HTTPS and SEO: how to move your site to HTTPS without losing traffic
Table of contents
To migrate HTTP to HTTPS is not to install a certificate: it is to leave nothing behind. A site with a valid certificate still shows "not secure" over one image loaded from http://, and since October 2026 Chrome asks permission before opening a public site without HTTPS.
What Google says about HTTPS in 2026 (and what it doesn't)
The conversation about HTTPS drags around a 2014 figure without context, plus a promise Google never made: that switching to HTTPS raises rankings.
- In 2014 HTTPS became a ranking signal, but the announcement itself calls it a very lightweight signal: it affects fewer than 1% of global queries and carries less weight than high-quality content.
- The current page experience documentation asks that pages be served in a secure fashion and notes that there is no single page experience signal; Core Web Vitals are used by ranking systems.
- If the same page lives on both HTTP and HTTPS, Google prefers to index the HTTPS version, according to the Search Console HTTPS report.
That reading sets up everything below: HTTPS rarely lifts you, but breaking it does cost you clicks.
Why HTTPS matters more for traffic than for rankings
As a ranking signal it is weak; as a front door it is decisive, because the warning appears before the content.
What changed in Chrome 154
Google announced on October 28, 2025 that with Chrome 154 (October 2026) the "Always Use Secure Connections" setting will be on by default: the browser will try HTTPS first and ask the user for permission before the first access to any public site without HTTPS, as explained on the Google Online Security Blog.
Two nuances change the urgency. The prompt is not repeated for sites the user visits often, so a loyal audience may never see it. And the variant does not apply to private sites: local IPs such as 192.168.0.1, single-label hostnames or shortcuts like intranet/. In the Chrome 141 experiment, the median user saw fewer than one warning per week.
Where it actually costs you visits
The full-page warning shows up before anything renders. And the most common case is absurd: sites that only use HTTP to redirect to HTTPS. That hop, even if it lasts half a second, is still an insecure navigation.
Mixed content: the mistake that breaks the padlock
Mixed content is the number one cause of "I have a certificate and I still show as not secure". There are two types:
- Active (scripts, stylesheets, iframes): the browser blocks it outright. The page breaks or loses its styles.
- Passive (images, video, audio): Chrome tries to auto-upgrade it to HTTPS; if the resource exists over a secure connection it loads, and if not, it does not load.
With passive mixed content present, the browser stops showing the page as secure even when the document arrives over HTTPS. The padlock does not fall because of the certificate; it falls because of an image. The technical explanation is on web.dev.
Finding mixed content in five minutes
Open DevTools and the Issues panel: it lists every mixed resource with its URL and status. Then search the served HTML for http://:
curl -s https://example.com | grep -o 'http://[^"]*' | sort -u
Also check inline CSS and JavaScript, email templates and URLs stored in the database. The rule: relative URLs or inherited protocol, never a hand-written http://.
The certificate: expiry, chain and hostnames it doesn't cover
Three different failures produce the same "your connection is not private" message, which is why they get misdiagnosed.
- Expired: the obvious one, and the most common on sites that renewed by hand.
- Incomplete chain: the intermediate is missing. It works in your browser and fails on mobile apps and APIs.
- Hostname not covered: the certificate is for
example.comand the visitor arrives atwww.example.com, or the reverse.
Prevention: automatic renewal through ACME (Let's Encrypt and friends) plus your own expiry alert, because provider emails do not always reach the person who watches the site. The concept from scratch is in what an SSL certificate is.
Three commands to inspect the real certificate
What the origin serves is not always what the hosting panel shows:
curl -vI https://example.com
openssl s_client -connect example.com:443 -servername example.com </dev/null 2>/dev/null | openssl x509 -noout -dates -subject -ext subjectAltName
curl -sI http://example.com | head -n 5
The first shows the full chain, the second the dates and covered names, and the third the status code and Location header of the hop from HTTP.
How to migrate HTTP to HTTPS without losing traffic: the right order
The classic mistake is the order: redirect first, certificate later, assets last.
The eight steps
- Backup and change window: database, files and a specific hour to roll back.
- Certificate installed and verified at the origin, before touching redirects: if it fails, every visit becomes an error.
- One 301 per URL, from HTTP to HTTPS on the same host, without passing through
wwwor another path first. - Canonicals, hreflang, sitemap and internal links already on HTTPS: if the canonical points to HTTP, you are telling Google the good version is the one that redirects.
- Search Console: add the HTTPS property and keep the HTTP one, which is where the hop shows up.
- CDN and theme assets without
http://, including fonts, icons and third-party scripts. - No
Disallowornoindexcarried over from staging. - HSTS last, with everything above stable for a few days.
What you don't do on the same day
Two things: enable HSTS before every subdomain works over HTTPS, and delete the redirects "because the migration is done". An old link in a forum or a PDF still arrives over HTTP and, without the 301, that traffic is a 404. The broader picture is in website migration without losing rankings and the mechanics in 301 redirect.
HSTS: what it fixes and when it takes your site down
The Strict-Transport-Security header makes the browser never try HTTP again for that domain. It removes the insecure hop, but it does not fix mixed content or indexing.
You raise it in stages, waiting the full max-age at each one: max-age=300, then 604800, then 2592000, adding includeSubDomains once every subdomain already serves HTTPS. Be careful with that parameter: it takes down any subdomain without HTTPS, including internal ones nobody remembers.
About preload, the preload list itself says preloading is not recommended: browsers already try HTTPS first and the extra benefit is minimal. It requires max-age of at least one year, includeSubDomains and every subdomain working. It is the part most copied from old guides.
The mistakes that cost the most traffic
http://tohttps://tohttps://www.chain: two hops per URL and two redirects counted in reports.- Mixed assets in a single template: the homepage is clean, the product page is not.
- Canonical pointing at the HTTP URL.
- Sitemap with HTTP URLs or with the old host.
- Certificate covering the apex and not
www, or the reverse. - HSTS enabled with half-migrated subdomains: the browser stops connecting and there is no way back from the server.
noindexorDisallowinherited from staging.
If the symptom is that new pages never show up, check the noindex tag before blaming HTTPS.
How to verify it went well
- Search Console HTTPS report: compares how many indexed URLs are HTTP and how many are HTTPS. It only exists for domain properties and for HTTPS URL-prefix properties; on an
http://property it shows as "HTTPS not evaluated". - Lighthouse or PageSpeed Insights: the "Does not use HTTPS", "Does not redirect HTTP traffic to HTTPS" and "Use a strong HSTS policy" audits are documented in Chrome for Developers.
- Your own crawl: walk the sitemap with
curland list any URL answering overhttp://or taking more than one hop. - Core Web Vitals: HTTPS is what enables HTTP/2 in practice, so a well-executed migration should not make LCP worse.
Final checklist
- Valid certificate, complete chain, every hostname covered.
- Zero mixed assets across all templates, not just the homepage.
- A single 301 per URL, on the same host.
- Canonicals, hreflang, sitemap and internal links on HTTPS.
- HTTPS and HTTP properties in Search Console, with the hop verified.
- HSTS rolled out in stages, with an expiry monitor in place.
- HTTPS report with no pending HTTP URLs.
- Old redirects kept alive.
The certificate is the requirement; the migration is leaving nothing behind.