SEO.In build

Templates rather than every post, because the failure is nearly always in the template.

Contrast, labels, headings, focus order, keyboard reachability, alt text.

The selector and the page, not a score out of a hundred.

Each failure with the change that resolves it. It publishes nothing.

Internal and external, including the ones inside page builder content.

A timeout is not a 404, and treating them the same creates noise.

For internal links, the page that replaced it. For external, the archived copy or nothing.

Only after you approve, and through post revisions so every edit can be undone.

Content, featured images, galleries, and page builder background images.

The actual file, not the attachment record, because the record can outlive the file.

A CDN outage, a failed migration and a deleted file need different answers.

Sorted by the traffic the page gets. It edits nothing.

Every reachable page, following the same rules a search engine does.

What is blocked, what is duplicated, what is orphaned, what is canonicalised to the wrong place.

Schema present, valid, and describing what the business actually is.

On the templates, on mobile, on the connection your customers actually have.

Ordered by what it is costing, not by severity in the abstract. It changes nothing.

Title, description, headings, alt text, schema, internal links, and what it currently ranks for.

Every one written out in full, so you read the actual words before anything happens.

It never publishes on its own. This step has no timeout and no default yes.

Through post revisions and the SEO plugin you already use, so nothing is bypassed and everything is undoable.

Confirm the change landed and nothing else on the page moved.

Every child sitemap it names, and whether the file is valid XML at all. A broken sitemap fails silently and nothing tells you.

Status code, canonical tag and meta robots for each. A sitemap full of redirects and 404s spends the crawl budget on your worst pages.

Pages the site links to that the sitemap leaves out, and pages in the sitemap that nothing on the site links to.

A last modified date that changes on every page every night tells a search engine nothing, which is the same as having no date at all.

What is missing and what should not be there. This is a sitemap audit, not an indexation report. Only Search Console can say what Google actually indexed. It changes nothing.

Nothing outside Search Console can tell you what Google has indexed. Without it the run exits here rather than reading a sitemap and calling it indexation.

Every page Google knows about, the state it assigned, and the date it last crawled it.

Pages you publish that Google has not indexed, and pages Google has indexed that you never meant to publish.

Crawled and not indexed, duplicate without a canonical, blocked by robots, and soft 404. Each of those needs a different answer.

A dropped service page matters. A dropped tag archive does not, and mixing them buries the one that does.

Each group with the change that resolves it. It requests no indexing and changes nothing on the site.

The tag in the head, the one in the HTTP header, and the one the sitemap implies. Three sources that can disagree with each other.

A page whose canonical names another address is asking not to rank. Sometimes that is correct, and often it is a plugin default nobody chose.

A canonical pointing at a redirect, a 404, or a page that canonicals somewhere else again is worse than having none.

http against https, www against the plain address, and a trailing slash that does not match. Each one splits a single page into two.

Ordered by what each page earns rather than by how many pages are affected. It edits nothing.

From the link to the final address, counting the hops on the way. If nothing chains, it exits here and edits no rule.

Every rule as it stands today, from whichever plugin or host layer holds it. Nothing is rewritten until that export exists.

One rule at a time, and only where the chain ends in a page that answers. A chain ending in a 404 is a different problem and is reported instead.

Two rules pointing at each other cannot be flattened safely. They are listed for you and nothing is touched.

The step two export puts every rule back exactly as it was, in one action, with no rule left half rewritten.

Confirm each one now reaches its page in a single hop, and that nothing which worked before this run now returns a 404.

From the page as it is served, not from the SEO plugin's fields, because a template can override what the field says. If nothing is duplicated, it exits here.

Twelve location pages differing by one town name are the same finding as twelve identical ones.

Every proposed title and description in the words that would be published. It waits here, with no timeout and no default yes.

Through the SEO plugin already installed and through post revisions, so nothing is bypassed and nothing is written twice.

Each change is one field in the SEO plugin, with the previous value held in the post revision. Putting a page back is one action.

Request each changed page again and confirm the new title is what actually comes back, not just what is stored.

In document order, with the level of each one recorded.

A page builder styling text to look like a heading is not a heading. It reads the tag, not the size.

An h2 followed by an h4 tells a screen reader and a search engine that a section is missing.

A page whose h1 says Home and whose title names the service is describing itself twice, differently.

