# Email-first, but your site is the home of record: choosing a newsletter platform without surrendering your canonical credit

> A 2026 guide to choosing a newsletter platform, Substack, Beehiiv, or Ghost, without handing your SEO and AI-citation credit to a domain you don't own. Run email as a growth layer on top of your own canonical site, never as the home of record.
>
> https://pravda.systems/blog/email-first-newsletter-home · 2026-06-28

You write the best teardown you've shipped all year. You publish it where it's easiest, the newsletter platform with the slick editor and the built-in audience, and it does numbers. Then, three months later, you search your own topic and there it is on page one: your words, ranking, under a domain that isn't yours. Or you ask an AI assistant the exact question your post answers, and it cites the platform, not you.

> You did the work. Someone else banked the equity.

That's the quiet tax nobody mentions when they hand you a newsletter recommendation. The platforms are genuinely good at growing an email list from near-zero, which is the one asset no algorithm can revoke. The trap is letting the platform become the version of your work the web treats as the original.

```mermaid
flowchart LR
  A["You write the post"] --> B["Publish on the platform"]
  A --> C["Publish on your own domain"]
  B --> D["Platform domain ranks and gets cited"]
  C --> E["Your domain ranks and gets cited"]
```

This note is about that single decision: which email-first home to pick, and how to wire it so the credit, the audience, and the citations stay on a domain you control. For the broader question of which channels pay you in leads versus citations, that's the companion playbook: [Where to publish for leads and AI citations](/blog/where-to-publish-for-leads-and-ai-citations).

## What's the real decision here?

**Pick exactly one email-first home, run it as a growth-and-engagement layer on top of your own site, and never let it become the canonical version of your writing. The platform's job is to bring you readers you couldn't reach alone. Your site's job is to be the original that ranks and gets cited.**

The honest first question isn't "which platform is best." It's whether you should open *another* surface at all. If you already have a site that converts, standing up a parallel newsletter stack beside it just fragments thin effort across two places. The way out of that contradiction is discipline: one email-first home, fed by your existing reach, pointing relentlessly back to your canonical articles. Not three half-tended publications.

Whichever platform you choose, the same backbone applies:

- Grow it on a steady cadence.
- Write consistently.
- Segment your list by what people actually engage with.
- Treat email as a relationship rather than a broadcast.

That part doesn't change between Substack, Beehiiv, and Ghost. What changes is one thing that decides everything downstream: **whether the platform lets you keep the canonical credit, and what it costs you to protect it.**

## Why does a missing canonical tag quietly hand away your credit?

**Two of the three contenders give you no way to tell search engines "the real version lives on my domain." Publish the same text on your blog and on the platform's archive, and you've created a duplicate-content problem with no signal saying which is the original, so the higher-authority domain can win, and it usually isn't yours.**

The technical lever is `rel=canonical`, a tag that points one URL at another as the authoritative source. [Medium](https://medium.com)'s import tool sets it automatically. [Substack](https://substack.com) doesn't let you set it at all. The canonical URL for any Substack post is effectively stuck on the Substack URL. [Beehiiv](https://www.beehiiv.com) exposes index/noindex controls but no per-post canonical pointing at your own domain either. So on both, pasting your full article word-for-word hands a higher-authority domain a copy that can outrank your own.

```mermaid
flowchart TD
  A["Same article on your blog and the platform archive"] --> B["No canonical signal saying which is original"]
  B --> C["Higher-authority domain can win"]
  C --> D["Usually not yours"]
```

The fallback everyone reaches for, an "originally published at →" line, is a *soft* signal, not a hard canonical. Search engines can still favor the higher-authority domain when the content is essentially identical, and some now ignore canonical tags altogether and pick their own representative URL. That uncertainty is the whole reason not to gamble with a clone.

