Redesigning without losing SEO: compare titles, canonicals, headings and links before and after

Before you start

• Crawl or save the old site as static HTML, or keep a copy of the build, so you have something to compare against.
• Keep URLs. If a URL must change, plan the map with changing URLs without losing rankings.
• Keep the content the same on purpose, so any ranking change can be traced to the design, not the copy.
• Keep staging out of the index with a login or noindex, and remember to remove it at launch.

The diff script

Point it at two folders of HTML (for example the old build and the new one). It lists removed and added pages and, for pages in both, which of eight facts changed:

import re, sys, os, html

def facts(path):
    s = open(path, encoding='utf-8').read()
    g = lambda p: (re.search(p, s, re.S | re.I) or [None, ''])[1].strip() if re.search(p, s, re.S | re.I) else ''
    strip = lambda t: html.unescape(re.sub(r'<[^>]+>', '', t)).strip()
    return {
        'title': strip(g(r'<title>(.*?)</title>')),
        'description': g(r'<meta[^>]+name=["\']description["\'][^>]*content=["\']([^"\']*)'),
        'canonical': g(r'<link[^>]+rel=["\']canonical["\'][^>]*href=["\']([^"\']*)'),
        'robots': g(r'<meta[^>]+name=["\']robots["\'][^>]*content=["\']([^"\']*)'),
        'h1': strip(g(r'<h1[^>]*>(.*?)</h1>')),
        'h1_count': str(len(re.findall(r'<h1[\s>]', s))),
        'jsonld_blocks': str(len(re.findall(r'application/ld\+json', s))),
        'internal_links': str(len(re.findall(r'<a[^>]+href=["\'](?!https?:|mailto:|#)', s))),
    }

def pages(root):
    out = {}
    for d, _, fs in os.walk(root):
        if '.git' in d.split(os.sep): continue
        for f in fs:
            if f.endswith('.html'):
                out[os.path.relpath(os.path.join(d, f), root)] = os.path.join(d, f)
    return out

old, new = pages(sys.argv[1]), pages(sys.argv[2])
print('Pages: %d before, %d after' % (len(old), len(new)))
for p in sorted(set(old) - set(new)): print('REMOVED  %s   (needs a redirect or a deliberate 404)' % p)
for p in sorted(set(new) - set(old)): print('ADDED    %s' % p)
changes = {}
for p in sorted(set(old) & set(new)):
    a, b = facts(old[p]), facts(new[p])
    for k in a:
        if a[k] != b[k]:
            changes.setdefault(k, []).append((p, a[k], b[k]))
for k, items in changes.items():
    print('\n%s changed on %d page(s)' % (k, len(items)))
    for p, a, b in items[:4]:
        print('   %s\n      before: %s\n      after:  %s' % (p, a[:90], b[:90]))

A real run on this site

I ran it on an earlier commit of this site and the current one, a worktree checkout of each. It reported 48 pages before, 58 after, ten pages added, none removed. It also reported, for 40 pages, that the canonical had changed, and for 33 that the internal link count had changed. Both were intended: the canonicals moved to clean URLs, and each page gained links to the new guides. The point is that I could confirm they were the only changes, instead of hoping.

At launch

1. Remove any staging noindex or password, and run the post-deploy checker from the traffic-drop triage on the key URLs.
2. Submit the sitemap again and spot-check pages with the URL Inspection tool.
3. Run the pre-launch checklist for everything the diff does not cover.
4. Watch the Page indexing and Performance reports for the following weeks; Google says to expect temporary fluctuation when a site changes.

What this does not cover