A Technical SEO Audit Checklist You Can Actually Run

Most SEO audits are a list of everything a tool flagged, sorted by the tool's own idea of severity. That is not an audit, it is a printout. What follows is the order we actually work in on client sites, from the checks that decide whether a page can rank at all down to the ones that decide how well. Each section starts with the answer, so you can act on it without reading the rest.

Start with what Google already told you

Open the Page Indexing report in Search Console before you open any third-party crawler. Google is telling you, for free and about your specific site, which pages it took and which it refused. Third-party tools guess at this; Search Console knows. Sort by the reason column and work down from the largest bucket.

• Discovered, currently not indexed means Google knows the URL exists but has not spent crawl budget on it. Usually a sign of weak internal linking or a site with no authority yet.
• Crawled, currently not indexed means Google fetched the page, read it, and chose not to keep it. This is a quality and priority judgement, not a technical fault.
• Duplicate, Google chose a different canonical means two of your pages look like the same page to a machine. Merge them or differentiate them.
• Page with redirect and Not found (404) in bulk usually mean an old URL structure is still being linked to from somewhere.

What to do about "crawled, currently not indexed"

Nothing technical, because nothing technical is wrong. If the inspection tool says crawl allowed, fetch successful, indexing allowed and the canonical matches what you declared, then Googlebot read the page and decided it was not worth an index slot. Requesting indexing again does not overrule that; it only asks for another look, and the answer will be the same until the inputs change.

Three inputs actually change it. The page has to say something the index does not already have, which on a service page usually means real prices, real constraints and real examples rather than the same paragraph every agency publishes. It has to be reachable by more than one link, ideally from articles that are about its topic. And the domain has to have some history, because a site with no crawl record and no external links is treated cautiously by design. On a new domain, secondary pages sitting in this state for several weeks while the homepage indexes is the normal shape of the problem, not a bug.

Content cannot rank if it cannot be crawled. But being crawlable was never the same as being worth indexing.

Crawling: can the bot reach the page

Check these first because everything downstream is wasted if they fail. Each takes under a minute.

• robots.txt does not block anything that should rank, and does not block /_next/, /static/ or your CSS and JS. Blocking assets makes Google render a broken page.
• No stray noindex in a meta tag or an X-Robots-Tag header. Check the header, not just the HTML, because a CDN or host can add one you never wrote.
• Every important page is reachable by a plain <a href> from somewhere. Links that only exist inside a click handler are invisible to a crawler.
• The XML sitemap contains only canonical, indexable, 200-status URLs. A sitemap full of redirects and 404s trains Google to trust it less.
• Redirect chains are collapsed to a single hop, and no canonical points at a redirect.

If you are moving a site or changing its URLs, this is where most traffic is lost, and the failure is nearly always redirects rather than content. We wrote that up separately in site migration without losing traffic.

Indexing: will Google keep the page

Assume a page competes for an index slot against everything already ranking for its topic. The questions worth asking are unglamorous. Does this page have a job no other page on the site does? If two pages answer the same query, one of them is dead weight and should be merged into the other with a 301. Is the canonical self-referencing and consistent with what is in the sitemap and what internal links point to? Do parameters, uppercase variants and trailing slashes all resolve to one URL?

Thin near-duplicate pages are the most common cause of a whole section failing to index. Twelve service pages built from one template with the nouns swapped read to a classifier as one page repeated twelve times. The fix is not more pages; it is fewer pages that each say something specific.

Keyword targeting: one page, one job

Pick the query a page is for before you write it, and pick it by intent rather than volume. A keyword with 200 monthly searches and clear buying intent is worth more than one with 20,000 and none. In practice we sort candidates into three buckets and treat them differently.

• Transactional queries name a service and often a place, like "web development company in Lalitpur". These map to service pages, not blog posts.
• Commercial investigation queries compare options, like "Next.js vs WordPress" or "React Native vs Flutter". These map to long articles that genuinely compare, including the cases where your own service is the wrong answer.
• Informational queries ask how something works. These map to guides, and they earn links and trust rather than direct enquiries.

