The SEO audit that shows you exactly what is broken

The Largest Contentful Paint threshold Google publishes for a good page experience, alongside Interaction to Next Paint at 200 milliseconds or less and Cumulative Layout Shift at 0.1 or less. All three are judged at the 75th percentile of real page loads, so one quick test on your own laptop proves nothing.

The ceiling on a single sitemap file: 50,000 URLs or 50MB uncompressed, whichever you hit first. Past that you need a sitemap index. A sitemap.xml quietly over the limit is one of the faster findings in a technical SEO audit.

Domains carrying schema.org Organization markup, counted by Google against its own web index in July 2026. Structured data stopped being an edge case a long time ago. It is the floor your competitors already stand on.

Crawling and indexing

• robots.txt, read line by line

  What it blocks, and what it blocks by accident. Google’s own documentation says robots.txt is not a mechanism for keeping a page out of the index, so a disallowed URL can still surface with no snippet under it.

• sitemap.xml and the sitemap index

  Whether it exists, whether it lists only live 200 URLs, whether it is submitted, and whether it has crept past the size limit and stopped being read at all.

• Index coverage in Google Search Console

  The pages Google found, the pages it skipped, and the reason it gives for each one. On most sites this is where the fastest wins are sitting.

• Canonical tags

  Every page needs a self-referencing canonical tag or a deliberate one pointing elsewhere. Conflicting canonicals split your signals across URLs that should have been one.

• Redirect chains and the redirect map

  Every 301 redirect should be a single hop. We trace the chains, flag the loops, and write a redirect map for anything retired. Bots give up on chains faster than browsers do.

• Crawl budget on large sites

  Where Googlebot actually spends its time. Faceted URLs, filter parameters, and near-duplicate pages eat the crawl budget you wanted spent on pages that convert.

• hreflang and country targeting

  If you run US, UK, or other regional versions we check the hreflang cluster returns properly and that x-default is set. Broken hreflang quietly serves the wrong page to the wrong country.

• Rendering and hydration

  We compare raw HTML against the rendered DOM on React, Next.js, Vue, and Nuxt builds. Content that only appears after hydration is content Google may never weigh.

Speed and page experience

• Largest Contentful Paint

  Which element is largest, when it paints, and what is holding it up. Usually an unoptimized hero image or a font that arrives late.

• Interaction to Next Paint

  How long the page takes to answer a real tap. Heavy client-side JavaScript is the usual cause, and it rarely shows up on a desktop test.

• Cumulative Layout Shift

  Images without reserved space, late banners, and fonts that swap. Cheap to fix, and it irritates buyers more than it costs you rankings.

• Lab scores measured against field data

  A Lighthouse run tells you what is possible. Field data tells you what your visitors got. We report both, and never quote only the flattering one.

• How the page is delivered

  Whether static site generation or server-side rendering would help, whether the CDN is doing edge caching on HTML or only on assets, and what is blocking first paint.

On-page

• Title tags

  Length, uniqueness, and whether the page targets a phrase anyone actually searches. Rewritten titles are often the first measurable lift after an audit.

• Meta descriptions

  Not a ranking factor, still a click-through factor. Duplicates, truncations, and the pages where Google is rewriting yours because it is unhelpful.

• Heading structure

  Exactly one h1 and a clean descent to h2 and h3. We check the tag the browser receives, not the component name, because heading wrappers get this silently wrong.

• Keyword cannibalization

  Two pages chasing one phrase beat each other instead of the competition. We find the overlaps and say plainly which page should win and which should be merged.

• Internal linking

  Orphan pages, dead ends, and the authority your strongest pages are sitting on instead of passing down. Internal linking costs nothing and most sites still waste it.

• Images, alt text, and URLs

  Missing alt attributes, decorative images that should carry empty alt, files served far heavier than needed, and URLs only worth changing when the gain beats the redirect.

Content

• Thin and near-duplicate pages

  Cloned service or location pages get filtered out of the index. Content pruning almost always comes before writing anything new.

• Topical gaps against who outranks you

  What the pages above you cover that yours do not. Topical authority is built by covering a subject properly, not by publishing more often.

• Answer-first structure

  Whether each page states its answer inside the first screen. AI answer engines pull disproportionately from the top third of a page, so burying the answer costs you twice.

• Freshness and author signals

  Whether dateModified in your structured data is honest, whether anything time-sensitive has quietly expired, and whether a named human stands behind the advice.

Structured data and entities

• JSON-LD, not Microdata

  Google recommends JSON-LD for structured data in its own documentation, so that is the format we check for and the format we write.

• Schema that matches the page

  The right schema.org types for what the page genuinely is, validated before it ships instead of after somebody notices a rich result disappeared.

• Schema drift

  FAQ schema markup saying one thing while the visible FAQ says another is the most common silent defect we find. To an answer engine it reads as cloaking.

• Entity consistency

  Your Organization markup, your sameAs profiles, and whether your business name is written the same way everywhere Google already trusts.

Links and authority

• Referring domains, not link counts

  How many real sites link to you, and how many of those are farms. A thousand links from forty domains is forty links.

• Toxic and bought links

  Anything a previous agency purchased, and whether it needs disavowing or simply ignoring. Most of the time the honest answer is ignoring.

• Broken inbound links

  Links pointing at URLs you retired years ago. A single-hop 301 redirect recovers authority you already earned and then forgot about.

Local visibility

• Google Business Profile and NAP

  Primary category, services, hours, photos, and whether name, address, and phone match across the directories that feed Google. For a local business this outranks a lot of the website.

• Location pages

  One genuinely different page per location, carrying LocalBusiness structured data and internal linking from the service pages that already hold weight.

AI visibility

• Whether the assistants name you

  We put the questions your buyers actually type into ChatGPT, Claude, Gemini, and Perplexity, and record whether you appear. It is measurable now, so we measure it.

• AI crawler access in robots.txt

  GPTBot, ClaudeBot, PerplexityBot, and Google-Extended each need naming. One over-broad wildcard that catches a retrieval bot removes you from AI answers with no error anywhere to notice.

• Extractable blocks and llms.txt

  Short definitions, real tables, honest lists, and a plain markdown file at your site root. The formats an answer engine can lift cleanly are the ones that get cited.

Measurement

• GA4 events that match reality

  Whether form submits, calls, and checkouts land in Google Analytics 4 as named events. We have opened audits where the tracking broke months before the traffic did.

• Google Tag Manager and server-side tracking

  What is firing, what is firing twice, and whether moving to server-side tracking with consent mode would protect the first-party data you still have.

• A year of Search Console history

  Twelve months of queries and pages. On a large site we push the bulk export into BigQuery and read it in Looker Studio, because the interface hides the long tail.