If you are wiring a blog onto an existing marketing site today, put it at yoursite.com/blog — unless your stack cannot serve it there without a proxy layer you would have to own forever, in which case blog.yoursite.com is a perfectly defensible choice you will not be punished for. Google has said for well over a decade that the two are treated the same, and the case studies back that up more than SEO folklore admits. What has changed in 2026 is a second scoreboard: one audit of 13 client sites found subdirectory pages pulled 2–5x more citations per page from ChatGPT and AI Overviews than near-identical subdomain pages. That is a real signal, from a small sample, and it still should not outrank the question of how your blog actually deploys.
Here is the honest version of the trade-off, with the parts most guides hand-wave.
The official position has barely moved since the George W. Bush administration. In a 2007 post on subdomains and subdirectories, Google's Matt Cutts wrote that "historically, it's been kind of a wash about when to use subdomains vs. subdirectories" — then added his own tiebreaker: "My personal preference on subdomains vs. subdirectories is that I usually prefer the convenience of subdirectories for most of my content."
A decade later, John Mueller said the same thing with the same footnote. Ahrefs quotes him flatly stating that "Google websearch is fine with using either subdomains or subdirectories", and in a May 2018 office-hours session he expanded on it in a line that Search Engine Journal recorded: "In general, we see these the same. I would personally try to keep things together as much as possible." He suggested reserving subdomains for content genuinely different from the main site.
Notice what both men did. They said the crawler doesn't care, and then both volunteered a personal preference for keeping things together. That gap is the whole article. Google's equivalence claim is about how a URL gets processed — not about what a site accumulates over three years of links, crawls, analytics configs and engineering decisions. The processing really is equivalent. The accumulation is not.
Strip out the folklore and a short list survives.
Backlink equity doesn't cross the host boundary for free. As NameSilo's breakdown of subdomain authority puts it, "backlinks pointing to a subdomain do not directly improve the SEO of your main domain". For a small SaaS site, that is the difference between your one genuinely great post — the one that gets picked up on Hacker News — strengthening the domain your pricing page lives on, or strengthening a host that only ever serves blog content.
Crawl attention gets fragmented. Bluehost's comparison notes that if you run several subdomains with thin or infrequently updated content, "Googlebot may deprioritize crawling them", while every external link to a subdirectory strengthens the root domain. On a 200-page marketing site with a weekly blog, this is a rounding error. On a subdomain that publishes twice a quarter, it is the difference between a post indexed on Tuesday and a post indexed in three weeks.
Third-party metrics score them separately. The same source notes that tools like Moz's Domain Authority and Ahrefs' Domain Rating "typically treat subdomains as separate entities." Those are vendor scores rather than ranking factors — but they are the numbers you'll be staring at when you judge whether the blog is working.
Measurement gets a tax. Google's own Search Console documentation is explicit that a URL-prefix property excludes other hosts: "Subdomains, such as m.example.com, are not included in this URL prefix." A Domain property does cover every subdomain, so this is a chore rather than a wall — one you repeat in analytics, where subdomain tracking needs deliberate setup to stop your own blog from appearing as a referral source and stealing credit for the signup it caused.
What does not differ: your ability to rank a great post. Which brings us to the newer question.
The genuinely new input is generative search. An analysis published this year tracked 13 client sites from January to March 2026 across ChatGPT search mode, Gemini, Google's AI Mode, Perplexity and AI Overviews, and found subdirectory content cited "2 to 5 times more often per page" than the subdomain equivalents. Its proposed mechanism is the interesting part: "different hosts get treated as different entities, even when they share a registered domain." An LLM anchoring on the host string sees docs.acme.com and acme.com as two organizations that happen to rhyme.
The sharpest data point in that write-up is a near-controlled one. A SaaS company kept nearly identical guides at docs.theircompany.com and theircompany.com/docs; over 90 days the subdirectory version earned 11 citations across ChatGPT and Gemini, the subdomain version 2. Same content, same company, roughly 5x the citations. If you are betting on being the source an answer is built from rather than a blue link, that gap is the strongest argument on the subdirectory side of the ledger.
Now the caveat the authors themselves lead with: it is a "small sample, far from a peer-reviewed study." Thirteen clients over one quarter is a signal to weigh, not a law to plan around — and citation behaviour in these systems is still driven mostly by page-level craft, which is why our guide to earning citations from ChatGPT and AI Overviews spends its time on extractable paragraphs and named inline sources rather than on hostnames. Treat host consolidation as a tiebreaker that costs you nothing when you are starting fresh, not as the lever.
If you go looking for proof that subdirectories win, you will find migration case studies. Read them closely and most of them stop proving it.
CognitiveSEO ran the cleanest test on itself: in late June 2017 it moved explorer.cognitiveseo.com to cognitiveseo.com/site-explorer/ and reported no improvement whatsoever — rankings hovered where they had been. The famous HotPads win from February 2014, where John Doherty saw an immediate impact after moving the blog to a subfolder, came bundled with a WordPress replatform, indexation fixes and a low-quality content cull. Iwantmyname moved its blog to a subdomain in July 2014 and traffic dropped, but the recovery after reversing it couldn't be cleanly attributed either.
Ahrefs made the structural version of this critique after examining Mention.com's 2016 migration and finding minimal direct impact — the traffic gains arrived later, when the company added substantial new content. Their list of confounders is the one to keep: migrations usually ship alongside better internal linking, content edits, tracking changes, platform redesigns and indexation fixes, any of which could carry the result.