Check the current results before committing. If page one is all national news sites, a five-page local business is not going to displace them and the effort is better spent elsewhere. If page one is thin, outdated or entirely foreign to your market, that is an opening. Reading the results honestly is more useful than any difficulty score a tool prints.

On-page: the parts that still decide most rankings

On-page SEO has changed less than the industry pretends, and most pages still skip half of it. The title tag should read like the answer to the query, front-loaded, under about 60 characters, and different from every other title on the site. One H1 per page, matching the promise in the title. H2s that are phrased the way a person would ask the question, with the answer in the first sentence underneath, which is also what gets a passage pulled into an AI overview.

• A meta description that earns the click. It is not a ranking factor, it is ad copy.
• URLs that are short, lowercase, hyphenated and stable. Renaming URLs later costs traffic.
• Alt text that describes the image for someone who cannot see it, not a keyword dump.
• Content that answers the query in the first screen rather than building to it after four paragraphs of preamble.
• Depth without padding. Cover the topic fully, then stop. Length helps only when every paragraph earns its place.

Internal links: the lever most sites never pull

Internal linking is the highest-leverage work available to a site with no external authority, and it is entirely within your control. If a page's only inbound link is one item in a homepage grid, you have told Google it is a minor page. If four articles link into it with descriptive anchor text because it is genuinely relevant to what those articles discuss, you have said something different.

Three rules cover most of it. Use anchor text that describes the destination, so "our SEO services in Nepal" rather than "click here". Link from the body of the text, where it is contextual, not only from a related-posts strip at the bottom. And build clusters, where a pillar page covers the topic broadly and the deeper articles link back to it and to each other.

External links: a few real ones beat a hundred bought ones

For a new domain, a handful of genuine links moves the needle more than any on-page change. The realistic sources for a small company are not glamorous and that is fine: your own company profiles on LinkedIn and the relevant national directories, a client or partner who will link to the work you did for them, a local business association, a supplier's customer page, a genuinely useful piece of data or research that someone in your sector wants to cite.

Skip anything sold by volume. Link packages, private blog networks and comment-spam services are cheap because they are worthless at best, and at worst they attach your domain to a pattern Google already has a classifier for. There is no version of this that is a shortcut.

Structured data: help the machine describe you

Add schema for the content types you actually have, and only those. Organization and LocalBusiness with a consistent name, address and phone. Service on service pages. Article with an author and a date on posts. FAQPage where you have real questions with real answers, BreadcrumbList so the path shows in results, and Product or Offer if you sell something with a price. Validate it in the Rich Results Test, and never mark up content that is not visible on the page.

Speed and Core Web Vitals

Field data from real users is the number Google uses, not a lab score from your laptop. On Nepali mobile connections the gap between the two is large. The three metrics and their usual causes are covered in Core Web Vitals: real gains from LCP, CLS and INP; the short version is that server response time and an oversized hero image account for most LCP failures, missing image dimensions account for most CLS, and a heavy JavaScript bundle accounts for most INP.

The order to run this in

Working out of order is how audits turn into six-month projects with nothing to show. This sequence gets you a result at every step.

• Week one: fix crawl blockers, redirect chains, the sitemap, and any duplicate canonicals. These are cheap and unblock everything else.
• Week two: pick one target query per page and rewrite titles, H1s and the opening paragraph of your ten highest-intent pages to match.
• Week three: build the internal link mesh. Every service page should have at least three contextual links from articles about its topic.
• Week four: merge or remove thin pages that compete with each other, with 301s to the page that survives.
• Ongoing: earn a small number of real external links, publish depth rather than volume, and re-inspect in Search Console every three to four weeks rather than resubmitting daily.

If you would rather have this run for you against your own numbers, that is what our SEO service does, and local SEO for businesses in Nepal covers the map-pack side that a technical audit does not touch.