Content Structure Checker

Start with a source and a clearly named review

Choose the document you actually want to inspect. A complete HTML export can include the page head and body. A copied article fragment may contain useful headings and links but omit the title, description and canonical declaration. Both can be reviewed, provided the handoff identifies the scope. A missing declaration in a fragment means it was not found in that supplied material, not that the published website definitely lacks it.

Juniper Clay names its project Beginner pottery page, draft 3. It obtains HTML through its normal website editing workflow and pastes that source into the checker. Alternatively, open a UTF-8 source file. Entering a page address is not a request to download it. The optional base URL supplies address context for relative references; you still need to provide the content separately. Keep the original draft available so the editor can implement decisions in the correct document.

Select HTML when the input contains actual markup. Select text when you want a source word count without structural extraction. Text mode does not guess headings from short lines, recognize pasted addresses as a link inventory or recover metadata from prose. A sentence describing a photograph is not an image record. Choose the mode according to the supplied format, not the richness of the report you hope to receive.

The platform selector labels the website project. All platforms, Instagram, TikTok, YouTube, Facebook, LinkedIn, Threads, X, Snapchat and Reddit use the same inspection rules. These contexts concern resources on your own website that support social activity. They do not connect accounts or inspect native profiles. For other preparation tasks, return to the free tools directory after saving this review.

Understand what local inspection includes and leaves out

Run analysis once the source, mode and optional base are ready. HTML is parsed as a document in a local browser worker. Supplied scripts are not executed, images are not loaded, and the checker does not insert your markup into a page preview. It extracts records from the parsed source rather than reproducing a visitor’s browser session. Ordinary site analytics remain separate; the pasted content is not sent to them.

For prose counts and the heading, link and image inventories, content in script, style, template, noscript and head is excluded. Supported metadata declarations are inspected separately. The word count uses whitespace-separated source tokens containing letters or digits. It is not a measurement of visible words, reading difficulty or how much content fits on a phone. A hidden HTML paragraph can contribute source words because CSS visibility is not evaluated.

Read counts as an inventory summary. Six headings means six extracted native heading elements, not six well-designed sections. Three links means three extracted link records, not three working destinations. A small count can expose an unexpectedly incomplete paste; a large one can reveal that navigation and footer material were included. Neither number earns a quality grade. Compare the scope with what you meant to supply before interpreting individual findings.

The checker refuses excessive workloads instead of presenting a shortened inventory as complete. Each source allows up to 200,000 UTF-16 code units and 512 KiB of UTF-8 data, with extraction caps of 200 native headings, 400 links, 200 images and 400 metadata records per source. The primary checklist allows up to 500 findings in total. If a limit is exceeded, prepare a smaller, clearly identified source. Cancel an unwanted analysis rather than treating an unfinished result as evidence.

Read the pottery page outline before changing it

Juniper Clay’s supplied body has this sequence: H1 Beginner pottery workshop; H2 Choose your class; H4 What to bring; H2 Choose your class; H2 Collection times; and an empty H2. That is six headings: one H1, four H2 elements and one H4. The checklist identifies an empty heading, repeated nonempty heading text and a descending jump from H2 to H4. There is no need to invent an overall score from those observations.

The editor checks each source location. The empty H2 came from a removed gallery heading, so the leftover element can be deleted in the website editor. What to bring describes a subsection of Choose your class, making H3 a reasonable proposed level. The repeated Choose your class introduces collection instructions rather than class options. Renaming it Preparing your finished piece could clarify the promise, but that decision requires reading its paragraphs, not merely obeying a duplicate flag.

Native H1 through H6 elements establish the extracted levels. Large bold text, a styled paragraph or an ARIA heading role does not become a native heading in this inventory. Conversely, a small-looking H2 remains an H2 here. The outline follows the parsed source order, not the order a layout may display. Check the actual page separately when visual order or assistive-technology behavior matters. This review is not an accessibility certification.

Use the source line and element ordinal to find the right occurrence, especially when text repeats. Lines belong to the supplied snapshot and can shift after editing. A finding about multiple H1 elements is a request to inspect their purpose and scope, not a prediction of search penalties. The heading outline review guide explains level changes, duplicate occurrences and exact comparison in more detail.

Separate metadata declarations from search appearance

Switch to the declarations inventory rather than expecting the visible H1 to tell you everything. The checker records supported title, description, canonical and robots declarations with their placement and source line. Juniper Clay’s draft contains one title, Beginner pottery workshop, one description summarizing the session and one canonical reference. Those records establish what this source declares. They do not establish which title or snippet a search engine currently displays.

Its robots declaration contains noindex because the draft was prepared for a private preview. The review needs a decision: should the eventual public class page retain that instruction? The tool reports the declaration but cannot tell whether a live server supplies the same HTML, adds an HTTP header or serves a different version. Noindex in a pasted draft is not proof that an existing public address has been removed from an index.

Missing, empty and multiple declarations are different situations. No description may mean the supplied fragment omits the head. An empty description means an extracted declaration has no useful value. Two descriptions warrant locating the source of both, perhaps a theme setting and an editorial plugin. A declaration outside the head is a placement review. Inspect the records before deciding which setting to change; a count alone cannot identify the intended owner.

Assess wording against the actual offer. A pottery class title should not promise a finished glazed piece on the same day if collection happens later. The checker does not apply arbitrary title or description length pass marks, verify claims or choose a canonical on your behalf. Use the metadata and link review guide for a focused declaration review, then verify any live implementation through your normal website workflow.

