The missing half
Somebody said yes. The message didn't arrive.
That is, more or less, the whole story of what changed about owned media between 2020 and 2026.
Seth Godin wrote Permission Marketing in 1999 and gave us the first half of the framework — the idea that you build an audience by earning explicit permission to show up in someone's attention. He was right. What he couldn't fully foresee, and what most of us missed for about fifteen years after reading him, is that permission could stay perfectly intact while the delivery mechanism quietly broke around it. A yes given is not a message received. Those turned out to be two separate facts.
Subscribers and followers, by the way, are the same gesture at different strengths. Both are people who, at some point, performed an act that meant I want to hear from you. Handing over an email and confirming a double opt-in is a heavier gesture than tapping a follow button — the way agreeing to meet someone at a party is a different commitment from inviting them into your home. But both are a yes. They are not two categories. They are points on the same scale.
What changed between 2024 and 2026 is not the permission. It is whether anyone is still delivering the mail.
A note on compounding
Audience isn't the only thing that compounds in marketing. Reputation does. Brand distinctiveness does. The body of work you keep getting better at does. Trust with specific people in specific rooms does. These are related but not identical — you can have reputation without an audience (the niche thinker everyone quietly respects) and an audience without reputation (the bloated mailing list nobody reads). What they share is the quality of gaining value while you're not actively spending money on them, which is rare enough in marketing that the ones you can measure tend to get reported in dramatic tones. The drama is overstated. The phenomenon is real. Audience is one compounding thing among several. Worth saying out loud so the rest of the post doesn't have to pretend otherwise.
The old deal
Through most of the 2010s the owned media playbook was four moves stapled together. SEO blog, email list, social, occasional gated content. Every marketer I knew ran two books at the same time — a glamorous one for rented reach and a small, quiet one for the people who had said yes. The glamorous book got the budget because the reporting was louder. The small quiet book got the compound interest.
This worked because of a hidden assumption nobody bothered to examine: that when someone said yes on any channel, the channel would deliver your message to them. Google would send the people who typed your thing into a search. LinkedIn would show your post to your followers. Gmail would drop your email into the primary inbox. The assumption was always fragile. It happened to hold up long enough for most of us to build careers on top of it.
What broke
The delivery mechanism broke on the rented channels, more or less at once.
LinkedIn cut organic reach by 60-80% from its 2023 peak. Average post reach is now 8-12% of followers, down from 15-20% a year ago. Company pages reach 2-4% of their own followers. That last one is worth reading twice. Your followers explicitly chose to see you. The platform decided not to show you to them. The relationship wasn't revoked by the people in the relationship. It was revoked by the mail carrier.
Google AI Overviews now appear on roughly 99.9% of informational queries. Zero-click searches climbed from around 56% in May 2024 to 69% a year later. Google used to deliver traffic to pages that answered the question. Now Google delivers the answer itself and keeps the traffic. Same structural move. Different lever.
And — this is the part I kept missing — even email is starting to become governed infrastructure. In February 2024 Gmail introduced stricter sender rules for anyone sending more than 5,000 messages a day: authentication via SPF, DKIM, DMARC, one-click unsubscribe, spam complaint rates below 0.3%. In November 2025 Google ramped enforcement from "light warnings" to actively rejecting non-compliant mail at the SMTP level. Microsoft rolled out parallel rules for Outlook in May 2025. Yahoo and Apple had already announced similar standards. The inbox is still the most direct channel we have, but it is no longer a neutral pipe. It is governed infrastructure with rising tolerances, and the governance is slowly turning into a second, quieter delivery layer of its own.
Meanwhile, the small, boring book mostly held. beehiiv's State of Newsletters 2026 report says paid newsletter subscriptions on the platform went from $8M in 2024 to $19M in 2025. Open rates holding around 38.7%. New launches hitting a median of 482 subscribers in their first month. These are not exciting numbers the way a viral post is exciting. They're exciting the way compound interest is exciting, which is to say only to people willing to stare at the same curve for five years without looking away.
Nobody is going to write a manifesto about compound interest. And yet.
Channels, ranked by delivery reliability
Still the most direct relationship you can have with someone who said yes, and still — by a margin — the channel where delivery reliability has held up best through the 2024-2026 period. The newsletter boom isn't a trend. It's a migration toward the mail carrier that didn't go on strike.
