Last updated: September 2026
Key Takeaways
- This page used to be nine predictions for 2023. We’ve kept the predictions, graded them honestly, and rebuilt the page around what the scorecard actually teaches.
- Four held up, three were wrong, two were half right. The pattern is clear: predictions about behaviour aged well, predictions about features aged terribly.
- “Stories will spread to every platform” was the worst call on the list. YouTube shut its Stories down in June 2023, and the copy-Stories wave receded everywhere else.
- “Go live or go home” inverted completely. Instagram now requires a public account with 1,000 followers just to start a broadcast.
- Social commerce grew as a category while the specific mechanism everyone predicted died. US social commerce passes 100 billion dollars in 2026, but you check out on the brand’s own site.
- The planning lesson: build processes on durable human behaviour, and treat every individual platform feature as something you’ll lose.
Nearly every social media prediction post has the same problem. Nobody ever goes back and marks it.
So we did. This page carried nine predictions, and rather than quietly replacing them with nine fresh ones, we’ve scored the originals against what actually happened.
Four held up. Three were flat wrong. Two were half right in a way that turned out to be more instructive than either.
And the pattern in the misses is the genuinely useful part, because it tells you which kinds of prediction are worth planning around and which are just noise you’ll pay for later.
What Should a Social Media Prediction Actually Tell You?
A useful prediction tells you where human behaviour is heading, not which feature will be popular. Behaviour changes slowly and predictably. Features get killed by product managers with no notice, which makes them unplannable.
That distinction is the entire finding of the scorecard below, and once you see it you can’t unsee it.
“People will consume more short video” was safe. “Stories will be everywhere” was a bet on a feature, and it lost.
Which Predictions Actually Held Up?
The worst call: Stories everywhere
The original prediction pointed out that LinkedIn, Pinterest, Twitter, YouTube and even Spotify had all shipped a Stories format, and concluded that this proved the format was the future.
It proved the opposite. That was peak copying, not early adoption.
YouTube shut Stories down on 26 June 2023, and its stated reasoning is the lesson. Among creators using both, Community posts drove many times more comments and likes than Stories, and Shorts drove many times more subscribers.
LinkedIn and Twitter had already dropped theirs. Anyone who built a Stories workflow across five platforms in 2021 rebuilt it twice by 2024. Our write-up on YouTube Stories now exists mainly as a record of that.
Stories on Instagram, for the record, are alive and useful. The prediction wasn’t that Stories would work. It was that the format would spread, and it didn’t.
The most instructive miss: go live or go home
This one inverted so completely it’s almost funny.
The prediction said you didn’t need a big following to run a successful live session, and that live would become a normal part of everyone’s mix. In August 2025 Instagram made a public account with at least 1,000 followers a hard requirement to broadcast at all.
Live shopping went too. Instagram removed live product tagging in 2023, and the wider Instagram live shopping push was wound down.
What actually happened is that live video consolidated into the places built for it, and became a format for people who already had an audience rather than a way to get one.
The half-right one worth studying: social shopping
This is the most useful entry on the scorecard, because the direction was right and the mechanism was wrong, which is the most expensive way to be partly correct.
Right: social commerce grew. US social commerce sales pass 100 billion dollars in 2026 on eMarketer’s forecast, an 18 percent year-on-year rise.
Wrong: the prediction’s specific claim was that customers would buy without leaving the app, everything under one platform, no second website. Instagram’s native checkout was phased out through 2025 and product tags now send people to your site. The Shop tab left the navigation in 2023.
So anyone who followed that advice literally built their storefront inside a product that no longer exists. Anyone who read it as “invest in social discovery pointing at your own store” got it exactly right. Same prediction, opposite outcomes, depending on how literally you took it. Our current guides to shopping on Instagram and how Instagram shopping evolved describe the version that survived.
What Never Changes?
Strip out five years of feature churn and a short list survives untouched.
- People share things that make them look good to someone specific. Every distribution system ever built rewards this, which is why sends now outrank likes on Instagram.
- Being useful beats being clever. The teaching format has topped engagement through every algorithm change in the period. Carousels still lead on Socialinsider’s 2026 figures at 0.55 percent, ahead of both reels and single images.
- Consistency beats intensity. Not because of a consistency bonus in the ranking, but because irregular posting means you never learn what works.
- You don’t own anything you didn’t build. Followers, features and reach are all rented. Email lists and websites aren’t.
