Where to publish in 2026: the platforms that turn technical writing into leads and AI citations

Yes. Post the whole thing everywhere that has an audience, with your name and a link home on every copy. The "duplicate content penalty" you're worried about isn't real — and at your stage, duplicating is how your site earns the authority to rank in the first place.

There is no penalty for publishing the same text in two places. What search engines do instead is pick one version to show and quietly filter the rest — they call it canonicalization, and nothing gets punished. The only real risk is subtler: with no signal pointing to the original, a high-authority platform can become the version that ranks and gets cited, so the credit lands on the platform instead of you. But read that carefully — it requires you to have a ranking worth stealing. From zero, you don't. There's nothing to cannibalize.

Here's what you do have. Every full copy on Medium, Dev.to, LinkedIn, or Substack carries a link back to your site — and a link from a high-authority domain is a backlink, which is precisely the raw material a new site needs to become rankable at all. So early duplication isn't a leak in the bucket. It's the pump that fills it. The platforms' authority is what bootstraps yours.

The one rule that is non-negotiable: your name and a link home on every single copy. Get that right and the reach, the attribution, and the lead all trace back to you no matter which copy someone reads. A buyer who finishes your full post on Substack with your contact under it is a lead you can close — the site converts a little better because it has the proof and the CTAs, but "a little better" is not "required."

Because it's free insurance, not because copying is dangerous. Where a platform offers it, set it and post the full text. Where it doesn't, post the full text anyway — at this stage your name and your link do the same job.

A rel=canonical tag tells engines "the real version lives here." Some platforms hand it to you for nothing: Medium's Import tool sets it automatically and even backdates the copy; Dev.to and Hashnode take a one-line canonical_url in the frontmatter. On those, a full identical copy plus the tag is strictly the best move there is — maximum reach, zero risk, and no reason to rewrite a word.

Substack and LinkedIn don't offer the tag, and that turns out not to matter much for you. LinkedIn lives mostly behind a login, so it's barely indexed — there's little SEO there to cannibalize in the first place; what matters on LinkedIn is the audience, not the search ranking. Substack is an email channel people read in their inbox, not something they find through Google. On both, a full copy that opens with "first published at [your link]" keeps your name on the work while you take the reach.

The single discipline that costs nothing and protects your future: publish on your own site first, even by an hour. When your rankings eventually arrive, yours is the version that's been the original all along.

The platforms that pay you in qualified buyers are mostly not the ones that pay you in AI citations. You need both columns, and they ask for different things from you. For B2B, LinkedIn is the lead machine and X is the credibility engine.

Reported 2026 benchmarks put LinkedIn at roughly 80% of B2B social leads, converting visitors to leads at about 2.74% — close to 4× X or Facebook. X has slipped to about 12.7% of B2B social leads (down from ~32%), converting at around 0.69%. That reads like a verdict against X until you remember they do different jobs: LinkedIn is where a decision-maker vets a vendor, X is where a founder builds a name and catches buyers in the act of asking for help. One converts; the other compounds.

The numbers are reported figures from 2026 GEO and B2B research, not my own measurements — treat them as the shape of the terrain, which is consistent across sources: leads cluster on LinkedIn and X; citations cluster elsewhere.

Engines quote a narrow, predictable set of surfaces — your own structured pages, Reddit threads, technical docs and GitHub, and YouTube transcripts — weighted toward recent, multi-sourced, concretely-stated claims. If you want to be the answer when a buyer asks Perplexity or ChatGPT who to hire, you write for those surfaces, not for follower counts.

The specifics, as reported, are unusually actionable. Reddit is the surprise heavyweight — by one count around 47% of Perplexity's top-10 cited sources are Reddit threads. Recency is a real signal — Perplexity is reported to boost content updated within the last 30–90 days, so a note you refresh quarterly beats a better one gone stale. Corroboration multiplies citations — claims that appear across 5+ domains reportedly earn about 67% more citations than single-source ones, which is one more reason to put the same work on Medium, Dev.to, and GitHub: each domain that echoes you, with a link home, raises the odds the engines quote you. And the bar is high — ChatGPT Search and Perplexity reportedly name only 2–8 sources per answer, and AI citations concentrate in roughly 50 top domains. There's no long tail to hide in; you're competing for a top-eight slot, which is exactly why being on more surfaces, more recently, with your name attached, is the whole game. (For the on-page mechanics — answer capsules, llms.txt, structured data, the freshness discipline — see the companion note, How to get cited by AI answer engines in 2026.)