The honest update: email is not the free, neutral channel it used to be. Gmail, Outlook, Yahoo and Apple have collectively raised the floor on what "a bulk sender" has to do to reach an inbox. Authentication is no longer optional. Unsubscribe links have to be one click. Spam complaint rates have a ceiling. If you're small and sending under a few thousand a day you barely notice. If you're large and careless you quietly disappear. Email still held up. It held up partly because the standards tightened, not because nothing changed.
One more thing email does that broadcast-style channels structurally can't. Once you have the permission, email carries identity — you know who each specific person is, what they've opened, what they've clicked, what they've already bought. That's the prerequisite for the whole lifecycle and CRM tradition: accumulate knowledge about people, then talk to them one at a time based on it. Podcasts can't do this. Substack barely can. In-product messaging and push share the property because they sit on top of identity too. Permissioned broadcasting and permissioned one-to-one are two different modes that happen to share the same prerequisite and otherwise live in different universes — this post is mostly about the first, and the second is a whole other essay.
Chat and broadcast channels
One of the less-discussed shifts of 2024-2026: messaging products quietly turned into media products. WhatsApp crossed 3 billion monthly active users in 2025. Instagram crossed 3 billion in September 2025. Telegram crossed 1 billion monthly active users in March 2025, up from 950 million in July 2024. Meta's entire Family of Apps now averages 3.58 billion daily active people.
Look one layer deeper, inside the platforms, and the media-product transformation is clearer. WhatsApp's Updates tab crossed 1.5 billion daily users in June 2025. Instagram's broadcast channels now carry over 1.5 billion creator-to-follower messages every month. Telegram channels generate more than a trillion views a month. These are not messaging stats. They are distribution numbers for surfaces that look like media, act like media, and are structurally treated as media by the companies running them.
But look one layer deeper still and the picture gets murkier again. The same people who tap "subscribe" on a Telegram channel also drag it into a folder named "Later" they will never open. WhatsApp broadcast lists sit in a status tab most users visit by accident. Instagram broadcast channels live inside an inbox that competes with DMs from friends and unsolicited pitches from brands. Call it the vegetable graveyard — the place where things people meant to read go quietly rotten. Delivery, in the narrow technical sense, happens. Attention is a separate question and does not arrive automatically.
Broadcast channels on these platforms sit somewhere between a feed and an inbox. Higher frequency than email, closer to the lock screen than the homepage, and — at least for now — a lot less filtered. They still sit on platform-governed rails, so they are not owned in any strong sense. But they are fast, direct, and built for episodic updates. The useful 2026 frame isn't "replace email with chat." It is: use chat for intimacy, frequency, and drops; use email or first-party product identity for permanence. Different jobs, same audience.
In-product messaging
If you make a product, you already have a channel nobody talks about enough. The user opens the tool; you deliver the message inside the tool. No feed, no algorithm, no platform interpreting your intent. The release screen, the What's New page, the in-app notification center, the empty state, the update dialog — these are all direct, permissioned, and embarrassingly underused as marketing surfaces. (I am writing this from inside a company where the release screen is arguably doing more audience work than half of what the marketing org produces elsewhere. I keep relearning this.)
The constraint is that it only applies if you ship software. The opportunity is that if you do ship software, you have a delivery channel with a 100% open rate among active users and no platform in the middle.
Push notifications — mobile, desktop, browser
A heavier gesture than typing an email address. Someone installed your app or your extension and explicitly agreed to be interrupted by it. The delivery is direct and the reach is near-total. The scale is easier to see once you look at the infrastructure numbers: Airship's 2026 benchmark report analyzed more than 681 billion push notifications sent to over 3 billion users across 15 industries. That is not a minor channel dressed up as a major one. It is a major channel most brands still treat like a minor one.
The abuse potential is also near-total, which is why most brands are either too timid or too aggressive and very few get the tone right. For the ones that do, push is one of the most valuable direct lines to a permissioned audience that exists — and one of the fastest ways to burn that permission if you treat it like an ad slot.
Community spaces you host
Discord, Slack, private forums. Delivery works because the people came to the room. Algorithm-proof by construction. Hardest to grow, easiest to keep. For most companies the right move is not to start a community of strangers but to give the existing list somewhere closer to gather. The community is the second chamber of the heart, not a replacement for the first.