Give link and image records the correct address context

Juniper Clay enters https://juniper.example/classes/beginner/ as the base. Its supplied anchors point to booking/, #collection and mailto:studio@juniper.example. The first resolves to https://juniper.example/classes/beginner/booking/; the second retains the page path and collection fragment. The email reference is a non-HTTP(S) link for review, not evidence of a broken destination. Different reference types need different decisions.

Relative addresses resolve only against the base you explicitly enter. A base element inside the source is ignored and reported as a review note. Without an entered base, relative references remain unresolved for review, while absolute HTTP(S) addresses can still be recorded but are unclassified as internal or external. With a base, internal means the resolved origin matches its origin. A similar brand name or related subdomain does not establish that equality.

Read the original reference, resolved value and source location together. A syntactically valid booking address may still be the wrong route. No HTTP request checks its status, redirects or content. A retained fragment does not confirm that the named target exists or works. Invalid or credentialed references need attention before sharing, but the report is not a vulnerability scan. Pasted destination values stay inventory text rather than becoming external clickable links in the tool.

The workshop also supplies two img elements. A photograph of a completed bowl has no alt attribute; a decorative clay-texture divider has an explicitly empty one. Those are two separate review situations. The editor can propose a description for the informative photograph and retain empty alt for genuinely decorative artwork. The checker does not see either image, judge the accuracy of a description or verify a file’s availability. An image count cannot prove the live page’s visual appearance.

Compare two supplied heading inventories literally

Add a second supplied source for comparison. Both inputs share the chosen mode and explicit base. The comparison receives a full inventory, but only the primary source receives a review checklist. Identify each version before interpreting the difference. HTML comparison matches native heading level together with normalized text and counts repeated occurrences. This is a literal multiset comparison, not a semantic reading of the two pages.

Suppose the workshop changes only H4 What to bring to H3 What to bring. Comparing the old inventory with the revised one gives one removed H4 occurrence and one added H3 occurrence. The text stayed the same, but the level is part of the matching key. If an editor instead rewrites a heading, a removal and addition can describe that single editorial revision. They do not necessarily mean one topic disappeared and an unrelated topic arrived.

Now imagine the two identical H2 Choose your class headings become one occurrence while everything else stays fixed. The multiset difference records one removed occurrence, not the loss of the phrase everywhere. On the other hand, moving Collection times above the class options without changing its level or normalized text need not create an addition or removal. The comparison is not reliable moved-heading detection. Inspect source order independently when sequence is the actual question.

Paragraph rewrites, dates, price corrections, image substitutions and destination changes can happen without changing a heading key. An unchanged comparison therefore does not mean the document is unchanged, current or approved. Text mode supplies counts without inferred heading records, so it cannot recreate this outline exercise from ordinary prose. For the pottery handoff, compare the intended structural changes, then read the altered booking and collection details separately. Keep both source snapshots if the revision needs to be explained later.

Turn findings into decisions with accountable next steps

Begin with reader consequences rather than the number of flags. Juniper Clay first checks the booking reference and collection wording because visitors use them to decide whether to attend. Next it resolves the leftover heading and reviews the draft-only robots instruction. Finally it considers whether the decorative image needs any change at all. This order comes from the page’s purpose; the checker has not calculated urgency or predicted a conversion improvement.

For each decision, identify the source location, proposed action and owner in your existing work record. A practical task says Draft 3, empty H2 after Collection times: remove leftover block in website editor. Another says Confirm whether noindex belongs on the public version before release. These instructions distinguish a specific source edit from an unresolved implementation question. A generic instruction to fix SEO gives the next person little evidence and can encourage unnecessary changes.

Mark an individual finding reviewed only after considering it. Reviewed means acknowledged in this project, not repaired in the source or verified on the website. The finding remains in the report with its status. Retaining empty alt on the decorative divider can be an intentional review decision even though no edit follows. Conversely, a colleague may accept a task while the actual source remains untouched; do not describe that acceptance as a completed repair.

Any source, mode or base change clears review exceptions because the evidence has changed. Rerun analysis after preparing a revised snapshot and inspect its current findings. Keep the earlier saved project if you need the previous review state. The checker does not edit your CMS, publish the page or automatically import tasks into SMTasker. A clean checklist is not a release certificate: someone still needs to check the implemented page and the visitor journey outside this tool.

Choose an export that preserves the right evidence

Save JSON for a later session. It preserves project context, both raw sources when supplied, and primary review state. A blank-source draft can be saved, but inspection requires a nonempty primary source. This is an editable project, not an automatic website backup. There is no autosave or localStorage recovery. JSON imports are validated as a whole; arbitrary CMS exports or files containing unknown fields are not supported substitutes.

Use CSV for all extracted inventory from both sources plus the primary findings. Keep scope and base context alongside it so a resolved address is not mistaken for a successful visit. Copy checklist provides the primary review tasks for a ticket or email. Neither action repairs markup, follows destinations or turns a reviewed status into proof of a completed change. The comparison inventory does not acquire a separate checklist through export.

Choose PDF for scope, complete extracted records from both supplied sources, primary findings and statuses, and guidance. It is not a rendered website screenshot and does not include raw source. Keep JSON for that material. PDF export has a 100-page and 40 MiB ceiling; if either is exceeded, use JSON or CSV rather than expecting an incomplete PDF to represent the full inventory.

Close the workshop review with a small handoff package: the dated JSON, a readable report and the agreed editing tasks. Give one person responsibility for the next source snapshot so parallel copies do not become conflicting approvals. Once the website work is implemented, prepare the separate social announcements with accurate class and collection information. SMTasker can support eligible social routines, but this local checker does not transfer the draft into an automation or publish it anywhere.

Questions about checking content structure