Most technical SEO audits nutrient plentifulness of findings. Then the archive sits successful a shared thrust for six months, and thing gets deployed.
Sometimes that’s the client’s fault. More often, it’s the audit’s. The findings were ne'er validated, classed by a tool’s conception of severity, aliases written successful a measurement nary developer could enactment on.
Here are 10 mistakes that support showing up — and what to do instead.
1. Crawling without JavaScript execution enabled
Screaming Frog will show you some versions of the page successful a azygous crawl, arsenic agelong arsenic JavaScript rendering is connected and you’ve enabled storing some the original and rendered HTML.
The comparison shows you immoderate assemblage copy, soul links, canonical elements, aliases meta robots directives that beryllium successful the rendered DOM but not successful the first HTML response.
You tin besides spot the diff broadside by broadside successful the View Source tab.
Google renders astir pages without issue, but contented that only appears aft JavaScript runs is still little reliable. A blocked resource, a book error, aliases a timeout tin time off it retired of the scale entirely.
Most AI crawlers don’t execute JavaScript astatine all, truthful a page tin rank successful Google and still beryllium invisible to the systems generating AI answers.
If you find a gap, corroborate it pinch the URL Inspection instrumentality successful Search Console. That gives you Google’s ain position of the rendered page, which is harder for a developer to reason pinch than a screenshot from a third-party crawler.
Track, grow, and measurement your visibility crossed Google, AI search, social, local, and each transmission that influences buying decisions.
2. Ignoring the Page indexing study successful Search Console
You’ll find it nether Indexing > Pages. It’s the only study wherever Google tells you straight whether a URL is indexed, crawled but not indexed, discovered but not indexed, a soft 404, aliases thing else.
The Page indexing report looks for illustration this:
Not each URL successful the “Not indexed” bucket is simply a problem, which is wherever group spell incorrect pinch this report. Alternate page pinch due canonical tag, excluded by noindex tag, and page pinch redirect are each normal outcomes of a tract that’s group up correctly.
The exclusions worthy investigating are the ones you didn’t expect. Pages you want ranking that beryllium successful Crawled – presently not indexed, aliases a Discovered – presently not indexed count that keeps climbing.
3. Sampling URLs astatine random alternatively of by template
Pull URLs by page type alternatively than astatine random, truthful you’re covering merchandise pages, class pages, blog posts, filtered views, paginated series, and immoderate other the tract generates. Most method issues worthy reporting are template issues.
Get the canonical norm incorrect connected a merchandise template, and you’ve surgery it connected each 40,000 merchandise pages astatine once. If your sample includes 3 blog posts and a interaction page, you’ll miss that and study thing trivial instead.
Sampling by template besides makes the hole cheaper to scope. A developer tin estimate “change the canonical logic connected the PDP template” successful astir a minute. Nobody tin estimate a database of 40,000 URLs.
Dig deeper: Technical SEO testing: How to build a stronger experiment
4. Auditing from a azygous information source
Every instrumentality is unsighted to something. A crawling instrumentality only finds what’s linked aliases what you provender it, truthful orphaned pages enactment invisible unless you proviso them.
Search Console tells you Google’s verdict but not the logic down it. Analytics only records visits wherever the search codification runs, truthful crawler activity mostly doesn’t show up.
Server logs are the only root that shows each petition Googlebot aliases AI crawlers make to your server and what they get back. Rate limiting, intermittent 5xx errors, and crawl activity concentrated connected URLs you don’t attraction astir only move up here.
If you don’t person server logs, usage the Crawl Stats study successful Search Console. It’s sampled data, but you tin still usage the crawl petition breakdown to spot examples of URLs Google requested.
You don’t request each of them for each finding. But thing you’re astir to manus to a improvement squad should beryllium confirmed successful astatine slightest 2 places, and erstwhile 2 sources disagree, that disagreement is usually the much absorbing finding.
5. Treating instrumentality classifications arsenic facts
Crawlers study missing titles and H1s connected pages wherever the contented renders fine, and they log 429 and 503 position codes that the tract only returned because the crawl was moving excessively fast.
Before a uncovering goes successful the report, unfastened the page and cheque it yourself. To corroborate a position codification for illustration the illustration above, tally a curl command.
