Are newsletters still worth building in 2026?

Yes, and for a narrower reason than the one people usually give.

The common answer runs like this. AI is eating search traffic, social reach is unreliable, so build the one channel you own. The first two parts hold. The third skips over what happened to the inbox, which is now the most interesting part of the question.

Two things changed in the last two years. Subscribers got harder and more expensive to find. And the inbox itself acquired a machine layer that reads your work before your reader does. Neither of those breaks the case for a newsletter. Both change what you should build and how you should judge it.

What changed is the growth math

Getting people onto a list used to be cheap. Publishers were buying subscribers for cents through ads. Search and social sent enough traffic that a decent signup form did the rest.

At the 2026 Newsletter Conference, one question kept coming up in the hallways. Had subscribers become harder and more expensive to find? The answer in the room was yes, and few of the people saying it had an incentive to.

The platforms answered with recommendation networks. You pay another publisher for each verified subscriber, or you swap recommendations for free. It works, in the sense that the number goes up. The quality is a separate matter. Readers often do not register that they signed up for a second publication. They arrive with no memory of you and no reason to open. One operator watched a company add more than 10,000 readers in a month through paid recommendations. The founder read that growth as evidence of product-market fit. It was evidence of a budget.

Platform-level growth has also cooled. Substack passed 5 million paid subscriptions in March 2025. More than a year later, the figure was still five million and change. The media analyst Simon Owens reads that as deceleration rather than a boom.

None of that is fatal, and the distinction that matters is a question of business model. Adam Ryan of Workweek frames it as the difference between a "how many" business and a "who" business. A newsletter funded by advertising is a how-many business. In 2026 it is fighting for cheap scale that no longer exists. An expert practice is a who business. You need a few hundred of the right people. The economics of scale were not yours to begin with.

AI arrived in the inbox, not only in search

Most of the argument for newsletters treats email as the safe harbor while AI reshapes everything else. That was true for about a year.

In early 2026, Google brought Gmail into what it calls the Gemini era. Summaries now sit above your message, and the inbox answers questions in plain language. A Manage Subscriptions view ranks senders by how often they mail, and unsubscribes with one click. Microsoft's Copilot triages Outlook. Apple has been sorting and summarizing since 2024.

Apple and Gmail alone account for nearly 90% of email opens. That figure comes from Litmus, measured across more than a billion opens in July 2026. The layer that now reads your work first is not a fringe product. It is most of the inbox.

Think about what that does to the moment of arrival. A model writes a summary card above your letter. That card is the first impression your reader gets. If your issue makes one clear argument, the summary sells it. If your issue is a roundup of five half-thoughts, the summary says so. The reader then has what they need without opening.

Assistants take the next step. Media strategists now describe the zero-click inbox. An agent reads the morning's 30 newsletters, briefs its owner on three, and surfaces one worth opening. That is not a forecast for 2030. Copilot and Gemini are doing early versions of it now.

So the ownership claim has a boundary on it. You own the addresses and the permission, and no algorithm can take those. What you do not own is the reader's attention at the moment of delivery. That attention now passes through a machine with its own opinion about what deserves a person's time.

For an information business, the asset logic held

One thing survived intact. It matters most when you sell knowledge for a living.

A list is a record of people who asked to hear from you. You keep it, and you can move it to any provider. That is the same thing it was in 2019, and it is the one audience no platform can revoke.

Better than that, the machine layer rewards the thing an expert already has. Summarizers reward clear structure and an argument worth summarizing. Prioritizing systems reward senders whose mail gets opened, replied to, and rescued from the promotions tab. Agents surface what their owner has shown interest in before. Those signals favor the person publishing a considered idea to people who chose them. They punish the person publishing volume to readers who half-remember signing up.

The filtering that worries mass senders is a tailwind for a specialist. Few readers summarize their way past a letter from the one person they want to hear from.

Write for the reader and for the summary

Start one now and four things should be different.

Make one argument per issue. That was already the better format. It is also the one that survives compression into three lines by a model.

Write in clean prose with the substance in the text. Clever preheaders and image-heavy templates give a summarizer little to work with. Structure and plain sentences give it plenty.

Grow from your own published work rather than from paid recommendations. Every article, talk, podcast appearance, and search result you already produce sends readers who know why they subscribed. It is slower. The people arriving that way open your mail, and openers are what the inbox now scores you on.

Stop treating opens as the number. A summarized message may register as an open with no human involved. Apple's privacy features already inflated the figure. Replies, direct visits, and revenue still measure something. And list size stopped predicting the outcome some time ago.

When the answer is no

Four cases, because a newsletter costs more in attention than in money.

You have no steady source of readers. A list needs a front door, which means articles, a podcast, search visibility, or a stage. Without one, the list has no way to grow.

You do not want to write, or think out loud, every two weeks. A letter is a standing commitment. Six issues and then silence is worse than no issues at all. It teaches your best prospects that you start things.

You need demand next quarter. A letter compounds trust with people who already have a reason to care. It does not manufacture buyers on a schedule, and anyone selling it that way is selling you something else.

Your plan was advertising revenue. That plan needs the cheap scale the market stopped offering. The arithmetic now punishes a general-interest list of moderate size.

None of those four is a reason to write off the channel. Each is a reason to fix something first.

The picture is still moving

The inbox features described here shipped in the first half of 2026. Google is testing further agents. Microsoft keeps extending Copilot's authority over the inbox. Anyone claiming to know how this settles by 2028 is guessing.

What holds through that is unglamorous. People give their address to someone whose thinking they want. They keep giving attention to whoever repays it. Machines are getting better at deciding what deserves a person's time. Clear thinking on a subject someone cares about is difficult to filter out.

Start a letter if you will write something a summary cannot replace. That was the standard before any of this. The difference in 2026 is that something else reads you first and decides whether the person ever sees it.