On most platforms the content carries itself, and liking strangers' posts does nothing for your reach — so don't. On exactly two, Reddit and X, the content has to take the form of a reply or an answer, and there the activity isn't optional. It's the only way in.

Pull "engagement" apart into three things, because they're not the same and only one of them is a waste of time.

Vanity engagement — liking other people's posts, leaving generic comments to be seen. Useless, on every platform. The algorithms reward what your content does — dwell, replies, saves — not what you do to someone else's. Spending time hitting like to "build presence" is a myth you can drop entirely.

Discovery-platform engagement — and this is the one people underestimate. On Reddit and X you cannot drop a link and leave: Reddit removes it, X buries it. The unit that actually gets seen — and gets cited by AI engines — is a genuinely useful answer inside someone else's thread: a deep Reddit comment, an X reply to someone asking for exactly what you do. That isn't vanity. It's content, delivered where the audience already is, and on Reddit it's also the price of being allowed to exist — communities enforce value-to-promotion norms around 9:1 or 10:1 and ban accounts that just self-promote. So the engagement is the content there. It's the same writing work you already do, living in a comment instead of a blog post.

Replying to people who engage with you — cheap, a few minutes, and it's where a comment becomes a DM becomes a client. Do this on every platform. It's not community theater; it's sales.

So: publish wide, engage narrow, and never spend a minute on likes-for-likes.

Build a publishing robot for the platforms where the artifact does the work, and keep your hands on the two where it's a conversation. Automating the conversation is how you get banned.

The split is clean once you see it:

Publishing is the cheap half. One markdown note → pushed to Medium (import + canonical), Dev.to/Hashnode (canonical_url), and a GitHub repo with links pre-set. The marginal cost is near zero, which is the real answer to "why not use all of them" — for publishing, you can and should use every one, because spreading wide costs you almost nothing once the script exists.

The engagement half can't be automated, but it can be made smart instead of endless. 🔒 authorize for free to reveal 85 words That's the line — automate the finding and the publishing; keep the writing of the reply human.

Publish on your site first, fan out the full copies within a day, post native to LinkedIn and X the same week, then keep answering on Reddit and X. One note, many surfaces, your name on all of them.

For each new note:

1. Publish canonical on your site — structured headings, tables, internal links, in the sitemap and llms.txt. This is the version everything resolves to, and it goes live first.
2. Within a day, auto-syndicate the full copies — Medium (import, auto-canonical), Dev.to/Hashnode (canonical_url set), and a GitHub repo or README that mirrors the note and links back. These are the corroborating domains that drive the citation lift.
3. The same week, post native — a LinkedIn post (with the link in the first comment so the algorithm doesn't throttle reach) and an X thread that extracts the sharpest part, both linking home.
4. Ongoing, by hand — answer real questions in your two chosen subreddits and reply to buyer-intent on X, linking your note only when it genuinely helps. Reply to everyone who engages.

Run it weekly. The single post never matters; the machine does. After a month you've stacked four canonical notes, eight or so canonical-safe copies feeding your authority, a string of GitHub artifacts, a steady LinkedIn and X presence, and early trust in a subreddit or two.

Publishing is cheap, so the only thing worth guarding is your attention. Three places drain it without paying you back.

• Vanity engagement — covered above. Liking strangers buys you nothing. Zero minutes.
• Closed Slack / Discord / Skool as publishing channels — their content is invisible to search engines and AI crawlers, so nothing you post there is ever indexed or cited. They're for relationships and warm intros, not distribution. Use them that way, and don't mistake activity there for reach.
• A second full newsletter stack — you already have your site and a LinkedIn newsletter. Standing up a parallel Beehiiv or Ghost operation beside them fragments your effort for little citation gain. If you want one email-first audience, pick a single platform and run only that.

And the meta-trap: trying to deeply engage on more than two communities at once. Publish everywhere; be genuinely present in two. Spread your human hours thin and you're forgettable in ten places instead of trusted in two.

Publish the same work everywhere that will have it, with your name and a link on every copy — that's how a new site earns the authority to rank and get cited. Set the free canonical tag where it's offered, publish on your own site first, and let a robot handle the spread. Then take the hours that saves you and spend them where the content is a conversation: answering the buyer on X who just described your exact service, and the person on Reddit asking the question your note already answers. Most people will keep hoarding their best work behind a caution they haven't earned. You won't — and in a year the engines and the buyers will both know your name, because it was on everything.