Technical SEO · November 13, 2026 · 8 min read
Traffic Dropped After Switching to HTTPS. What Broke?
Two causes: Google is still swapping the indexed URLs, which fixes itself, or your redirect drops the path and sends every old page to the homepage.
By FluxWriter
Traffic dropped after switching to HTTPS most often for one of two reasons: Google is still swapping one set of URLs for another, or the redirect your host added is sending pages somewhere they do not belong. The first resolves on its own. The second does not, and every week it sits there is another week your best pages arrive at the wrong address. This walks through how to tell the two apart, what to check and in what order, and what a normal recovery looks like.
Google Treats The HTTP And HTTPS Versions As Two Different Sites
Google treats http://example.com/page and https://example.com/page as two different URLs. Not two versions of one page. Two addresses. Moving to HTTPS therefore changes every URL on your site at once. Google has to fetch each old one, follow the redirect, fetch the new one and carry the ranking signals across. That work is neither instant nor evenly spread.
There is no HTTPS penalty and no migration penalty. What you are watching is a queue. Google recrawls pages roughly in proportion to how often it crawled them before, so your homepage and your most-linked posts move across quickly while the long tail waits. As a rough rule, a site of a few dozen pages settles within a couple of weeks, and several thousand pages can take a month or more. Traffic dips in that window because part of your index is in transit.
The tell is the shape of the dip, not its size. A normal dip is broad and shallow. Most pages lose a little, nothing goes to zero, and the line climbs again without you touching anything. A fault looks different: named pages fall off a cliff and stay there while everything around them holds. If one page is at zero and its neighbours are fine, stop waiting and go and look at that page's redirect.
The Redirect That Lands On Your Homepage Is The Expensive One
The most common migration fault is a blanket rule that sends every http:// request to https://yourdomain.com, the root, with the path thrown away. Hosts and plugins ship it as a one-click option because it is simple to write and it does make the padlock appear. It also means every link pointing at your best post now lands on your homepage.
Google handles that the way it handles any redirect to something unrelated. The old URL does not hand its signals over, because the destination is not the same page. The old page falls out of the results and the homepage collects nothing useful in exchange. From the outside it looks like your best post being dropped for no reason, which is why owners spend a week rewriting content that was never the problem.
Checking costs one command. Run curl -sI http://yourdomain.com/your-best-post and read the Location header. It should return https://yourdomain.com/your-best-post: the same path, new protocol. If it returns https://yourdomain.com/ with nothing after the slash, that is your fault, and it is worth fixing today rather than at the weekend. The correct rule preserves the requested path and changes only the protocol.
Every Extra Hop Is Somewhere To Lose A Page
There are four front doors to a site: http and https, each with and without www. All four should end at the same place, and the route matters. A request travelling from http://example.com/page to https://example.com/page to https://www.example.com/page has taken two hops where one would do.
Google will follow a short chain, so one extra hop rarely kills a page by itself. The problem is that chains are where rules disagree. One adds HTTPS, a second adds www, a third strips a trailing slash, and somewhere in that sequence a URL gets rewritten in a way nobody tested. Chains break quietly on the pages you look at least: an old category page, a landing page with a query string, a PDF.
Test all four variants of a real deep URL, not just the homepage. The homepage almost always works, because it is the one thing everybody tests. If you are enabling HSTS, understand it first: browsers remember the instruction for the max-age you set, so a short max-age while you debug leaves room to change your mind.
