Timber survey pegs and a taut string line cutting straight through an established garden bed of mature planting.

The year platform risk stopped being theoretical

A retrained US TikTok recommender, rationed Google clicks and a monthly Instagram originality check all landed within twelve months of each other.

Every audience you have on a platform is rented. Everyone knows that, nobody prices it, and then a year like this one arrives and the invoice shows up all at once.

Four things changed inside twelve months, each one out of your control, each one capable of moving a distribution channel a business had come to rely on.

Google started rationing the click

Zero-click searches hit 68.01 per cent of US Google searches between January and April 2026, up from 60.45 per cent in 2024. Chartbeat measured global Google search traffic down 33 per cent year on year across more than 2,500 publisher sites. Smaller sites lost roughly 60 per cent of their search referrals over two years, against 22 per cent for large publishers.

The click did not disappear, it got rationed, and the rationing favours size. Seer's data puts organic click-through with an AI Overview present at about 2.1 per cent when your page is cited inside it and about 0.9 per cent when it is not, against about 3.3 per cent with no overview at all.

Nothing you did caused that and nothing you can do reverses it.

Instagram started assessing accounts monthly

On 30 April 2026 Instagram extended its originality system from Reels to photos and carousels. The assessment is at account level and monthly, not per post. If most of a month's output is reposted material, the account is classed as an aggregator and loses recommendation surfaces including Explore and suggested posts. Follower feeds are unaffected. Instagram stated plainly that crediting the original creator or adding superficial edits does not meet the threshold.

Read the mechanism. One month of lazy output, and the penalty applies to the account, not the posts. Brands running repost-heavy calendars, or resharing user content as the bulk of their feed, are exposed to a rule that judges them in aggregate.

The three signals Mosseri named in January 2025 still hold: watch time, likes and sends. The useful detail is the split. Likes matter slightly more for reach among your followers, sends matter slightly more for reach among non-followers. If the brief is growth, design for the send. If the brief is nurture, design for the like.

One housekeeping note. Instagram's own ranking explainer page still carries a 31 May 2023 date and the old four-bucket framing with follower count listed as a Reels signal. Do not cite it as current.

LinkedIn replaced its ranking system with a language model

LinkedIn published 360Brew, a 150 billion parameter decoder-only foundation model that replaces the separate models previously handling feed, people, jobs and ads, in January 2025. It confirmed the production rollout on 12 March 2026, describing generative recommenders paired with large language models.

LinkedIn's own example: a member interested in electrical engineering who engages heavily with small modular reactors now gets matched correctly, because the system reads topical relationships instead of keywords. It also surfaces expert commentary on breaking news within minutes rather than hours.

The consequence is that hashtags, keyword stuffing and topical tagging are close to worthless as distribution levers now. The model reads the content. Semantic specificity replaced metadata gaming, and most advice about the LinkedIn algorithm has not connected the engineering paper to the March announcement.

TikTok's US recommender was rebuilt

As part of the ownership restructure, the US app moved to a recommendation system retrained on US data. Set aside what happened to any individual account's reach. The point is structural: the mechanism that decided who saw your work was replaced by a corporate and political process that had nothing to do with your marketing, your budget or your customers.

That is the definition of platform risk, and it is now demonstrated rather than argued.

What owning an audience actually means

Not a follower count. Followers are a permission the platform can revoke by changing what it shows them.

Owned means you hold the contact detail and the relationship. In practice, five things.

  1. 1.Email, with a reason to open it. Not a newsletter nobody asked for. Something with information in it that the reader would miss.
  2. 2.Phone numbers, used sparingly. Highest reach of anything you can own, and the fastest to burn.
  3. 3.A customer database that marketing can query. If your CRM cannot answer who bought what and when, you do not have first-party data, you have a filing cabinet.
  4. 4.A site people return to on purpose. Tools, calculators, stock availability, project galleries. Something with a reason to come back that does not depend on a feed.
  5. 5.Direct relationships with people who have their own audiences. Journalists, association editors, industry podcasters. Slower to build and impossible for a platform to switch off.

The allocation question

None of this means retreat from social. Distribution is where audiences are, and a million views a month is the kind of reach that makes everything else cheaper.

It means treating platform reach as leased and building a conversion into ownership at every touchpoint. Every campaign should answer one question: what did we keep? A view you cannot contact again is rented. A name and an email is yours.

Rent the reach, own the list.

Written by David Eid. Published .