This matters double for [AI citations](/blog/how-to-get-cited-by-ai-answer-engines), and the data is blunt about where they land:

> **Citations accrue to whatever domain hosts the content, not to the platform brand.** By one reported 2026 analysis of a large prompt dataset, Substack content accounted for a vanishingly small share of AI citations (well under a tenth of a percent), and the same analysis noted that [Ghost](https://ghost.org) has no platform-level citation metric at all, because Ghost citations land on each creator's own domain rather than a shared one.

Read that together and the strategy writes itself. The citable, rankable thing is the *domain*. If your own site is already the canonical home, the platform is a list-building decision, not a citation strategy. The engines were always going to cite your domain, as long as you don't undercut it with an indexed duplicate.

So the safe move differs by platform, and getting it right is most of the game:

```figure src=platform-canonical-comparison.webp w=1536 h=1024
Three newsletter platforms compared by canonical support, noindex control, and the safe move to protect your site's credit.
```

| Platform | Canonical to your site? | Can you noindex its web archive? | The safe move |
|---|---|---|---|
| **Substack** | No | No | Send a differentiated **cut** (~40–60% of the full piece) + a "first published at" link |
| **Beehiiv** | No | Yes ("Remove Indexing") | **Noindex the archive**, then send the full text in the email (private email isn't indexed) |
| **Ghost** | It's your own domain | You control it | No conflict, *unless* you go headless, which breaks membership (see below) |

That table is the spine of the whole decision. Everything else is detail hanging off it.

## Substack, Beehiiv, or Ghost: the one-screen comparison

**Each platform is excellent at exactly one thing and exposes you on exactly one thing. Substack gives you discovery you can't replicate but no ownership of the SEO or the paid relationship, Beehiiv gives you operator-grade growth tooling but the same canonical gap, Ghost gives you total ownership but you run the infrastructure, and the one feature you came for breaks the moment you try to keep your own front end.**

| Platform | Distinct strength | The failure mode | Platform fee on paid revenue |
|---|---|---|---|
| **Substack** | Networked discovery, recommendations, Notes, "subscribe to similar" bring new readers | You don't own the list *relationship* (paid subs don't migrate) or the SEO (no canonical) | **10%** + Stripe (~13–16% effective) |
| **Beehiiv** | Operator growth + monetization tooling, referrals, Boosts, ad network, deep analytics | The same canonical gap: its archive can compete with or outrank your own site | "Media company" model (ads, referrals, subscriptions) |
| **Ghost** | You own everything, content, list, code, custom domain, **0% fee** | You run the infrastructure, and going headless to own your front end breaks native membership | **0%** (your own Stripe) |

Pick by which exposure you can live with. If you need readers you can't otherwise reach, Substack's network is the answer and the canonical discipline is the price. If you want growth and revenue levers under one roof, Beehiiv, with indexing locked down. If owning the whole stack matters more than a discovery network, Ghost or a thin layer on your own site. The rest of this note is each platform's distinct case, then the craft that applies to all of them.

## Substack: the discovery engine nothing else replicates

**Substack's networked discovery is the one capability a bare email tool simply cannot give you, which makes it the best single place to grow an email audience from near-zero. The catch is that the same engine is what makes leaving expensive, and it surrenders both your SEO and your paid-subscriber relationship to a domain you don't own.**

A standard email service delivers to people who already found you. Substack actively introduces you to people who haven't. The recommendation system is the quiet workhorse: writers curate publications that get surfaced during the subscribe flow, and readers now see *many* recommended publications instead of just three, so cross-pollination jumped. Those recommendations appear in more than one place:

- during signup
- on your publication page
- inside digest emails

Then there's Notes, which stopped being a Twitter clone and became a discovery feed: the Home tab shows notes from writers you follow *and* writers they recommend, plus related notes, built to put creators you've never heard of in front of you.

> By late 2025, one creator cited Substack's own number: **32 million new subscriptions** generated from inside the app in a single three-month stretch, with recommendations and Notes "actually working" as a discovery engine.

