I did this audit four weeks after launch. Every item on it was broken. Here it is, in the order I'd run it, with the specific check for each.
1. Trailing slash consistency (5 minutes)
/page and /page/ must resolve to one URL, not two. If both 200, you have duplicate content.
curl -sI https://yoursite.com/page | grep -i "^location\|^HTTP"
curl -sI https://yoursite.com/page/ | grep -i "^location\|^HTTP"
In Next.js this is trailingSlash in next.config. Pick one, redirect the other. I'd also check: does the sitemap match what you chose?
2. The www and protocol question
Four URLs for every page if you're not careful:
http://yoursite.com
https://yoursite.com
http://www.yoursite.com
https://www.yoursite.com
Three should 301 to one. Check with:
for u in http://yoursite.com http://www.yoursite.com https://www.yoursite.com; do
echo -n "$u -> "; curl -sI "$u" | grep -i "^location" || echo "no redirect"
done
3. Does your 404 actually 404?
curl -sI https://yoursite.com/this-page-does-not-exist-12345 | head -1
If that returns 200, you have a soft 404 — a catch-all route or a "not found" page rendering with a success status. Google treats those as thin content, not as missing pages, and they can hurt the whole site.
4. robots.txt
Two questions: is it blocking something you want indexed, and does it point at your sitemap?
curl -s https://yoursite.com/robots.txt
Common self-inflicted wound: a staging Disallow: / left in the production deploy. Check it after every deploy, not once.
5. Sitemap: URL count vs. actual pages
curl -s https://yoursite.com/sitemap.xml | grep -c "<loc>"
Then count how many URLs you actually have. If sitemap says 113 and your site has 90 pages, 23 entries are stale — Google will waste crawl budget on redirects and 404s.
Also: are all the locale variants present? If you have four languages and the sitemap only lists one, you're invisible in three markets.
6. Orphan pages (the one nobody checks)
Crawl from the homepage, build a graph, compare against the sitemap. Anything in the sitemap but not reachable is orphaned.
Mine:
crawled 102 / reachable 102 / orphans 12
Twelve pages with zero internal links pointing at them. They existed, were in the sitemap, and had no link equity whatsoever. Fixing it was adding three index cards.
If you skip every other item on this list, do this one.
7. hreflang: reciprocal, self-referencing, and correct
Every page in a group must:
- Reference all the others
- Reference itself
- Use absolute URLs
<link rel="alternate" hreflang="ja" href="https://site.com/ja/page" />
<link rel="alternate" hreflang="fr" href="https://site.com/fr/page" />
<link rel="alternate" hreflang="x-default" href="https://site.com/en/page" />
If A points to B but B doesn't point back to A, both are ignored. This is easy to get wrong when the map is hand-maintained — I had one, and it silently pointed at 404s for weeks.
Generate hreflang from data, not by hand.
8. Is the canonical tag self-referencing?
Every indexable page should have <link rel="canonical" href="that same page">. Missing canonicals are survivable. Wrong ones are not — a canonical pointing at the homepage deletes the page from the index.
curl -s https://yoursite.com/some-page | grep -o '<link rel="canonical"[^>]*>'
9. Structured data: does it match the visible page?
Test your JSON-LD in the Rich Results Test. Then go further and check the harder question: does the markup describe something a user can actually see?
FAQPage with questions that aren't rendered. Article with a datePublished that's a lie. Product with an aggregateRating from no ratings. All of those are at best ignored and at worst penalised.
One pattern that prevents drift — generate both from one source:
{faq?.length > 0 && <FaqJsonLd items={faq} />}
{faq?.map((f) => <details key={f.q}><summary>{f.q}</summary><p>{f.a}</p></details>)}
10. Does the page render without JavaScript?
curl -s https://yoursite.com/page | grep -c "meaningful content string"
For a Next.js SSG site this should be fine. For a client-rendered page it won't be, and Google will eventually index the empty shell.
I checked this and found my pricing page needed no JS to show prices, but the FAQ blocks were behind a <details> element — collapsed by default. Visible in the DOM, arguably not "visible" to a user. I left it, because <details> content is indexed, but it's worth knowing the distinction.
11. Real 404 handling, not just a status code
Fetch a few URLs that should be 404 (deleted pages, typo'd slugs) and check:
- Status is 404, not 200
- The page tells the user what happened
- The 404 isn't
noindex(it should be allowed to be 404 and stay 404) - There's a link home
12. Check the deployed env variables
This one isn't SEO, but it fails the same way: silently.
# In production, is the config what you think?
curl -s https://yoursite.com/api/health 2>/dev/null || echo "no health endpoint — add one"
I once had a PAYWALL_ENABLED set differently between local and production and only noticed because a user paid for something that was free on the deployed build.
What all of these have in common
None of them show up as an error. The site works, pages load, nothing crashes. The damage is invisible and it compounds — a soft 404 here, an orphan page there, an hreflang pointing at nothing.
That's why the audit has to be a checklist and not a feeling. You cannot tell by using the site.
Time budget: 45 minutes if you're systematic, and it's the highest-leverage 45 minutes available after launch — because everything you build afterwards lands on top of it.
(Site this was run on: Speech Kobo, four languages, Next.js App Router.)
Top comments (0)