I have opinions about SEO audits. Specifically: most of them are too long, most of them aren't prioritized, and most of them include a huge pile of warnings from Screaming Frog that the client then spends six months trying to clear, none of which actually moves traffic.
This is roughly the list we run when we onboard a new enterprise client. It's in order of what tends to matter most. We don't always check every item — sometimes we're 3 bullets in and realize the problem is somewhere else entirely. But this is the default starting point.
Can Google get to your pages?
Seems basic. Is not. Every couple of months we onboard someone whose robots.txt is silently blocking a high-value directory because someone added a rule in 2021 and nobody remembers why.
- robots.txt reviewed line by line. If you don't know why a Disallow is there, delete it (after backing it up).
- XML sitemap exists, returns 200, and matches the URLs you actually want indexed. Sitemaps listing redirected or noindexed URLs are surprisingly common.
- Orphan pages — anything indexable but unlinked from the rest of the site. Usually old campaign pages or legacy blog posts.
- Click depth: your money pages should be reachable within 3 clicks of the homepage. Past that, Google deprioritizes them.
Is it getting indexed, and are the right URLs being treated as canonical?
This is where most enterprise sites lose the plot. Faceted nav, session IDs, tracking parameters, internationalization — any one of these done wrong can bloat Google's crawl budget and leave your real pages under-crawled.
- Canonical tags are self-referencing on the URL you want indexed. Not a redirect chain away.
- Faceted / filter URLs are canonicalized or noindexed — one of the two, consistently.
- GSC Pages report: indexed count vs sitemap count should be close. A big gap is a signal to dig.
- No accidental noindex on important pages. Check templates, not just samples. We've seen a single misplaced noindex tag knock an entire category off the index.
Core Web Vitals
Still a ranking factor in 2026. More importantly, slow sites convert worse — so even when rankings don't budge, you usually see a CVR lift from fixing these. Use CrUX data (real users) not Lighthouse lab scores.
- LCP under 2.5s at the 75th percentile on mobile. Hero images are the usual culprit.
- CLS under 0.1. Watch for late-loading ads, embedded videos without dimensions, and web fonts swapping.
- INP under 200ms. This one is harder than the old FID. Usually it's third-party JS — tag managers, chat widgets, session replay tools all stacked up.
A fun one we hit last year: a client's chat widget was adding 1.8 seconds of INP on every page. It had been there for two years. Nobody had connected it to traffic decline. We disabled it, traffic came back in six weeks.
Structured data
Easy to implement, easy to get wrong, genuinely does help with rich results and (we suspect, though Google is cagey) with general understanding of your pages.
- Organization schema on the homepage.
- BreadcrumbList on every inner page.
- Product / Article / FAQPage schema where applicable.
- Validate in the Rich Results Test, not just Schema.org. Google is stricter than the spec.
What we don't audit (on purpose)
Meta description pixel length. Title tag tweaks for "power words." Keyword density. LSI keywords, whatever those are supposed to be in 2026. Most content plugins' "SEO scores." These are all either noise or actively outdated.
If you're getting an audit back that spends more time on those than on the four sections above, ask for your money back.