The Shape Of The Drop Tells You Which Fault You Have
Before changing anything, match what you are seeing to a pattern. The percentages and timings below are illustrative rather than measured, and they move with the size of your site and how often it is crawled, but the shapes hold.
| What you are seeing | Most likely cause | Normal or broken | First check |
|---|---|---|---|
| Broad dip of roughly 10-30%, recovering across two to four weeks | Index swap still in progress | Normal | Count indexed https URLs week over week |
| Two or three top pages at zero, homepage sessions up | Redirects collapsing to the root | Broken | curl -sI an old URL, read Location |
| Slow drift down over a month, no single page at zero | Redirect chains through www or trailing slashes | Broken | Count hops on all four URL variants |
| Both http and https versions appearing in results | Canonical tags and sitemap still say http | Broken | View source, read the canonical tag |
| Rankings look normal, but reports show near-zero | Reporting pointed at the old property | Not a real drop | The property prefix in Search Console |
| Pages load but look wrong, console shows blocked requests | Mixed content in a shared template | Broken | Browser console, one page per template |
The two worth separating hardest are the broad shallow dip and the single-page cliff. A broad dip where everything still ranks somewhere is a queue emptying, and patience is the right response. A cliff on named URLs while the rest of the site holds steady is a redirect fault, and patience fixes no redirect fault.
Your New Pages Are Still Voting For The Old Ones
A redirect tells Google where a page went. Everything else on the page should say the same thing, and after a rushed migration a lot of it does not.
Start with the canonical tag. If https://example.com/page carries a canonical pointing at http://example.com/page, you have shipped a contradiction. The server says the page moved; the page says it did not. Google picks one, and you do not get a vote. The same applies to the URLs inside your sitemap, your RSS feed, your og:url tags and any structured data that embeds an address.
Then internal links. Templates, hardcoded navigation and old posts with absolute http:// links in the body all push readers and crawlers through a redirect that need not exist. A search and replace of http://yourdomain.com with https://yourdomain.com across the content, run on a backup first, clears most of it.
Mixed content belongs in the same sweep. If a template pulls a script or an iframe over http, the browser blocks it, and a page missing its script may be missing its layout, its images or its checkout button. That is a conversion problem before it is a ranking problem, and it never shows up in a rankings report.
Search Console Is Measuring A Site You No Longer Have
This one catches people who did the migration properly. URL-prefix properties in Search Console are protocol-specific, so a property registered as http://example.com does not report on https://example.com. The day your redirects went live, that property's clicks began falling towards zero, which looks identical to a disaster. Nothing was lost. The reporting is aimed at an address that redirects away.
Add a Domain property if you can verify through DNS, since it covers both protocols and every subdomain at once. Otherwise add the https:// prefix property and treat the old one as an archive. Resubmit your sitemap under the property that covers https, with https URLs inside it, and remember that a disavow file belongs to one property and does not follow you across.
Check the rest of your stack for the same fault. Rank trackers configured with an http:// domain, uptime monitors and any dashboard that requests your site by a hardcoded URL will all report something strange. Analytics usually keeps working, because the tag travels with the page, but the channel mix can shift when protocols change, so treat the first week of referral data as unreliable.
FAQ
How long should I wait before deciding something is broken?
Waiting is right only for a broad, shallow dip where pages still rank and the trend is flat or improving. Assume a couple of weeks for a small site and longer for a large one. If one page went to zero while its neighbours held, do not wait at all. A page whose redirect points at the wrong destination does not recover with time.
Should I leave the old HTTP URLs live so nothing breaks?
The old URLs should keep answering, but only with a redirect. If both versions return real content, you have two copies of every page competing for the same query and Google picks between them. Keep the redirects permanently rather than retiring them after a few months, because links and bookmarks pointing at http:// arrive for years.
Will I get exactly the same rankings back?
Redirects are designed to carry signals to the new URL and usually do, but nobody can promise a specific position on any site. Expect the same general shape rather than identical numbers. If a page is still flat two months after a clean migration with a verified one-hop redirect, treat it as a content or competition question rather than a migration one.
The Practical Takeaway
This week, run the checks in the order that finds expensive faults first. Run curl -sI against three of your best old http:// URLs and confirm each Location header returns the same path on https. Test all four protocol and www variants of one deep page and count the hops. View source on a live page and read its canonical tag. Open Search Console and confirm the property you are reading covers https. Most migrations that look broken fail at exactly one of those four points, and the first is both the cheapest to fix and the most costly to leave. Start with the curl check on your single best page.