The AI Stack for a One-Person Newsletter
A solo newsletter dies for one of two reasons: the writing takes longer each week than the week before, or nobody new ever arrives. AI can compress the first — it cannot fix the second, and most stacks sold as 'newsletter automation' quietly assume otherwise. This is the six-stage chain that actually holds up weekly, including the two stages where using AI makes the issue worse.
The chain at a glance
Every arrow above is a file handing off to the next tool. Those handoffs — not the tools — are where this workflow usually breaks, so each stage below states exactly what format has to come out of it.
Stage by stage
Research
30 minFive to eight raw inputs for the issue — links, quotes, numbers — each with a source you'd be comfortable being quoted on.
Cited answers mean the research pass ends with links already attached. For a weekly issue you're running a handful of queries, which sits comfortably inside the free allowance.
Hands off as
A flat list of links with a one-line note on why each matters. Not a summary — the summaries are what you'll write, and pre-summarising here is how issues start reading like everyone else's.
Costs you an hour if you miss it
Researching the topic instead of the angle. Five sources on 'AI in hiring' produce a generic issue; five sources on 'why screening tools reject the same candidates twice' produce one worth forwarding.
Draft
45 minA structured first draft in your voice: an opening that earns the scroll, three to five sections, one clear takeaway.
Long-context drafting matters here because the fastest way to keep your voice is to paste your three best past issues and ask for the same cadence. Short-context tools force you to re-describe your style every week and it drifts anyway.
Hands off as
Plain text or Markdown, headings intact. Whatever you draft in, do the final formatting inside the email tool — pasting styled HTML from a doc is the single most common source of broken mobile rendering.
Costs you an hour if you miss it
Asking for a finished issue. Ask for the structure and the transitions, then write the opening and the takeaway yourself — those two paragraphs are the entire reason someone stays subscribed.
Edit
30 minThe draft cut by 20–30%, claims checked, and every sentence that could have come from anyone removed.
The useful function at this stage is tightening sentences you've already committed to, not rewriting them. Paraphrase-plus-grammar in one pass is faster than round-tripping the whole issue back through a chat model, which tends to re-inflate what you just cut.
Hands off as
The final text, one document, no tracked changes. Read the opening aloud before it leaves this stage; that is the whole quality gate.
Costs you an hour if you miss it
Letting a paraphraser touch quotes or statistics. It will smooth a number into a different number. Lock quoted material before the edit pass and re-check it after.
Swaps that work
- Claude — better for a structural edit — 'which section is doing no work?' — and worse for line-level tightening.
Visuals
15 minOne header image and, at most, two in-body graphics. More than that and the issue loads slowly and gets clipped by Gmail.
A locked template beats a better-looking one-off. The value at this stage is that issue 40's header looks like issue 1's, which is what makes the newsletter recognisable in a crowded inbox.
Hands off as
Web-sized JPG or PNG under 200KB each, hosted by the email platform rather than hotlinked. Total issue weight under 102KB of HTML or Gmail clips it and your call to action lands below the fold of a 'view entire message' link.
Costs you an hour if you miss it
Putting the takeaway inside an image. Most clients block images by default on first open, and screen readers get nothing — anything load-bearing has to be text.
Send
20 minThe issue delivered, with a subject line tested against a second option and a plain-text fallback that reads properly.
The free tier goes far enough that a new newsletter never pays before it has an audience, and referral and recommendation features are built in rather than bolted on — which matters because at this size, existing readers are your only distribution.
Hands off as
A public web version URL for the issue. That URL is what you promote in the next stage and what search engines and AI assistants can actually read — an email-only issue is invisible to both.
Costs you an hour if you miss it
Sending without opening the plain-text version. It's what a meaningful share of deliverability scoring reads, and a broken one quietly costs you inbox placement long before you notice opens sliding.
Swaps that work
- Substack — no monthly fee, takes a cut of paid subscriptions; the discovery network is real but you're renting the audience relationship.
- ActiveCampaign — the right choice when the newsletter feeds a product funnel and you need behavioural automation, not just a broadcast.
Promote
20 minThree to five posts across the week pointing at the web version — not one announcement on send day.
Scheduling is what converts 'I'll post about it when I remember' into a week of steady entry points. The free tier covers a solo operator's three surfaces, which is more than most newsletters use well.