That's why working the feed isn't vanity activity layered on top of the real work. On Substack, a meaningful share of distribution *is* the feed:

> Per Substack's own reported figures, roughly **40% of all subscriptions and 15% of paid subscriptions** now originate inside Substack (the feed, Notes, and recommendations), not from anything you link to externally.

The 2026 algorithm rewards continuity of behavior (did the reader keep reading, stay engaged, and remain inside Substack) over raw reach. The signal hierarchy is unusually legible once you lay it out, and it tells you exactly where to spend effort:

```figure src=substack-signal-strength.webp w=1536 h=1024
Five Substack algorithm signals ranked from weakest to strongest, showing where to spend effort to open the discovery graph.
```

| Signal | Strength | Why it matters |
|---|---|---|
| Impressions → scroll-past | Weakest | Mostly tells the system what people ignore |
| Likes | Low | A soft acknowledgement, a weak predictor of subscriptions |
| Comments, saves | Medium | Comments take thought, threads a few exchanges deep count as strong |
| Restacks, clicks to your post | Strong | A restack is a "trust transfer", a click signals content worth opening |
| New free / paid subscriptions | Very strong | The outcome the system is actually optimizing for |

You work two native surfaces, and they do different jobs:

| Format | Best job | Length / cadence |
|---|---|---|
| Long-form Post (the newsletter cut) | Deep teaching, converting readers to subscribers | 800–1,600 words, 1–2× / week |
| Mini-series Post | A complex topic split across reading sessions | 2–3 parts, 800–1,200 words each |
| Note: text | Fast insight, conversation, discovery | 1–4 short paragraphs, 1–3× / day |
| Note: image / media | Visual explanation, higher reach | a screenshot or diagram + 1–3 sentences |

That last row is an open goal for a technical writer. Image and media Notes outperform text-only ones. One analysis found they drew the most engagement in its dataset. A redacted architecture diagram or a clean code screenshot is exactly the kind of artifact that travels, and you already produce them. Mine each long-form piece for 3–7 Notes and spread them across 7–10 days. The multiplier when your audience is small is engagement *outward*: restack other writers with commentary, reply with something substantive. A Note that earns a restack from a writer with an audience borrows that audience for free.

```mermaid
flowchart LR
  N["Note"] --> P["Substack post"]
  P --> S["Your site"]
  P --> Sub["Reader subscribes"]
  Sub --> G["Opens the recommendation graph"]
```

In Notes, link to your Substack post, not directly to your website. The two-step path (Note → Substack post → your site) keeps the reader inside the network long enough to subscribe, and that subscription is the signal that opens the recommendation graph to you. A Note that fires someone straight off-platform spends your reach on a single click. Then close the loop with the data: Growth → New Subscribers → Sources → Substack → Notes shows you which Notes actually produced subscribers, so you double down on what converts instead of what racks up likes.

On cost, Substack is free to start, cheap to run, and stickier than it looks to exit:

| Cost | Amount |
|---|---|
| Platform fee on paid subscriptions | **10%** |
| [Stripe](https://stripe.com) processing | ~2.9% + $0.30 per transaction, plus a ~0.5–0.7% recurring-billing fee |
| Effective take on gross paid revenue | **~13–16%** |
| Custom domain (drops `substack.com` from the URL) | **$50**, one-time |
| Free subscribers / hosting | $0 |

The lock-in is subtle and worth naming precisely. Your email list exports as a CSV in one click, but the *monetization relationship* doesn't. Leave, and your paid subscribers revert to free readers you have to re-onboard into a new billing system, and you lose the in-network discovery. So 13–16% on an optional future "backstage" tier is fine upside for [a services business](/blog/productized-ai-service-offer) that earns through projects, not the newsletter, but keep high-ticket offers anchored to your own site, not to Substack's fee rails, and export the addresses periodically so the asset is always unambiguously yours.

