What is replacing blogging?

Nothing has replaced blogging outright. What changed is that a blog used to do three jobs at once (get found, get followed, get cited), and those jobs now get split across newsletters, video, social feeds, docs sites and AI answers. The long-form post survived because it's the one format you can host on a URL you control. For most developers in 2026 the sensible setup is a blog on your own domain as the source, with everything else treated as distribution.

Disclosure: I work on VeloCMS, a small hosted publishing platform. It's one of eight options in the comparison near the end, listed alphabetically, with its weak spots. This piece is also published on Hashnode, which is another of the eight.

Every figure below came from a page I opened on 2026-10-18. Where a page wouldn't open (Medium's and Substack's help pages returned a 403 to my tools), I say so. Most of the audience data describes news readers in the US or worldwide, not developers specifically, and I flag that where it matters.

What did blogs actually lose, and to whom?

Attention moved to feeds. That much the numbers support, though they measure news habits, not tech writing.

Pew's social media and news fact sheet (published 25 September 2025, survey fielded 18 to 24 August 2025) says 53% of US adults get news from social media at least sometimes. Asked where they regularly get news, 38% named Facebook and 35% YouTube. Instagram and TikTok each got 20%, X 12%, Reddit 9%, Threads 3% and Bluesky 2%. The list I read doesn't include LinkedIn or Mastodon.

The Reuters Institute's Digital News Report 2026 (16 June 2026, about 97,500 respondents across 48 markets) puts the same shift in a global frame. Social and video networks reached 54% of respondents for news, ahead of news organisations' own sites and apps at 51%. YouTube was used for news by 34%, news podcasts by 11%, and 27% got news from news-focused creators (46% from creators of any kind).

So the first honest answer to the question is "feeds and creators took the discovery job." If your blog's role was being the front door, that door moved. But look at what the surveys can't say: they don't count how many people write long posts, and they don't claim publishing stopped. WordPress.org says its software runs "over 43% of all sites across the web," which is a vendor claim about all sites, not blogs, yet it's hard to square with a dead format.

Did newsletters take the blog's place?

They took the "follow" job, not the "find" job. Pew's February 2026 report on email newsletters (5,153 US adults, same August 2025 fieldwork) found 6% get news from them often and 24% sometimes. That's a real habit and a minority one.

The line between the two has blurred. Ghost, Substack and others publish each issue as a web page with a feed, so a newsletter platform is often a blog with an email button. The question becomes "does each piece live at a stable URL." I covered the measurement side in are newsletters still relevant in 2026 and the platform side in which platform is best for newsletters, so I'll leave email there.

Did short posts on X, LinkedIn, Bluesky and Mastodon replace the blog post?

For quick takes, yes. For anything you'd want to find again in two years, no.

A thread is fast, invites replies and reaches people who'd never visit a new domain. Its weaknesses are structural: the platform sets the reach, search inside it is poor, and a post can vanish with an account. Pew's regular-news shares (X 12%, Threads 3%, Bluesky 2%) show how different those audiences are in size, but they measure news, not developer reach.

My take, and it's a judgment, not data: social posts are best treated as the trailer and the blog post as the film. A trailer can be excellent. It still points somewhere.

What about video, podcasts and communities like Discord and Reddit?

Video is the biggest competitor for explanation. Reuters' 34% global YouTube figure and Pew's 35% US figure are news numbers, so I won't stretch them to "developers prefer video"; I couldn't find a sourced developer-specific number. Podcasts are a smaller slice, 11% for news podcasts in the Reuters data, and suit conversation more than reference. Neither can be searched, skimmed or pasted into a terminal the way text can.

Communities took over the comment section. Reddit sits at 9% for regular news in Pew's list, and I found no sourced Discord figure. For developers, GitHub Discussions is the closer analogue; GitHub's docs call it "a collaborative communication forum for the community around an open source or internal project." That's the job blog comments used to do, moved next to the code.

Where do developers actually write now?

Mostly in places that aren't called blogs, which the "blogging is dead" take misses.

Docs sites and READMEs carry the explanation people actually need when they use a tool. Nobody calls the Hugo or Astro docs a blog, but they're long-form writing that gets read and cited.
Changelogs carry the "what changed" story. Keep a Changelog's first principle is that "changelogs are for humans, not machines," and GitHub, for one, serves a repository's releases as an Atom feed (/releases.atom).
TIL notes carry the small stuff. Simon Willison's site describes itself as "Things I've learned, collected in simonw/til," listed 583 entries when I looked, and offers an Atom feed. A TIL is a blog post with the ceremony removed.
Digital gardens carry the unfinished stuff. Maggie Appleton's history page defines one as "a collection of evolving ideas that aren't strictly organised by their publication date," and traces the idea through Mark Bernstein's 1998 essay and Mike Caulfield's 2015 "The Garden and the Stream."
Hosted developer platforms carry the audience. DEV calls itself "a community of software developers getting together to help one another out." Hashnode and Medium play similar roles.