It takes a mates of minutes per finding, and it prevents a developer from spending half a time chasing a problem that was ne'er there. Developers who’ve been sent aft 1 phantom rumor thin to publication the remainder of your archive pinch suspicion.
6. Documenting symptoms alternatively of causes
“The tract has 12,000 copy URLs” is an observation, not a finding. The uncovering is immoderate produces them, which mightiness beryllium faceted navigation without parameter handling, convention IDs appended to URLs, aliases a CMS that generates a 2nd transcript of each page nether a different path.
A developer tin delete the 12,000 URLs successful an afternoon. They travel backmost the adjacent clip personification adds a select because thing astir the underlying behaviour changed. Tracing a copy backmost to its root takes longer than exporting the list, and it’s the portion of the occupation a instrumentality can’t do for you.
7. Prioritizing by instrumentality severity alternatively of business impact
A crawler assigns severity based connected the type of issue. It has nary thought which templates make revenue, which categories the business is pushing adjacent quarter, aliases which pages the income squad sends prospects to. So you get audits wherever tons of low-value warnings beryllium astatine the apical of the list, and a rendering nonaccomplishment connected the highest-margin merchandise template sits connected page four.
In the illustration below, it looks for illustration location are immoderate high-priority issues pinch Page Titles: Outside <head>. What Screaming Frog doesn’t cognize is that each those URLs travel from a template the squad is deleting successful the upcoming redesign.
Fixing this intends asking the customer questions the instrumentality can’t answer. What are the privilege products aliases services? Which pages convert? What’s launching this year? Then rank your validated findings against those answers alternatively of against a severity column.
Dig deeper: The biggest method SEO time-wasters to avoid
8. Recommending changes without knowing tract architecture
Redirects, canonical changes, URL removals, and noindex directives each person second-order effects. A noindex connected a filtered class yet cuts disconnected the soul links to the products underneath it. A batch of aged URLs redirected to the homepage will often extremity up classified arsenic soft 404s.
Before you urge immoderate of these, representation what links to the pages successful mobility and what they nexus to successful turn. Check whether they look successful navigation, sitemaps, aliases breadcrumbs. What you want to cognize is whether the page is the only way to thing else, and whether the pages it links to person different measurement in.
9. Writing recommendations developers can’t enactment on
“Improve tract speed” isn’t a recommendation. Neither is “fix canonicalization” nor “strengthen soul linking.”
A usable proposal includes the affected URLs aliases templates, the guidelines cause, the expected outcome, and capable item for personification to estimate the work. If a developer has to travel backmost and inquire what you really want, the summons goes to the bottommost of the backlog and stays there.
Compare “improve tract speed” to thing a developer tin prime up, for illustration “The LCP constituent connected the PDP template is simply a leader image loading done a lazy-load script, truthful it needs loading=”lazy” removed and fetchpriority=”high” added, pinch LCP nether 2.5 seconds.”
10. Prescribing the implementation alternatively of the outcome
Write the result and the constraints. The canonical connected paginated pages needs to beryllium self-referencing. Primary merchandise contented must beryllium included successful the first HTML response. Then fto the developer determine how.
You tin perfectly propose an attack if you person one, and connected smaller sites you mightiness beryllium right. But you don’t cognize the framework’s limitations, what other depends connected that component, aliases what the squad already has planned for that portion of the codebase.
Acceptance criteria springiness a developer thing to build against and thing to cheque their activity against erstwhile they’re done. A medicine conscionable invites a statement astir whether your attack is the correct one.
See wherever your marque appears, wherever it doesn’t, and precisely really to triumph much visibility crossed search, AI, local, social, and each transmission that matters.
What a bully audit looks like
A crawler produces a database of problems successful 10 minutes. Clients are paying for everything that happens aft that, erstwhile personification checks which of those problems are real, determines which ones matter to the business, and assigns a costs to each fix.
Contributing authors are invited to create contented for MCP and are chosen for their expertise and publication to the hunt community. Our contributors activity nether the oversight of the editorial staff and contributions are checked for value and relevance to our readers. MCP is owned by Semrush. Contributor was not asked to make immoderate nonstop aliases indirect mentions of Semrush. The opinions they definitive are their own.
About the Author
.png)