## Beehiiv: operator tooling with the same canonical gap

**Beehiiv was built by ex-[Morning Brew](https://www.morningbrew.com) people for "operator"-type publishers who care about growth levers, monetization, and analytics, not just sending pretty emails. It's the pick when growth and revenue infrastructure are the goal. But it carries the identical canonical gap as Substack, and the fix isn't a tag you don't have. It's noindex.**

Reviews consistently place Beehiiv ahead when growth and monetization matter (referral program, Boosts, recommendation network, ad network) and place Substack, Ghost, and [ConvertKit](https://kit.com) ahead for simplicity, deep customization, or complex automation. One 2026 analysis argues Beehiiv beats Substack specifically for B2B founders, because it's optimized for list growth and revenue infrastructure where Substack optimizes for a social feed.

### The fix is noindex, not a canonical tag

Here's the move, made concrete. Beehiiv has no per-post canonical (the canonical chapter covered why), only a global "Discoverable on the web" toggle and a "Remove indexing" option, so with a full copy the noindex toggle is your only protective lever.

Beehiiv's *own* guidance for people who already run an external blog is explicit: enable "Remove Indexing" so the Beehiiv site isn't crawled, which avoids duplicate content and protects your main site's SEO. Because duplicate-content penalties apply to indexed web pages, not to private email content, search engines then only ever see and rank your site. Structure each issue simply:

1. An attribution block up top ("the field-notes version of a system originally published on your-site →").
2. A short context intro.
3. The full body text (safe, since it's private email).
4. A CTA pointing home.

(If you ever decide you *want* the Beehiiv pages indexed for their own discovery, you can't ship near-duplicates, you'd make that version a shorter, reworded cut, the recipe the formatting chapter below covers. It's strictly more work, so default to noindex.)

The growth and monetization machinery is real but built for scale, so most of it is upside to tap later, not the engine that gets you off zero:

| Tool | What it does | At a small niche B2B list |
|---|---|---|
| **Referral program** | Rewards readers for bringing readers, near "set and forget" | Worth turning on early with *modest* rewards, the social proof matters more than the economics |
| **Recommendation network** | Cross-promotes newsletters at subscribe time | Thrives in big consumer verticals, in a narrow niche the pool is smaller, so treat as upside until you find 3–5 overlapping technical newsletters |
| **Boosts** | Pay other newsletters a CPA to recommend you | a low single-digit dollar CPA per active subscriber (Beehiiv-reported), only sensible once you know subscriber LTV |
| **Ad network** | Sponsor placements paid on verified opens/clicks | Beehiiv reports it pays out over $1M/month across all publishers, but a small list sees only modest payouts (a rough estimate, not a guarantee), and generic ads clutter technical content |

The load-bearing point for a solo engineer selling high-ticket work: you will almost certainly make far more per subscriber by routing readers into services than by running generic ad inventory. The platform-wide monetization numbers are impressive and exactly why they're a trap if you read them as a to-do list:

- **Per Beehiiv's self-reported 2026 "State of newsletters," paid subscriptions on the platform more than doubled from $8M to $19M between 2024 and 2025.**
- A median time to first dollar of about **66 days** for 2025 launches.
- Creators combining ads, Boosts, and subscriptions earning **roughly triple** single-channel creators.

All true, and all premised on an audience you don't have yet. The reported Boosts case studies that show big returns, $840 in ad revenue and 1,200+ subscribers over 90 days in one 2025 example, come from broad consumer verticals, not a narrow B2B niche.

The discipline that protects you from day one is acquisition tagging:

Tag every subscriber by source from the first send:

- `source:pravda`
- `source:linkedin`
- `source:x`
- a reserved `source:boosts_future`

This is what later tells you whether a paid channel sends you buyers or junk. Without it you can't isolate a paid cohort's engagement from your organic readers'. Then, once you have three or four issues and real engagement data, run exactly one experiment: a tiny Boosts test of **$50–$100** against one or two adjacent automation/engineering newsletters, and watch that tagged cohort like a hawk before spending another dollar.