None of those is a replacement. They're narrower than a general blog, and a developer often runs three or four at once.

How do AI answer engines change who finds a post?

They reduce clicks, and they make being indexable more important, not less.

Pew analysed browsing from 900 US adults in March 2025 (68,879 Google searches, 12,593 of which showed an AI summary). On visits with a summary, people clicked a traditional result 8% of the time, against 15% without one. They clicked a link inside the summary on 1% of visits, and ended the session on 26% of visits with a summary versus 16% without. That's one browser panel, one month and Google only, so read it as a direction.

Reuters' 2026 data adds the self-reported side: 10% use AI chatbots for news weekly (up from 7%), 16% among under-35s, and 42% of chatbot news users say they always or often click through to the sources. The comparable figures were 44% for search and 36% for social. Self-reported clicks and measured clicks needn't agree.

Cloudflare's August 2025 analysis of its own network shows the crawler side. Its crawl-to-referral ratio (pages fetched per visit sent back) went from about 286,930 in January to 38,066 in July 2025 for Anthropic, 1,217 to 1,091 for OpenAI, 54.6 to 194.8 for Perplexity and 3.8 to 5.4 for Google. Cloudflare sells bot management and the sample is its customers, so I read this as "AI products fetch far more than they send back," not as a forecast of your referrals.

Developers add a trust angle. In Stack Overflow's 2025 survey, 84% use or plan to use AI tools, yet only 3.1% highly trust their accuracy (29.6% somewhat), and 66% cite "AI solutions that are almost right, but not quite" as a frustration. People use these tools and then check them, which is when a clear, dated, linkable source gets opened.

What does Google say you need? Its AI features page (updated 2025-12-10) states there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary," and that a page must be "indexed and eligible to be shown in Google Search with a snippet." It adds that no special schema.org markup is needed and that this traffic is counted in the Web search type in Search Console.

Why does owned long-form still matter?

Because the things people are now fighting over (citation, search, permanence) all attach to a URL.

A citation needs a page to point at. Whether an assistant or a person quotes you, the thing quoted has to live somewhere that stays put. Tim Berners-Lee's 1998 note on the subject starts with "A cool URI is one which does not change," and a platform that shuts a feature or changes its domain rules can break yours. A September 2018 post quoting Medium's own notice has Medium saying it is "no longer offering new custom domains as a feature" (the help page itself returned a 403 to me, so I can't tell you today's policy).

A feed is the other durable layer. RSS 2.0.11 was published on 30 March 2009 and is still the current spec on the RSS Advisory Board's site, and Atom is a December 2005 IETF standard (RFC 4287). JSON Feed 1.1 followed on 7 August 2020. Readers, aggregators and plenty of tools still consume them, and nobody's algorithm sits between you and a subscriber.

Webmention, a W3C Recommendation since 12 January 2017, covers the third piece: telling a page that someone linked to it, with no network in the middle.

One caution so I don't oversell this. I know of no measurement showing that owning a domain increases your odds of an AI citation. The argument is about control and permanence, not a guaranteed lift.

What should a developer do in 2026?

Here's the shortest list I'd stand behind. It's a practical checklist, not research.

Own the domain. Even if the posts live on a hosted service, put them on a name you can repoint. Moving host should never change a URL.
Keep URLs permanent. Skip dates and categories in slugs, and when you must move something, 301 it. This one habit does more for long-term findability than any tag.
Publish a feed and advertise it. Add autodiscovery links in the page head so tools can find it:

<link rel="alternate" type="application/atom+xml" title="Posts (Atom)" href="/atom.xml">
<link rel="alternate" type="application/feed+json" title="Posts (JSON Feed)" href="/feed.json">

Serve real HTML. Google's bar is "indexed and eligible to be shown in Google Search with a snippet." Put the article text in the first response, not behind a client-side render, and keep a clean sitemap.
Add article markup that matches the page. Google's Article documentation (updated 2026-09-08) says there are no required properties, recommends author, datePublished, dateModified, headline and image, and warns it can't guarantee any feature will appear. A minimal version:

{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "headline": "Running Postgres on a Raspberry Pi",
  "datePublished": "2026-10-18",
  "dateModified": "2026-10-18",
  "author": { "@type": "Person", "name": "Your Name", "url": "https://example.com/about" },
  "image": "https://example.com/img/pi-postgres.png"
}

Practise POSSE. The IndieWeb defines it as "Publish (on your) Own Site, Syndicate Elsewhere": post at home first, then share copies or links on other services that point back. The IndieWeb page lists Ghost, Jekyll and WordPress setups doing this, so tooling exists.
Be careful with canonical tags on syndicated copies. DEV's RSS import offers "Mark the RSS source as canonical URL by default," and Hashnode's site says imports keep canonical URLs. Google's canonicalization troubleshooting page (updated 2026-08-21) has a "Syndicated content" entry that reads differently: the canonical link element "is not recommended for those who want to avoid duplication by syndication partners, because the pages are often very different. The most effective solution is for partners to block indexing of your content." My reading: syndicate a summary and a link, or accept that a bigger platform's copy might outrank yours.
Split your writing. Notes in a TIL stream, essays on the blog, release notes in the changelog, all on one domain.
Keep the source in Markdown in Git. Being able to leave in an afternoon is the best insurance, whatever tool you pick.
Watch Search Console. AI feature traffic is counted under Web, so no new tool is needed.

