Claude blog post validator

This route is a generator. The draft you paste is treated as the brief, the server returns finished content matching the format the tool name implies, and that content is a written review of the draft. There is no rule engine underneath, no rubric file and no threshold, which is why the result reads as an editor's notes rather than as a report.

The operating envelope is fixed: 5 runs per rolling 24 hours, input bounded at 1 to 12,000 characters, an optional instruction capped at 300 characters, a live character and word meter above the input, and a Copy control on completion. Nothing is stored between runs and no account exists.

The draft you submit defines the scope of the review, so what you leave out is a decision.

Step 1

Submit the draft with its heading markers intact. Headings are the only structural signal a text tool receives, and stripping them removes half the available analysis.

Step 2

Use the instruction field to state a criterion rather than a topic. "Does each section deliver what its heading claims?" is testable; "check my post" is not.

Step 3

Treat each finding as a claim about the text and verify it. A review that cannot be checked against the draft is not worth acting on.

The Apify blog post SEO check takes a URL array and reports title, description, headings, word count and Core Web Vitals, with output columns of URL, Score, Title, Issues and Words. The AIOSEO analyzer also starts from a site address and covers metadata, content quality, internal links, schema, technical SEO, mobile usability, page speed and indexing. PublishPress Checklists takes the third route, living inside WordPress and reading the post record directly.

All three need something the draft does not contain: a served response, or a database row. That is not a failing, it is the nature of those checks. The consequence is a genuine blind spot in the category, since none of them can look at a piece of writing that has not yet become a page. This route occupies exactly that window and nothing else.

A text-level validator is bounded by what survives the paste. These four do.

Hierarchy

One top-level heading or several, and whether levels descend without skipping. PublishPress puts single-H1 and logical header order behind its Pro tier; in pasted text they are simply visible.

Promise and delivery

Whether each heading describes the section under it, and whether the title's claim is met by the body. This is a reading task, not a measurable one.

Redundancy

Sections that restate each other, the characteristic failure of a long draft assembled over several sittings.

Distribution

Word and character counts are live in the meter, and the review can say which sections are disproportionate to their importance.

PublishPress publishes the most complete checklist in the set, and most of its items are properties of a record rather than of prose: featured image present, its alt text and caption, image dimensions and counts, internal and external link counts, minimum and maximum tags or categories, excerpt length. A pasted draft carries none of those fields.

Then there is the served page: page speed, Core Web Vitals, mobile usability, schema, indexing status. And link resolution, which requires a request this route never makes, so a URL in your draft is a string. Anything claiming to validate these from pasted text would be asserting knowledge it does not have.

A score is the most requested feature in this category and the least defensible. Apify returns one; so does most of the audit tooling. The trouble is that a composite number over heterogeneous checks has no unit, so a rise from 72 to 81 does not describe anything in particular, and a threshold invites editing toward the threshold rather than toward the reader.

Written findings are harder to skim and harder to fake. Each one is either true of the draft or not, and the check takes seconds. That is the trade this route makes deliberately, and it is also a consequence of the family: a generator returns content, so a scoring system would have to be invented rather than measured.

Long-form drafts exceed the input ceiling, and truncation at the boundary would silently change what is being reviewed. The meter above the input makes the overflow visible before the run rather than after it.

The efficient decomposition is structural first, prose second. One run on the heading tree plus each section's opening sentence answers the questions about shape, sequence and duplication for the whole piece. Subsequent runs then go to individual sections on their merits. Reviewing every section in full spends five runs to learn what the first run already established.

No URL input, no crawler, no site audit, no link resolution, no schema inspection, no Core Web Vitals, no indexing check. No score, grade or pass state. No CMS integration with WordPress, Yoast, Rank Math or All in One SEO, and no ability to make publication conditional the way PublishPress does when a task is set to Required.

No keyword volume, ranking data or competitor comparison, because no search index is reachable from here. No rewriting, no plagiarism matching, and no verdict about whether a detector will consider the post machine-written, which we do not claim on any route. No stored history, no API, no bulk queue and no account.