Lock-in, as with Substack, isn't your list (it exports as a CSV with custom fields and stats). It's the growth network, Boosts, referrals, recommendations, which doesn't follow you out. Lean on it for acceleration, but make sure your owned reach is what's actually compounding, so a future move costs you a channel rather than your audience.

## Ghost: you own everything, until headless breaks membership

**Ghost wins the ownership contest decisively: a custom domain, a 0% platform fee, full access to your list and your Stripe relationship, open-source code you can self-host. But the one new thing it offers someone who already has an optimized site, integrated membership, only switches on if you make Ghost your main public site. Try to keep your own front end, and the feature you came for breaks.**

### What the bundle gives you

Take the bundle seriously, because it's well-built. Any Ghost post can be sent as email to everyone, to free or paid members, or to a segment, using a built-in template. Since Ghost 5.x it also runs multiple newsletters, premium tiers, offers, and audience segmentation, all in Ghost Admin, with analytics that break paid members down by cadence and tier. The ownership story is just as clean: Ghost is open-source under the MIT license, so you own the content, the list, and the code. On revenue it takes **0%**, you connect your own Stripe and pay only Stripe's processing charges. The SEO half (sitemaps, canonical tags, clean permalinks, schema.org data, Open Graph cards, redirect/noindex controls) is rated excellent in 2026 reviews, but if you've already built that yourself, it's the same capability a second time, not an upgrade.

### Where headless breaks it

Here's the crux the rave reviews skip. **Ghost's membership lives in its frontend, not its API.** Run Ghost as a headless backend behind a separate Next.js site, and the Stripe paywall, Portal, comments, offers, and member analytics have nothing to attach to. Ghost's own docs say membership is not compatible with a headless/JAMstack setup. Worse, to avoid serving duplicate content, Ghost recommends marking its own frontend "Private," which switches off its SEO features and can break unsubscribe and Portal flows.

```mermaid
flowchart TD
  H["Run Ghost headless behind your own front end"] --> S["Ghost SEO or a headless front"]
  H --> M["Ghost membership or a headless front"]
  S --> X["You cannot run Ghost's paywall through your own site"]
  M --> X
```

So you can have Ghost's SEO *or* a headless front, Ghost's membership *or* a headless front. What you cannot have is Ghost's paywall running through your own site. That leaves two real options: move most of your site *into* Ghost and live on its frontend, or run a fragile dual stack with split UX and split analytics, both at odds with keeping one canonical surface everything resolves back to.

On pure ownership, the three sort cleanly. Ghost gives you the content, the subscriber list, and even the MIT-licensed code, all on your own domain, which is exactly where the SEO and attribution value then lands. Substack gives you an exportable list but sends the attribution to substack.com unless you pay for a custom domain. Beehiiv gives you the list, but on its hosted platform. (The fee differences are in the comparison table above.)

And the cost of *running* Ghost depends entirely on the path:

| Path | What you pay | What you operate | Newsletter sending |
|---|---|---|---|
| **Ghost(Pro)** | Managed fee, tied to member count (reported ~9–35/mo up to ~199/mo at 10k) | Nothing, it's managed | Included |
| **Self-host Ghost** | Free software + a VPS, commonly 5–25 USD/mo | Node.js, database, backups, Ghost updates | **[Mailgun](https://www.mailgun.com) required**, basic SMTP can't bulk-send |
| **Next.js + email provider** | Provider API fees, tens of dollars/mo at ~10k subscribers | Just API calls and a DB table | Your provider |

Note the self-host footgun: "free and open-source" still means a VPS, a database, ongoing Ghost upgrades, *and* a Mailgun account before a single newsletter goes out. Ghost earns its place when the business itself is a publication, multiple newsletters, multiple tiers, offers and trials, and you'd rather buy an integrated stack than assemble one, and you're willing to let Ghost be the front door. If publishing *is* the product, that's coherent. If publishing is the marketing for some other product, the headless incompatibility usually makes it redundant-plus-complexity. In that case, build a thin layer on the stack you already run:


1. **Phase 1, free list, no paywall.** Pick a developer-friendly provider, [Buttondown](https://buttondown.com) (Markdown-first) or [Loops](https://loops.so) (simple campaigns), and POST signups to its REST API, letting it own double-opt-in. For a fully-owned route, deploy [Listmonk](https://listmonk.app) or Keila, or a Next.js + [Resend](https://resend.com) waitlist that stores addresses in Postgres.
2. **Phase 2, membership, once demand exists.** Add auth with [NextAuth.js](https://authjs.dev) or [Supabase](https://supabase.com) Auth, wire Stripe subscriptions for one "membership" product, and gate premium content with server-side checks on session + subscription status. Everything, SEO, content, members, attribution, stays on your domain.


## How do you format each issue so the credit stays home?

**The newsletter cut is its own craft: a tighter, inbox-built version, not a copy of the blog post. Publish the complete canonical piece on your own site first (self-referencing canonical, timestamp of record), then on the platform ship the differentiated cut where you can't noindex, or the full text in private email where you can. The structure of that cut is the rest of this chapter.**

Aim for 800–1,600 words built for the inbox, with short paragraphs, a cold-open hook, and a single clear ask. The structure that holds attention:

1. **Cold open (2–4 sentences).** Make the value explicit. A reader should instantly know who you are, what this is, and why it matters.
2. **Positioning line.** One short, recurring, voice-forward paragraph: "I build and debug automation systems for B2B teams."
3. **Compressed value (3–5 sub-sections, ~150–250 words each).** A scenario, the pattern or failure it exposes, then a concrete prescription.
4. **TL/DR (3–6 bullets).** "If you only do three things, do these."
5. **One precise question.** "Where is your automation stack most fragile right now?" Replies are a strong signal.
6. **One primary CTA.** A single button after the value, generic "subscribe for more" underperforms a value-aligned reason.

```figure src=fig-newsletter-issue-structure.webp w=1536 h=1024
The six-stage structure of one newsletter issue, from the cold open through the single call to action that sends the reader back to your own site.
```

On link placement, there's no published evidence Substack throttles external links, so place them where they're useful: one text line after the hook, one to three contextual links in the body, one primary button. Don't build a link farm, but make the path home obvious.

## Will the inbox even deliver in 2026?

**The infrastructure is solid on every platform, but the bar for staying out of spam has risen sharply, and it now bites small lists too. The technical setup is mostly handled for you. The list discipline is not, and it's non-negotiable.**

[Gmail](https://www.gmail.com)'s bulk-sender rules expect spam-complaint rates below **0.10%**, and hitting **0.30% per 1,000 emails** can trigger blocking or leave you "ineligible for mitigation" until you sustain lower rates for several days. On a 500-to-1,000-person list, that's roughly one complaint starting to matter. Who handles the plumbing varies:

| Platform | Who handles deliverability |
|---|---|
| **Substack** | Manages SPF, DKIM, DMARC, and one-click unsubscribe for you |
| **Beehiiv** | Expects you to configure a custom *sending* domain with SPF/DKIM/DMARC so your reputation sits on your domain, not its shared one (the shared default domain is the single most common own-goal) |
| **self-hosted Ghost** | Requires Mailgun |

The discipline is identical everywhere:

- never import a cold or scraped list
- keep sending volume reasonable
- prune unengaged subscribers
- optimize for engaged opt-ins over raw size

A tight welcome email that asks the reader to reply or whitelist you lifts inbox placement *and* hands you copy inputs from the people you write for.

And stop grading yourself on open rates. Substack has told creators outright that email is becoming less deterministically reliable, pixel blocking distorts opens while clicks and paid conversions stay stable. Judge content by clicks, replies, and what readers do back on your site.