Podcasts
Delivery is partial. You upload an episode; the platform pushes it to people's feeds; the listener has to open the app to hear it. Global listeners hit 619M in 2026, up from 547M in 2024. In the US, 55% of the 12+ population now listens monthly, which is the first time the majority has ever crossed that line. YouTube became the biggest podcast platform in the world with over a billion monthly users, which is a sentence that sounds wrong even after you check the source twice. Discovery glacial, loyalty unreasonable. Small position, long horizon, don't check the price often.
YouTube subscribes and LinkedIn follows
These are the channels where the gap between permission and delivery is widest. People tapped subscribe, people tapped follow — that's a real yes. Then the algorithm sits in the middle and decides, for every single post, whether any given follower should actually see it. On LinkedIn the answer is increasingly no: organic reach down 60-80% from peak, company pages at 2-4% of followers, sporadic posters' visibility down from 57% to 28%.
And yet. (There is always an "and yet" with LinkedIn.) Average engagement rate actually climbed to 5.20% in 2025, up from 4.48% in January 2024. Multi-image carousels hit 6.6%, PDF documents 6.1%, native video 5.6%, text-only at the bottom. Employee posts outperform company-page posts by 6-8x in reach. Top creators now occupy 31% of the feed, up from 15% in 2022. Fewer people see any given post, but the ones who do interact more. The game didn't end — it consolidated into a very specific shape: named humans, niche topics, dense visual formats, personal profiles. YouTube is doing the same trick in a different costume.
Useful discovery surfaces. Never assets. If you want the asset, point them at something where delivery still works.
Blog and search
This one sits outside the framework because there is no yes in it at all yet. The blog is the pre-permission stage — the place strangers spend enough time to consider giving you permission. AI Overviews changed the traffic, not the purpose. Don't grieve the wrong number.
The hybrid pattern
Substack is a clean working example of how all this fits together in 2026. Its paid subscriptions grew from about 2 million in 2023 to 4 million in November 2024 to over 5 million by March 2025. Roughly a quarter of those paid conversions now come from inside Substack's own discovery engine — Notes, Recommendations, the app feed — rather than from external traffic.
That is the pattern worth naming. Substack is simultaneously a rented-reach surface (you're discovered through the network, the platform decides what surfaces) and an owned-relationship channel (once someone subscribes, you can email them directly, export the list, leave if you want). The two halves of the old owned media playbook are reunited inside a single product — rented discovery feeding directly into permissioned delivery, with the handoff happening in one click instead of across a funnel.
The strategic move is the same with or without Substack: use rented surfaces to create first contact, convert that first contact into a yes, and then deepen the yes on surfaces where delivery still works. Most brands are still running this backward — treating the rented reach as the destination and the email list as a slightly annoying legal requirement.
The deeper question
Permission decays. Somebody who subscribed last year and hasn't opened anything in nine months is, in every way that matters, no longer saying yes, and at some point you should do them the courtesy of removing them from the list. Same for followers — the person who followed you on LinkedIn in 2021 for some specific reason from a specific context is not obliged to keep caring. The yes was real. It has an expiration date. Renewing it is the actual work.
What makes 2026 different from 1999 is that we now have to manage both halves at once. The permission — earning it, renewing it, not abusing it. And the delivery — understanding which channels actually carry your message to the people who said yes, and which ones just store their names in a database and show them something else instead.
Different channels do different jobs. Search and social create discovery. Email holds identity. Podcasts build trust. Communities create belonging. Chat adds frequency. In-product messaging builds habit. You don't pick one. You stack them, and the stack is the actual asset.
There's one more question hiding underneath all of this, and it is probably the more interesting one. Platforms aren't tightening delivery because they are evil. They are tightening it because their users asked them to. Success on the platform's side of the equation is measured by whether the average person feels their feed and their inbox contain things they actually wanted. A message that gets filtered out was, almost always, a message the recipient didn't want enough to defend.
Which reframes the whole thing. The real question isn't "how do I get my message through the filter." It is: what did my audience think they were saying yes to? If the answer is a person, a point of view, a body of work, a topic they care about — the delivery mostly holds. That is why Substack and podcasts keep compounding. People subscribed to a human, or an opinion, or a topic they trust, and the platforms honor that permission because honoring it is their job.
If the answer is "a brand, in case it ever has something worth selling me" — the filter is doing exactly what it was built to do. The permission was vague, and the recipient knew it. What the recipient said yes to was never advertising.
Godin named the permission in 1999. 2024-2026 finally named the delivery. Both have to be true for a relationship to work. Neither one is free anymore.