- Attention is the only genuinely scarce input. Everything else, including content production, has got cheaper every year. That asymmetry is the whole game.
Every prediction that held up on the scorecard is a restatement of one of those. That isn’t a coincidence, it’s the method.
What’s Actually Changing Right Now?
Four shifts, stated as behaviour rather than features, so they might survive contact with next year.
1. Sharing is moving private
The public feed is no longer where most sharing happens. Adam Mosseri has said more photos and videos are shared in Instagram DMs than to Stories or Feed on any given day.
Which means a chunk of your best performance is now invisible to you. Something forwarded 200 times in DMs shows almost nothing on a public post.
2. Distribution is detaching from audience
Recommendations now decide most reach, which means your follower count matters less each year and the individual post matters more. Good for new entrants, uncomfortable for anyone who spent a decade accumulating.
We’ve written up the mechanics in how the Instagram algorithm works and what it means longer-term in the future of Instagram marketing.
3. Content production got nearly free, so curation got valuable
AI collapsed the cost of making things. Platforms responded by penalising volume without originality.
Instagram’s April 2026 rule is the clearest example: post content you didn’t create or materially enhance 10 or more times in a rolling 30 days and you lose recommendation eligibility until 30 days after the last one.
The strategic read: producing more is no longer an advantage, because everyone can. Having a point of view is.
4. Discovery is splitting between search and answers
People increasingly get answers from AI summaries rather than clicking through a list of links. That changes what content is for: being quotable and extractable now matters as much as ranking.
Our guide to zero-click SEO covers the practical side, and it applies to social captions as much as web pages.
How Do You Build a Strategy That Survives a Platform Change?
Three rules, all derived from the scorecard rather than invented for this section.
Write your strategy in behaviours, not features. “We help people discover us through short video” survives a format change. “We post Reels” doesn’t. Same activity, one of them outlives the product decision.
Never let a platform hold the only copy of anything. Your customer list, your content archive and your analytics history should all exist somewhere you control. Every organisation that lost work to a discontinued feature skipped this.
Assume the feature you like most will be removed. Ask what breaks if it disappears tomorrow. If the answer is “our whole process”, that’s not a strategy, it’s exposure. Our broader guide to social media business strategies goes through building the rest of it, and 25 content ideas covers filling it.
FAQs
Is social media marketing still worth it?
Yes for businesses whose customers make visual or impulse-led decisions, and increasingly questionable as a standalone channel for long-cycle, research-driven purchases.
The question almost always means something narrower than it sounds: is it still worth the hours, given that the results got worse. That’s a fair reading of what happened, and worth answering with numbers.
Per-post engagement genuinely fell. Socialinsider’s 2026 benchmarks put Instagram’s average at 0.48 percent, down 24 percent year on year across 35 million posts. At the same time Pew found half of US adults on Instagram and 84 percent on YouTube in late 2025. The audience didn’t leave, the easy reach did.
The failure mode is running social media as a channel with no defined outcome, posting consistently for a year, getting reasonable engagement and having nothing to show a finance director. Engagement is not revenue and was never a proxy for it.
The honest cost: social media is now a skilled ongoing expense rather than a cheap growth hack. If nobody owns it properly, it will underperform the same money spent on search.
Decision rule: pick one commercial number, enquiries or sales, and give it 90 consistent days. Flat after that means your customers don’t decide here, and the budget belongs elsewhere.
What should a social media strategy include?
At minimum: one commercial objective, two platforms you can actually sustain, a defined audience, a repeatable content format, a posting cadence you can hold, and one number that tells you it’s working.
What people usually mean is “what am I missing”, and the answer is nearly always the last item. Most strategy documents are long on content ideas and silent on how anyone would know if they worked.
Two specifics worth getting right. Cadence should be the highest frequency you can hold at full quality for three months, not an aspiration, because volume at lower quality lowers the engagement ratios that drive distribution. And formats should be written as behaviours rather than products, since a strategy naming specific features dies with them.
The failure mode is the 40-page deck with a platform-by-platform matrix that nobody opens again. It looks like rigour and functions as a filing exercise.
The tradeoff to accept explicitly: a strategy narrow enough to execute will feel like it’s leaving opportunities on the table. It is, deliberately, and that’s why it works.
Decision rule: if your strategy doesn’t fit on one page and name the number you’re moving, it’s a wish list.
How do you decide which social media platforms to be on?