The practitioner consensus lands in a sensible place. Eric Enge notes that "Google is pretty good at seeing a subdomain as being an integrated part of the main domain," recommending subfolders for new content while cautioning against moving established blogs purely for SEO. Cyrus Shepard grants that "empirical evidence shows that moving content...to a subfolder often results in increased visibility," then adds that the effort may not justify the result.
Translation for a founder: on a fresh blog, pick the subdirectory because it's free to pick. On a three-year-old subdomain with real backlinks, the SEO case alone does not clear the bar.
Here is the part that rarely makes it into SEO guides, because their authors aren't the ones on call.
The most useful public account of subdomain regret comes from the team behind Clubhouse (now Shortcut), who called their per-organization subdomain architecture "a mistake worth rewriting a lot of code for". Their reasons were entirely operational: Google requires a domain whitelist for OAuth access, which makes wildcard domains awkward; Extended Validation certificates can't be issued to a wildcard like *.clubhouse.io; and every environment needed its own URL-parsing code path. Moving the org slug into the path — app.clubhouse.io/myorgname/stories instead of myorgname.clubhouse.io/stories — deleted all three problems at once. That's an app-architecture story rather than a blog story, but the failure modes rhyme: each additional host multiplies certificate, cookie, auth and environment configuration.
Cutting the other way, your platform may simply make the choice for you. HubSpot only serves a blog on your subdirectory if the main marketing site is also hosted on HubSpot Content Hub; otherwise a subdomain is the supported path. GitHub Pages enforces one custom domain per repository — "only one repository at a time can be associated with a given custom domain" — so multi-repo subdirectory routing isn't a setting you flip.
If you do want /blog on a separately built site, three documented paths exist, in ascending order of what they cost you:
basePath in next.config.js prefixes routes and assets automatically. It's a build-time setting, not a runtime one — decide before you ship./blog/:path* to the blog deployment from the main project's vercel.json. Kevin Goedecke's write-up documents both the config and its edge — deeper nested subfolders don't cooperate.href, src, link and meta values so nothing leaks the origin host, plus 301s from the old domain. It works, and the same write-up flags CORS handling, template complexity and reported subrequest latencies past 700ms.Option 3 is a permanent piece of infrastructure that sits between users and every blog post you publish. If your answer to "who fixes this at 2am" is "me, and I also do sales," that cost is real and it is recurring — which is exactly the calculus Cutts described in 2007 when he told people to pick whichever was more convenient.
Four questions, in order:
/blog without a new proxy layer? If yes, subdirectory. Done — you just took the SEO tiebreaker and the AI-citation tiebreaker for free.Most founders reading this land on option 1 within about ninety seconds, which is the point: this decision deserves ninety seconds, not a week.
An empty subdirectory ranks worse than a subdomain that publishes every week. The URL structure sets a ceiling of a few percent; cadence and depth set everything else — and the reason the choice feels heavy is usually that it's the last decision before the hard part starts.
That hard part is where Magic Share fits. Point the agent at your site, it builds a brand profile from your actual pages, researches a topic, and holds the draft until you approve it — with 100% of cited links verified and a verification report attached, because a post full of dead citations damages you regardless of which host serves it. If your blog lives in a repo, the agent opens a GitHub pull request and merging it is the publish step, which pairs neatly with the pull-request review workflow for Astro and Next.js blogs we've written up separately. Whichever structure you chose, the posts land in it on a schedule you set — and our look at how often a small blog should publish explains why a cadence you can hold at full depth beats a burst you can't.
Is a subdomain bad for SEO? No. Google has said consistently since 2007 that subdomains and subdirectories are treated equivalently in search, and John Mueller reiterated in 2018 that "in general, we see these the same." The measurable differences are in link equity consolidation, crawl attention and measurement overhead — not in a ranking penalty.
Should I move my blog from a subdomain to a subdirectory? If the blog is new or has little accumulated authority, yes — the move is cheap. If it has years of backlinks and rankings, the SEO case alone is thin: expect a 2–4 week traffic dip and 4–6 week recovery, and note that CognitiveSEO's own migration produced no measurable improvement at all. Bundle it with a replatform if you can.
Do subdirectories really get more AI citations? One 2026 study of 13 sites found subdirectory pages cited 2–5x more often per page across ChatGPT, Gemini, AI Mode, Perplexity and AI Overviews, with one near-identical docs comparison at 11 citations versus 2 over 90 days. The authors call it a small sample, so treat it as a tiebreaker, not a mandate — page-level structure still drives citations more than hostnames do.
How do I serve a separately deployed blog at /blog?
Three documented options: basePath if the blog lives in the same Next.js app, platform rewrites (Vercel's vercel.json can rewrite /blog/:path* to a second project), or a reverse proxy such as a Cloudflare Worker with HTMLRewriter. The third works but adds a permanent component to your critical path.
Does a subdomain need its own Search Console property?
A URL-prefix property excludes other hosts, so https://example.com/ won't report on blog.example.com. A Domain property covers every subdomain at once. Either way you'll spend extra setup time on both Search Console and analytics that a subdirectory gives you for free.
Pick the subdirectory when your deployment allows it, accept the subdomain without guilt when your platform or your team structure demands it, and don't migrate a healthy ranking subdomain on SEO grounds alone. The evidence for subdirectories is real but modest; the evidence that consistent publishing matters more is overwhelming.
Once the structure is settled, the bottleneck is drafting, sourcing and fact-checking every week. Start with three free posts and see what a verified draft looks like before you commit to a cadence.
Magic Share researches, writes and fact-checks posts like this for any site — point it at your URL and review your first draft today.
Plant your first post