## What does the first 30 days actually look like?

**One flagship issue a week, a daily habit of short posts, and a deliberate push into the platform's discovery surface, in that order. Any single post barely matters, the machine is what compounds. The realistic target is your existing reach converted into the first 50–100 subscribers, with zero paid growth.**

- **Week 1, setup and first issue.** Nail your positioning, configure deliverability (and, on Beehiiv, flip "Remove Indexing" on), build your attribution-safe template once, and ship the first flagship: full on your site, a differentiated version on the platform. Announce it on [LinkedIn](/blog/linkedin-b2b-lead-generation-for-technical-founders) and [X](/blog/x-twitter-growth-system). Start the daily short-post habit.
- **Week 2, discovery.** Keep posting. Identify 5–10 complementary publications and enter their graph (Substack recommendations, or Beehiiv's network once you've found genuine overlaps). Reach out to 2–3 with a specific, non-spammy mutual-recommendation proposal. Ship a second flagship. Realistic target: **50–100 organic subscribers**.
- **Week 3, nurture.** Tag subscribers by what they click. Refine the welcome email from replies. Put a "reply with your biggest automation bottleneck" CTA in every issue, it trains engagement and feeds future case studies.
- **Week 4, scale what hits.** Turn your best section into a recurring series. Prune recommendations to your ICP. Add gentle service CTAs that keep monetization off the platform's rails while building a pipeline. Hold off on Boosts, ads, and paid tiers entirely.

## What kills it?

**The same handful of own-goals, made over and over, each easy to commit by accident and expensive to undo.**

```figure src=own-goals-taxonomy.webp w=1536 h=1024
Four newsletter own-goals arranged by the axis they violate: canonical credit, discovery signals, feed strategy, and audience ownership.
```

- **Dumping the full article 1:1 on an indexed platform with no canonical.** You're letting the engines decide which domain ranks, and it might not be yours. Send a differentiated cut, or noindex and send the full text in email.
- **Chasing likes.** They're soft signals. Comments, restacks, and subscriptions move you.
- **Shipping the big post and skipping the feed.** That forfeits the large share of growth that comes from inside the app, the easiest mistake for a solo builder.
- **Stacking CTAs.** One primary ask after the value beats five scattered ones.
- **Over-investing in Boosts and ads at near-zero.** Scale tools. Your early hours pay off more in better writing and GEO'd service pages.
- **Letting the platform become your "main website."** It's not a CMS replacement for a technical product/services site. Keep your own domain canonical.
- **Building your core business on platform payments before validating fit.** Paid subs don't migrate cleanly, so early monetization should bias toward off-platform services.

## The bottom line

**Make the platform your growth engine and your own site the home of record. Publish the full piece on your domain first, send a version with your name and a link on it, work the discovery feed by hand, and pull the email addresses out as a CSV every so often so the asset is always, unambiguously yours.**

Three reported results from 2025–26 show the shape of what works, and they map onto any technical niche:

- A self-directed-learning newsletter grew past **10,000** email subscribers (roughly **9,000** of them organic in 2025) on a simple rhythm, 1–2 in-depth issues a week plus a Note every couple of days, with most growth coming from the platform's built-in network rather than external SEO.
- A technical machine-learning newsletter sat flat at about **15** paid subscribers for seven months, then jumped to about **115** paid over four months, not from more effort, but from keeping the free content deeply technical while shifting the paid tier toward career strategy and compensation, the meta-value readers paid for. The lesson: build trust with free depth, if you add a paid tier later, charge for the meta-value, not just more code.
- And the pattern that should change your behavior: that large in-feed share from earlier means a big chunk of discovery now happens inside the platform, not from links out. For every article, think "one issue plus a small internal campaign of posts," never "one cross-post with a link out."

Substack if you need a discovery network nothing else replicates. Beehiiv if you want operator-grade growth tooling, with indexing locked down. Ghost or a thin owned layer if owning the whole stack beats the network. Whatever you pick, the rule is the same one you started with: email-first, but your site is the home of record. Run it that way and you buy networked growth a near-zero account couldn't get any other way, without handing the credit, or the audience, to a platform you'd have to rebuild from if you ever walked away.