Open SEO Skill.

What the audit checks

Five areas, assessed across your whole site rather than one page. Below is every individual check, and more usefully, why each one matters and what a failure actually costs you.

How it decides what to look at

Before checking anything, it maps your site. It reads your homepage and pulls every internal link out of the navigation, header and footer, then fetches robots.txt and sitemap.xml to catch pages the navigation doesn't mention.

From that map it picks six to ten pages by signal: your About page, your main service or product pages, a couple of individual ones, a blog index and some real posts, contact, and any FAQ. It skips privacy policies, terms, carts and thank-you pages, which tell you nothing.

This matters more than it sounds. An audit that only reads your homepage will confidently tell you to add a Team page you already have at /about-us. Checking first is the difference between advice and noise.

1. Findability and technical basics

Whether a search engine can reach your pages, understand what each one is about, and show it properly in results. Most sites lose more here than anywhere else, and it's the cheapest area to fix.

CheckWhat it looks for and why
Title tagPresent, unique across pages, roughly 50–60 characters, main topic near the front. Too long and Google cuts it off mid-word. Duplicated across pages and Google can't tell them apart. The audit quotes your actual title with its character count.
Meta descriptionPresent, unique, roughly 150–160 characters. It doesn't affect ranking, but it's the ad copy under your listing. Leave it blank and Google writes its own from whatever text it finds.
H1 headingExactly one per page, matching what the page is about. Several H1s or none at all muddies the page's main subject.
Heading orderH2s and H3s in sequence without skipping levels. This is how both search engines and screen readers follow the structure of a page.
Canonical tagPresent and pointing at itself, unless the page is a genuine duplicate. Missing canonicals are how the same content ends up competing with itself on several URLs.
Robots directivesAny accidental noindex in the page or Disallow in robots.txt blocking pages you want found. A leftover staging block is the single most expensive SEO mistake there is.
URL structureLowercase, hyphenated, readable, no ID numbers or query strings. Also whether the structure reflects how the site is organised.
Internal linksWhether important pages are linked from somewhere other than the sitemap, and whether the link text describes the destination. Pages reachable only from the sitemap barely rank.
Image alt textDescriptive text on meaningful images. Required for accessibility, and it's how images get found in image search.
Social tagsOpen Graph and Twitter card tags, which control the title, description and picture when someone shares your link.
HTTPSSecure, with no HTTP resources mixed in. Mixed content triggers browser warnings.
Sitemap healthWhether it exists, is referenced from robots.txt, and lists only real pages that return a 200 and are meant to be indexed.

2. Content and trust

Whether a person landing on the page gets what they came for, and whether there's enough evidence behind it to believe. Google frames this as three questions: who made this, how was it made, and why does it exist.

Of Google's four quality factors — experience, expertise, authoritativeness and trustworthiness — trust carries the most weight. The audit weights it accordingly rather than treating all four equally.

3. Structured data

Schema markup is a block of machine-readable code that tells search engines what a page is: a product with a price, a business with opening hours, an article with an author. It doesn't change your content, only how clearly it's labelled.

The audit reads every piece of structured data on your pages, lists the types it finds, checks the required properties are actually filled in, and flags anywhere the markup claims something the page doesn't show. It then names the types you should have and don't.

It won't recommend retired types

Several schema types stopped producing any visible result in Google and are still recommended constantly by older guides and other audit tools. HowTo and FAQPage are the big two. The audit works from a dated list and skips them. See the current facts.

Where something is missing, it writes the complete JSON-LD block for you rather than telling you to add one.

4. AI search readiness

Whether your pages can be picked up and quoted by AI Overviews, ChatGPT, Perplexity and similar. This is sometimes sold as a separate discipline with its own acronyms. It mostly isn't.

AI answers are grounded in ordinary search results. A page has to be indexed and eligible to appear as a snippet before it can be quoted in any AI answer at all. So the foundation is everything in section 1, and the extra work is about making your content easy to lift out and cite.

  • Self-contained passages. Whether each key question gets a complete answer of roughly 40 to 170 words that still makes sense pulled out of the page.
  • Question-shaped headings where the content beneath genuinely answers that question.
  • Attribution. Whether you cite sources, name people, and reference checkable facts. Pages that cite get cited.
  • Entity consistency. Whether your business is named the same way everywhere, with Organization or LocalBusiness markup and links to your real social profiles.
  • Extractable formats. Comparison tables and plain definitions, which get pulled into answers far more often than prose.

What it deliberately won't tell you to do

Add an llms.txt file. Rewrite your content with "AI keywords". Chop your pages into fixed-length chunks. Add AI-specific schema. None of these are supported by how AI search actually retrieves pages, and the audit won't recommend them no matter how often you've seen them suggested.

5. Local

Only runs if the business serves a physical place. For a plumber, a clinic or a café this section usually matters more than everything above it.

And then it writes the fixes

The checks above produce findings. Findings on their own don't change anything.

Once the report is done, the audit sorts everything into three groups: what you can change yourself today, what needs your web developer, and what needs more data or a specialist. Then it writes out the highest-impact fix as finished text and asks which platform you're on so it can tell you exactly where that goes.