscription, whether or not you open eight of them.
This tool
One job. Buy ten of them for $9 and stop.
Workflow
The whole check fits between finishing a draft and publishing it. Each step hands you evidence and leaves the editorial call where it belongs.
One query, one draft. Paste the text or hand over a published URL — there is no project to set up, no keyword list to import, and a URL that refuses to import costs you nothing.
You decide
Which single query this page is really trying to answer.
The draft comes back as a map of its own sections, each marked Aligned, Mixed, Off-topic or Unknown — alongside the wording that earned the mark and the subject the section leans on instead.
You decide
Which flags are real, and which ones you meant to write.
The weakest sections are ranked and each arrives with a direction to take — reframe it, move it, tighten it, cut it — instead of a list of terms to sprinkle back into your sentences.
You decide
What actually changes before this goes out, and what stays.
Read the full method and its limits — how sections are split, how verdicts are decided, and what the report cannot tell you yet.
What drifting looks like
A report marks the section and shows the text behind the verdict. Naming the pattern is the part you do — these are the six worth knowing.
The section is about a neighbouring subject, not yours.
A page targeting AI headshot generators spends three paragraphs on general photo editing.
Right subject, wrong job.
The reader came to use a tool; the section explains what the category is instead.
The lens is too wide or too narrow.
A piece about LinkedIn headshots turns into a general portrait photography guide.
Something the subject needs never appears.
A headshot page never mentions business profiles or corporate portraits at all.
A side feature crowds out the main one.
Filters and background removal get more room than the generation flow itself.
The heading promises one thing, the text delivers another.
An H2 reads “How it works” and the paragraphs underneath only discuss pricing tiers.
The report
Every signal remains explainable. If the data is not available, the report says Unknown instead of inventing certainty.
Scan Aligned, Mixed, Off-topic, and Unknown sections at a glance.
Compare the likely query job with the emphasis of your draft.
Start with the weakest section and a concrete editing direction.
Unavailable SERP evidence stays visibly Unknown.
Against score chasing
Give a writer one score and a suggestion list, and the work becomes clearing the list. Terms get wedged into sentences that did not want them, the bar turns green, and the draft reads worse than when it started.
So this report refuses the single number. Each section gets a verdict and the text that produced it. You read the evidence and make the editorial call — which is the part a tool should not be taking from you.
A priority fix, in full
Industry history drifts away from the target query
Pricing
A renewal can get more expensive even when the screen you use has not changed. Here, $9 buys 10 analyses that never expire. Subscriptions exist if the volume suits you.
Free
2 reports
Validate the workflow with real content.
Credit pack
No subscription$9 once
10 analyses, yours until you use them.
Creator
Most popular$19 / month
40 analyses every month.
FAQ
Including the ones where the answer is no.
The three kinds of text analysis, and what each one costs you →
Surfer is a suite. Its plans bundle an AI writer, a writing assistant, plagiarism checking, site audits, internal linking, cannibalization reports and — as of 2026 — AI search visibility tracking, from $49 to $299 a month. If you use most of that, it is good value and you should keep it.
This tool does one thing: it reads your draft section by section and tells you which parts have wandered away from your target query. There is no writer, no audit, no rank tracking. That is the product, not a roadmap gap.
The full comparison, with both monthly floors side by side →
Because a single number invites you to game it. The common failure mode with scored optimizers is writers adding suggested terms until the bar turns green, ending up with stiff, repetitive copy that reads worse than the draft they started with.
A report here gives you per-section verdicts — Aligned, Mixed, Off-topic or Unknown — each with the text that produced it. You are meant to read the evidence and make an editorial call, not chase a target.
No, and that is the main difference in how this is sold. Content optimizers generally have a monthly floor: NeuronWriter starts around $23, Frase and Surfer around $49, Clearscope at $129 for 20 reports. You pay it whether or not you publish anything that month.
Here, $9 buys 10 analyses and they never expire. Subscriptions exist if you want them — Creator at $19 a month for 40 analyses — but nothing about the tool requires one.
No. The current analyzer runs entirely on our own server using deterministic text analysis. No language model sees your draft, and nothing is sent to a third-party AI provider.
This is worth comparing directly: any tool built around an AI writer or AI assistant necessarily processes your draft on its servers and through a model to deliver those features. If you are working on client content or something unannounced, that difference matters.
Lexical text analysis, computed on our own server. Your draft is split at its headings, stop words are dropped, and the concepts in your target query are pulled out. Each section is then measured three ways against that vocabulary: how much of it appears at all, how often, and how much of it reaches the heading. Those three numbers are the alignment figure behind every verdict.
Separately, each section's own most frequent terms are ranked, which is what lets the report name the subject a drifting section leans on instead of guessing at one. Because it is a text analysis tool and not a model, the same draft and query always return the identical report — so re-running it after an edit tells you something real.
That also fixes the limits. This is a topical text analyzer, not a readability grader, a word counter or a general NLP toolkit, and every judgement it makes is relative to the one query you gave it.
The three kinds of text analysis, and what each one costs you →
No. It tells you which section is the problem, what it appears to be about instead, and what direction the edit should take. The writing stays yours.
This is a deliberate scope decision rather than a missing feature. Diagnosis and generation are different jobs, and there is no shortage of tools that will write for you.
No, and be wary of any content tool that implies otherwise. Nobody outside Google can see how a page is ranked, and no third-party score is a ranking signal.
What the report gives you is a reproducible reading of whether your draft stays on the topic you targeted. That is a useful editorial check. It is not a ranking prediction and we will not dress it up as one.
The report says Unknown, and it says it in the place the number would have been. Sections shorter than about 50 words, and headings that only make sense in context, get marked Unknown rather than scored.
Live search-results comparison is not wired into the analysis path yet, so SERP coverage currently reports Unknown for every report. We would rather show you the gap than fill it with an estimate.
A target query and a draft. That is the whole input.
You do not need a keyword research subscription first. Most people arrive with a query they already picked from Search Console, Ahrefs, Semrush or a client brief, and that is exactly what this expects.
The credit comes back automatically. Refunds are idempotent and read the amount from the original ledger entry, so a retry or a replayed webhook cannot double-refund or leave you short.
Failed URL imports never cost anything in the first place — importing is free, and only a completed analysis is billable.
English drafts only, and one target query at a time. Analysis assumes a US English context.
Other languages are not on a near-term roadmap. Getting section-level judgements right in one language is the current job.