Because one theme file is usually behind two hundred pages. It changes nothing.

Every internal link on every page, including the ones inside page builder content and navigation menus.

The shortest path in clicks, following only the links a crawler is able to follow.

Pages with a single internal link, and pages only the sitemap knows about.

A service page four clicks deep is the finding. A 2019 blog post four clicks deep is fine where it is.

Which existing page is about the same subject and has the traffic to pass on.

A list of links to add, page by page. It adds none of them.

Every public post, page and product read through the connector, with its address and its date.

Every page that can be reached by following links from the homepage.

What is left is orphaned. Nothing links to it, so nothing tells a search engine it matters.

An orphaned page with traffic needs a link. An orphaned page from a theme demo import needs deleting, and it says which is which.

Which page should link to it and where in that page. It publishes no link and deletes no page.

From every template, whether it sits in a script tag, in microdata, or is spread across both.

Against the published schema.org types and against Google's own required and recommended fields.

Product schema on a site that sells nothing, an organisation with no address, a review with no reviewer. Each one is either ignored or held against you.

The types this business should have and does not. A local contractor with no LocalBusiness and no opening hours is invisible in the map results.

The SEO plugin and the theme both writing an organisation block, with different names, is worse than neither of them doing it.

With the exact field to add or remove. It edits nothing.

Which translation plugin is active and which languages it serves. On a single language site it exits here.

From the head, the HTTP header and the sitemap, on every page in every language.

If the English page names the German one and the German one does not name it back, Google ignores both of them.

en-UK is not a language code. en-GB is. This mistake is common and completely silent.

One x-default, pointing at a page that answers with a 200 and does not redirect.

Which direction is broken and on which pages. It changes nothing.

The robots.txt file, the WordPress discourage search engines setting, and the SEO plugin's own indexing rules. If all three agree and none of them blocks, it exits here.

A robots.txt blocking a folder the SEO plugin is asking to have indexed is a conflict, and the crawler settles it against you.

The exact file contents and every setting, written down before anything changes.

The discourage setting, and any disallow rule covering pages you want found. One at a time, and only the ones you approve.

The file and the settings return to what step three recorded, in one action. Nothing else in robots.txt is rewritten at any point.

Request robots.txt and the affected pages again, and confirm the file being served is the file that was written and the meta robots tag now agrees with it.

With its alt attribute as served and the page it sits on. If coverage is already complete, it exits here and writes nothing.

A background flourish should carry an empty alt. A product photo should not. Treating both the same is why automatic alt text is usually wrong.

Every alt string in the words that would be saved, with the image next to it. It waits here and publishes nothing on its own.

Into the media library record and into the post content, through post revisions.

Any string you reject afterwards is cleared from the media record, and the post revision returns the content exactly as it was.

Request the changed pages again and report the share of meaningful images carrying alt text, before and after.

Whether the theme or the SEO plugin already outputs a breadcrumb list, and whether it validates. If a valid one is present, it exits here.

The SEO plugin's breadcrumb configuration and any theme option, before either one changes.

Through the SEO plugin already installed, so that plugin stays the single place breadcrumbs are configured.

The path a visitor would actually take, not the folder structure of the address, because those two are often different.

One setting returns to what step two recorded and the markup goes with it. No template file is edited at any point.

Fetch a post, a service page and an archive, and confirm the breadcrumb list parses and names the right parents.

Blog, categories, tags, author archives and any product listing, with how many pages each one runs to.

Page two through to the last page, checking each one answers and each one links to the next.

A page two canonicalised to page one is asking to be dropped, and everything only reachable from page two goes with it.

Every page in a series sharing one title is a duplicate finding that has nothing to do with the content on them.

An archive that keeps serving pages past the last real one hands a crawler an unlimited number of empty pages to find.

Which archive, which page, and what to change. It edits nothing.

The content area only, with navigation, sidebar and footer taken out, because those inflate every word count by the same amount.

Pages under the threshold you set, with the template that produced them named.

Near duplicates, where only a town name or a product code differs. Twenty location pages built from one paragraph compete with each other and none of them wins.

A thin page with steady traffic is not a problem. A thin page with none, duplicating a page that has some, is.

Expand it, merge it into the stronger page, or remove it and redirect. One decision per page, never a bulk instruction.

Grouped by template, because the fix is usually one template rather than two hundred pages. It changes nothing.