Typographic errors in content

A word repeated immediately — the the, of of — which survives proofreading because the eye skips it.

Doubled punctuation, where a full stop or a comma appears twice in succession.

A missing space after a sentence-ending mark, which usually comes from a paste or a bad line join.

A space before a punctuation mark, which is correct in some languages and not in the one the page declares.

A block of text in capitals, which reads as shouting and is announced letter by letter by some screen readers.

Spelling requires a dictionary per language and produces false positives on names, products and technical terms.

Grammar requires understanding, and a checker that misjudges it generates noise that teaches people to ignore the report.

The defects above are structural rather than semantic: they are wrong regardless of what the sentence means.

Which makes them safe to report automatically, and it is why the check stops there rather than attempting more.

A human reading the page remains the only way to catch what these cannot.

Content pasted from another application, which brings its own spacing and punctuation conventions.

A translation pipeline, where the source language's punctuation spacing survives into the target.

An import from a legacy system, which is where whole-page capitals usually originate.

Editing without re-reading, which produces the repeated-word case almost exclusively.

Automated content assembly from fragments, where the join between two fragments loses a space.

A block set in capitals in the source text cannot be un-shouted by anyone reading it.

Some screen readers announce a capitalised word letter by letter, treating it as an initialism.

It is also harder to read for everyone, because word shapes are what fast reading relies on and capitals remove them.

Where capitals are a design choice, the text transform belongs in the stylesheet so the source stays readable and the presentation is a decision.

That change costs nothing and fixes the screen reader behaviour at the same time.

Fix the instances, which is quick — these are one-character edits.

Find the source: a repeated pattern across many pages points at an import or a template rather than at an author.

Move presentational capitals into the stylesheet and restore the source text.

Add the check to whatever review a page passes before publishing, since it costs nothing and catches them before a reader does.

Re-check after any bulk content operation, which is where they arrive in quantity.

Read the page aloud, which catches repeated words better than reading silently.

Search the rendered text for a doubled space, a doubled full stop and a space preceding a comma.

Look for paragraphs where every letter is capital in the source rather than in the styling.

Check the pages produced by an import separately, since they cluster there.

Check translated pages against the punctuation conventions of the language they declare.

`structure.html_quality` flags the unambiguous defects in the stored markup sample and emits the corresponding issue codes.

`structure.duplicate_content` and `seo.title` cover the adjacent content findings — duplication and titles rather than typography.

`seo.lang` reports the declared language, which is what decides whether a spacing convention is correct.

`accessibility.contrast` and `mobile.text_size` cover readability from the presentation side rather than the text side.

This is not a full grammar or spelling check, and the report says so: only unambiguous structural defects are reported.

The same mechanical detection finds placeholder strings left in published pages — dummy paragraphs, a heading still reading as a to-do, a default sample title.

Those are the same failure at a larger scale: a page published without being read, where the evidence is a whole block rather than a character.

They cluster on the pages nobody visits, which is why they survive for years and are found by a crawler rather than by a person.

A repeated placeholder across many pages usually means a template default that was never replaced, rather than many separate omissions.

Which is the same diagnostic as the typography: a pattern across pages points at a template, and a scattered one points at editing.