Which options can hold a long-form home?

Eight options, alphabetical, not ranked. Prices are USD from each vendor's own page on 2026-10-18. I couldn't verify a few things (marked), and I list those honestly rather than guess.

Astro, Hugo or Jekyll on a static host

You write Markdown, a generator builds HTML, and a host serves it. Hugo calls itself "a static site generator written in Go, optimized for speed and designed for flexibility" and is Apache 2.0 licensed. Astro documents an @astrojs/rss package, and the jekyll-feed plugin produces an Atom feed for Jekyll. The trade-off is that you build everything else: no editor, no email list, no comments. A future-dated post also needs a rebuild after its date, which is how static builds work.

Price: the generators are free. Netlify lists Free at $0 with 300 credits, Personal at $9 and Pro at $20 a month. GitHub Pages has a 1 GB site limit, a soft 100 GB monthly bandwidth limit, and says it isn't for running an online business.
Feed: yes, via the packages above.
Exit: your files already sit in Git.

DEV

DEV describes itself as a developer community built on open source Forem software. Writing there puts you in front of an existing audience, and the site serves an RSS feed (I fetched dev.to/feed). The DEV team's "Revamped RSS Feed Imports" post says individuals can add several RSS feeds from an external blog, with the canonical choice noted earlier (DEV's help page on RSS import that I read is written for organizations). The trade-off is the usual one for a hosted community: your posts sit under dev.to, and I found no custom-domain feature on the pages I read.

Price: I found no writer price on the pages I read.
Feed: yes, site-wide and RSS import.
Exit: I couldn't find a documented export on the pages I read.

Ghost

Ghost is MIT licensed on GitHub and sold hosted as Ghost(Pro). It's built around publishing with members and email, and it serves RSS (I fetched ghost.org/blog/rss/). The README says 100% of revenue from Ghost(Pro) goes to the Ghost Foundation. The trade-off is that pricing follows member count, and custom themes are off the lowest tier.

Price: the page's headline prices are per month, billed yearly, and move with an audience-size slider. At the default setting: Starter $18 (1,000 members, default theme only, no paid subscriptions), Publisher $29 (1,000 members, custom themes, paid subscriptions), Business $199 (10,000 members in the plan comparison). I couldn't read monthly-billing prices.
Own domain: yes on all listed plans.
Exit: open source, and you can self-host the same software.

Hashnode

Hashnode is a developer-focused hosted blogging platform with a Markdown workflow, and it serves RSS (I fetched velocms.hashnode.dev/rss.xml). The key ownership features sit behind Pro: its Pro page lists custom domain, GraphQL API and headless mode, webhooks, GitHub backup of posts and drafts as Markdown, branding removal and scheduled publishing. The trade-off is that the free tier gives you the audience but not those controls.

Price: Pro is $5 a month or $50 a year, per publication. A free account gets one publication.
Own domain: listed as a Pro feature on its Pro page (its main page lists "Your own custom domain" without naming a plan, so check when you sign up).
Exit: GitHub backup is a Pro feature.

Medium

Medium's own pages (plans, creators, help) returned a 403 to my tools, so here's what I could confirm. A feed exists (medium.com/feed/@medium returned RSS). Search results quoting Medium's help centre say the Partner Program pays writers based on member reading time, and that eligible stories sit behind the paywall. Custom domains were closed to new setups according to a 2018 post quoting Medium. The trade-off is plain: you get a reading audience and a revenue option, and you give up control of the URL. I won't print a price I couldn't read.

Price: unverified, page blocked.
Own domain: not available for new setups as of the 2018 notice; current status unverified.
Exit: unverified.

Substack

Substack is newsletter-first. Its page says writers keep 90% of their revenue minus card fees, and that creators "own all of your content and your subscriber list." Its publications serve an RSS feed (I fetched on.substack.com/feed). The trade-off is the 10% cut on paid revenue, and a format built around email posts more than a site you design. The going-paid page didn't mention exports, which isn't evidence that they don't exist.