Pick based on where your buying decision gets made, then cut the list to the number of platforms you can post to weekly without dropping quality. For most small teams that’s two.
The instinct is to be everywhere in case you miss something. That instinct is what produces four neglected accounts.
Use the actual reach data rather than what feels current. Pew’s late-2025 survey of 5,022 US adults: YouTube 84 percent, Facebook 71, Instagram 50, TikTok 37. TikTok is culturally loud and substantially smaller than Instagram, and Facebook is unfashionable and far larger than both.
Then match to decision type. Visual and impulse-led goes to Instagram and TikTok. Research-led B2B goes to LinkedIn and search. Anything with a how-to component goes to YouTube, which is where that behaviour lives.
The failure mode is cross-posting the same asset everywhere and reading the weakest result as proof the platform doesn’t work. It’s proof the asset was made for somewhere else.
Decision rule: two platforms done properly, reviewed at 90 days. Add a third only after both have run for two consecutive quarters without slipping. Our comparison of the main social platforms covers the rest.
How do you plan ahead when platforms change every few months?
Plan the behaviour on an annual horizon and the tactics on a quarterly one. The mistake isn’t planning too far ahead, it’s planning the wrong layer that far ahead.
The scorecard above is the evidence. Predictions about behaviour held up over five years. Predictions about features were wrong within two, and the people hurt weren’t the ones who planned too far out, they were the ones who planned the wrong thing.
Practically, separate your plan into two documents. The annual one names audiences, objectives and behaviours you’re betting on. The quarterly one names platforms, formats and features, and you expect to rewrite it. Nothing in the annual document should mention a product name.
The failure mode is the opposite and very common: an annual plan full of feature commitments, which forces a full rewrite every time a platform ships an update, so the team stops planning at all and just reacts.
The cost of doing this properly is that your annual plan will look vague to anyone expecting a content calendar. That vagueness is what makes it durable.
Decision rule: if a platform announcement forces you to rewrite your annual plan, the plan was written at the wrong altitude.
Is social commerce worth setting up?
Worth using social platforms to drive purchases, yes. Worth building your storefront inside one, no, and the last five years are a fairly brutal argument for that.
The category is real. US social commerce passes 100 billion dollars in 2026 on eMarketer’s forecast, up 18 percent year on year and around 6.9 percent of US retail ecommerce.
But look at what happened to the in-app version. Instagram’s native checkout was phased out through 2025, live shopping product tagging was removed in 2023, and the Shop tab left the navigation the same year. Product tags now route to your website. Every business that made Instagram its checkout rebuilt that flow.
The failure mode is a catalogue integration maintained for a surface nobody visits, which is real ongoing work for product feeds, stock sync and pricing, producing almost nothing.
The tradeoff on the other side is friction. Sending people to your own site costs you conversions at the hand-off, and you accept that in exchange for owning the transaction, the customer data and the ability to keep selling when a platform changes its mind.
Decision rule: keep product tagging on, keep checkout on infrastructure you own. Never let a platform hold the only route to buying from you.
Do you need to be on every new platform that launches?
No. Claim the handle, post nothing, and wait for evidence the audience is actually there. Early-mover advantage on social platforms is real but far rarer than the people selling it suggest.
The fear driving this is missing the next Instagram. Worth noticing how often that fear has been wrong: the list of platforms that were going to be essential and weren’t is considerably longer than the list that mattered.
Set a threshold in advance so you’re not deciding emotionally. A reasonable one: your actual customers mention it unprompted, or a competitor is demonstrably getting business from it, not just posting there. Until one of those is true, the new platform is a hobby.
The failure mode is adding platform five while platforms one and two are already slipping. Attention is the scarce input for you exactly as it is for your audience, and a new account always costs more than it looks.
What early entry genuinely buys, when it works, is cheap reach before the platform starts charging for it. That’s a real prize, and it’s a bet, not a plan.
Decision rule: register the name on day one, commit resources only when a customer names it or revenue demands it.
Conclusion
The scorecard’s lesson is narrower and more useful than any fresh set of predictions would be.
Predictions about people held. Predictions about products failed. So write your strategy in the language of what people do, keep your own copy of everything that matters, and treat every feature you rely on as temporary.
Do that and the next platform change is an inconvenience rather than a rebuild.
Read next: building a social media strategy, how often you should post, and when to post.
If you’d rather have someone building the durable version of this for you, talk to AB Media. We’ll happily be graded on it in five years.