Hands off as
Nothing downstream — but log which post drove signups. The one number that decides whether this newsletter grows is where new subscribers come from, and no tool in the chain will tell you unless you tag the links.
Costs you an hour if you miss it
Posting 'new issue is out' with a link. Post the single best idea from the issue as a standalone thought, and link underneath — the first version asks for a click, the second earns one.
Swaps that work
- Typefully — stronger if X is your primary surface — thread composition and analytics are better than a general scheduler's.
Paraphrase, check grammar, and cite sources in one pass — the edit stage is where a weekly issue gets its hour back.
What the month actually costs
Under 1,000 subscribers
$0
Perplexity free, Claude free, QuillBot free, Canva free, Beehiiv free to 2,500 subs, Buffer free for three channels. There is genuinely no reason to pay for anything in this chain until either the list or the cadence forces it.
Weekly at ~5,000 subscribers
≈ $50–70
Beehiiv's paid tier once you pass the free subscriber ceiling, plus QuillBot premium and Canva Pro for the template library. The email platform is the only line item that scales with the list; everything else is flat.
Twice weekly at ~10,000
≈ $120–180
Higher email tier, a paid Claude plan for the drafting volume, and usually a paid scheduler once you're on more than three channels. At this point the newsletter should be earning — if it isn't, the cost problem is a monetisation problem.
Prices are list prices checked against each vendor's public pricing page and change without notice. Annual billing typically cuts 15–20% off the paid rows.
Where this chain breaks
Nobody publishes the failure modes, so people quit at stage three assuming they did something wrong. These are the ones that show up on the first real run.
- •The draft stage swallowing the voice. Every week you accept slightly more of the model's phrasing, and around issue 15 long-time readers stop replying. The fix is structural: write the opening and closing paragraphs by hand, always, even when the middle is assisted.
- •Gmail clipping. HTML over roughly 102KB gets truncated with a 'view entire message' link, and everything below it — including the subscribe and reply prompts — effectively disappears. Heavy image use at the visuals stage is the usual cause.
- •The paraphraser rewriting quotes. Line-level editing tools do not know which sentences are yours and which are quoted, and a smoothed statistic is a correction email. Lock quoted material before the pass.
- •Deliverability decay after a growth spike. A batch of new subscribers from a giveaway or a viral post opens at a much lower rate, which drags your sender reputation and lands the whole list in Promotions. Suppress non-openers after three issues rather than mailing everyone forever.
- •Never publishing the web version publicly. If the issue only exists in inboxes, it can't be found by search or cited by an AI assistant, and the newsletter has exactly one growth channel — you, posting. The web version is what makes the archive an asset.
Common questions
How much of a newsletter should AI actually write?
The middle. Research, structure, transitions and the line edit compress well. The opening paragraph, the takeaway and any opinion should be yours — those are the parts subscribers reply to, and replies are both the growth signal and the deliverability signal. A fully generated issue reads fine and gets forwarded by nobody.
Can I run this whole stack free?
Yes, up to about 2,500 subscribers and one issue a week. Every stage here has a free tier that ships real work, and the email platform is the first thing that forces payment — which is the right order, because by then the list is worth paying for.
Beehiiv, Substack or an email platform like ActiveCampaign?
Beehiiv if the newsletter is the product and you want the referral mechanics without a revenue share. Substack if you want its discovery network and are comfortable with a cut of paid subscriptions and a weaker export relationship. ActiveCampaign if the newsletter is a channel into something else you sell and you need behaviour-triggered sequences, not broadcasts.
What's the realistic time saving?
Roughly two hours per issue against writing unassisted — most of it at research and the line edit, not drafting. Drafting time tends to shift rather than shrink: you spend it steering and rejecting instead of typing. The saving is real but it is not the 15-minute newsletter the tooling ads imply.
Which stage do people skip that they shouldn't?
Promote. Six of these stages produce the issue and one distributes it, and the one is where solo newsletters stall. If you only have 30 minutes spare this week, spend it on the promote stage of last week's issue rather than starting this week's.
Related reading
📬 Get the best new AI tools delivered weekly
One concise email with fresh launches, trending picks, and featured standouts.