Price: the page doesn't list a platform subscription; the fee is a share of paid revenue.
Own domain: Substack's own older announcement post says a one-time $50 fee per publication (its help page returned a 403, so check the current fee).
Exit: not stated on the page I read, and Substack's help pages wouldn't open.

VeloCMS

VeloCMS is a hosted blog platform; its developer page lists Next.js and PocketBase as the stack. It serves /feed.xml, a sitemap and BlogPosting JSON-LD (I checked a post on learn.velocms.org, a site we run), and a REST API exists from Pro with posts:write keys. Importers cover WordPress, Substack and Ghost. The trade-off: it's young, it's hosted only, and its docs say it's "managed SaaS only, no self-hosted core, no source access."

Price: Pro $9 a month ($86.40 a year), Business $29, Agency $69. Every plan starts with a 14-day trial that takes a card up front and charges on day 15. There is no free plan.
Own domain: yes, all plans.
Exit: I found no public page that documents an export. The REST API docs list GET /api/v1/posts under a posts:read scope (Pro and up), so a script can pull posts out. The Markdown help page covers importing only.

WordPress

WordPress is two things. The software at WordPress.org is GPLv2-or-later and open source. WordPress.com is the hosted service. Its free plan shows WordPress.com ads to visitors, Personal is $9 a month on monthly billing or $4 a month billed every 12 months, and the page says a custom domain is free for one year with annual plans (the free plan stays on a wordpress.com subdomain). WordPress feeds work out of the box (I fetched wordpress.org/news/feed/, a self-hosted WordPress site). The trade-off is that self-hosting means you own updates, backups and security, and the hosted route trades that for tiers.

Price: self-hosted software is free plus hosting; WordPress.com as above.
Own domain: paid hosted plans (the free plan has none); yes self-hosted.
Exit: open software and a large ecosystem of migration tools.

Who is each option best for?

Astro, Hugo or Jekyll: developers who like Git and want total control, and don't mind building the rest.
DEV: writers who want an existing developer audience more than they want their own domain.
Ghost: publications that want members and email in an open source product, and will pay by headcount.
Hashnode: developers who want a reading audience and a custom domain, and will pay $5 a month for the controls.
Medium: writers who want reach and a revenue option and can accept limited URL control (details unverified, see above).
Substack: writers whose product is the newsletter, who accept a percentage of paid revenue.
VeloCMS: people who want a blog, members and newsletter under one hosted login, and can accept a young, hosted-only platform.
WordPress: anyone who needs the biggest plugin and hosting market, or already knows it.

Where is VeloCMS the wrong pick?

Several places.

If you need self-hosting or source access, VeloCMS can't do it. The static generators, Ghost and WordPress can. If you want a large plugin or theme market, WordPress is far ahead; VeloCMS's developer page says third-party plugin development isn't open to outside developers today. If your workflow is Markdown files in Git, the help page says there's "no batch Markdown importer yet" and no front-matter parser, so you'd script POST /api/v1/posts yourself with a Pro key. If you want readers delivered to you, DEV and Hashnode come with a community and VeloCMS doesn't, so every reader arrives through your own distribution.

Cost matters too: with a trial and no free plan, the lowest price is $9 a month, while GitHub Pages (public repositories on a free GitHub account), Netlify Free, WordPress.com Free and a Hashnode free publication cost nothing. On speed, its /performance page shows a PageSpeed Insights run dated 2026-09-27 with desktop LCP at 841 ms and mobile LCP at 3.90 s, and it says mobile misses its sub-1s target; the field LCP at p75 across live VeloCMS blogs was 2.30 s (1,227 samples, 2026-07-20 to 2026-10-18).

One smaller gap I found: the feed autodiscovery link was on the blog index page of learn.velocms.org but not on the post page I checked.

FAQ

Is blogging dead in 2026?

No, but its role narrowed. Feeds, video and creators took much of the discovery job (54% of Reuters' respondents use social and video networks for news), while the long-form post remains the format at a URL you control.

Are AI answers killing blog traffic?

They reduce clicks on classic results in one Pew study: 8% of visits with an AI summary versus 15% without. It covers Google only, for one month, so it's a signal, not a forecast.

Should I cross-post to dev.to and Hashnode?

If you do, publish at home first, link back, and don't expect a canonical tag to settle which copy ranks. Google's troubleshooting page doesn't recommend it for syndication partners.

Do I need structured data to be cited by AI?

Google says there's no special schema.org markup required for AI Overviews or AI Mode. Article markup helps search results show titles, images and dates, so it's cheap to add, but nobody has shown it causes citations.

Is there a free way to start?

Yes, on other platforms. Netlify Free costs $0 and GitHub Pages is available to free accounts on public repositories, WordPress.com has a free plan with ads, and Hashnode has one free publication. VeloCMS has no free plan, only a 14-day trial that takes a card.

Last verified: 2026-10-18

Sources

All accessed 2026-10-18, except where I say a page wouldn't open.