Category: The Enshittification Machine: Platforms That Rot After Winning

A study of what happens after a platform wins.

This category follows the familiar cycle: attract users with a useful service, attract businesses with access to those users, then gradually squeeze both sides once leaving becomes inconvenient. Features disappear, ads multiply, search gets worse, subscriptions creep in, moderation becomes inconsistent, interfaces grow hostile, and the original product gets buried under monetization machinery.

The focus is on platform decay as a system rather than a single bad redesign. Network effects, lock-in, investor pressure, acquisitions, algorithmic control, subscription creep, API restrictions, creator dependence, and the slow conversion of useful public spaces into extraction machines all belong here.

The machine usually works beautifully right up until you notice what it is producing.

Posted on by

Bundled products that make users pay for unwanted services

Bundling is convenient when you want the bundle.

It gets stranger when the package grows, the price rises, and the new thing is something you never asked for.

Microsoft provided a useful example in January 2025 when it announced that Copilot and Microsoft Designer would be added to Microsoft 365 Personal and Family subscriptions in most markets. At the same time, Microsoft said U.S. prices for those plans would increase by $3 per month.

Microsoft explained the change in Copilot is now included in Microsoft 365 Personal and Family.

The bundle changed rather than remaining a menu

A Microsoft 365 customer might primarily want Word, Excel, PowerPoint, OneDrive storage, or Outlook.

Another customer might value Copilot enough to pay extra for it.

Bundling collapses those separate preferences into one package.

The standard Personal and Family plans gained the AI features and the standard subscription price rose. Microsoft did provide alternatives for existing subscribers: it said people with recurring billing could switch to Basic or, for a limited time, Personal Classic or Family Classic plans without Copilot or AI credits.

That qualification matters. This was not literally “use Copilot or lose Office tomorrow.”

But the default product direction was clear: AI became part of the main consumer subscription rather than remaining a separate optional purchase.

Bundles hide individual prices

Suppose a subscription contains six services and rises by $3 after a seventh is added.

A user who loves the seventh service may see a bargain.

A user who never opens it may see a price increase on the six things already being used.

Both are looking at the same invoice.

That makes bundles difficult to evaluate because the provider can point to the total value of everything included, while each customer assigns a different value to the pieces.

The bundle may objectively contain more features and still be worse for a particular user.

Standalone choice is what disappears

Software companies have good reasons to bundle products. One subscription is easier to market, support, bill, and integrate. New features also become available to a huge installed base immediately instead of waiting for people to discover and buy them separately.

The cost is granularity.

When a feature is sold independently, its price is exposed to a simple test: do enough people want this badly enough to pay for it?

When it is bundled into an established subscription, that test becomes fuzzy. Revenue from customers who mainly want the older products can help fund the new one too.

The useful question is not whether the bundle is “worth it”

There is no universal answer because customers use different parts of the package.

The better questions are measurable:

  • What was included before?
  • What was added?
  • How much did the price change?
  • Can the added service be declined without losing the older product?
  • Is an unbundled alternative permanent, temporary, or unavailable?

Microsoft’s 2025 change is interesting because it exposes all of those questions at once.

A subscription can become more capable and less optional at the same time.

Posted on by

Previously included features moved into higher-priced subscriptions

A feature does not have to disappear to be removed from your plan.

It can simply move one column to the right on the pricing table.

Evernote has repeatedly adjusted the boundary between its free and paid tiers. In late 2023, the company announced that Free accounts would be limited to 50 notes and one notebook for new creation. Existing users above those limits could still view, edit, export, share, and delete their existing material, but creating more required getting back under the limit or subscribing.

In 2024, Evernote described “virtually unlimited note and notebook creation” as a paid benefit while also limiting Free accounts to one connected device at a time.

Evernote documents the changes in its 2023 Free account limits announcement and its 2024 device-limit update.

The feature being sold may be ordinary use

There is nothing exotic about creating a 51st note.

That is what makes tier changes interesting.

Subscription software often begins by separating advanced features from basic ones: collaboration controls, automation, administration, large uploads, analytics. Those boundaries are easy for users to understand.

But a company can later redraw the line around behavior that had previously been normal everyday use.

Once that happens, the paid tier is not only selling something new. It may also be restoring freedom the user remembers having before.

Existing data makes the decision asymmetric

Evernote handled the 2023 change more gently than simply locking old notes. Users above the new Free limits retained access to their existing content and could export it.

That matters.

Still, a person with hundreds or thousands of notes has a different decision from a new user evaluating the Free plan today.

The new user can walk away with almost nothing invested.

The established user may have years of clipped pages, scans, task lists, tags, notebooks, links, and personal organization built around the service.

The subscription price is therefore compared not only with competing software, but with the inconvenience of migration.

A paid tier can also fund genuine improvement

Evernote argued that subscriptions supported infrastructure work, real-time editing, AI features, stability improvements, security work, and faster synchronization.

That is a legitimate part of the comparison. Software has continuing development and hosting costs. A free tier does not create a permanent entitlement to every capability it once included.

The useful question is more precise: what additional value is being sold, and what old capability has been moved behind payment?

Those are not the same thing.

If a higher tier adds substantial new functionality, the upgrade is easy to describe as a new product.

If the main practical effect is that an established user must pay to keep doing what the old tier already allowed, the economics feel different.

Nothing vanished from the software.

The line moved around the user.

Posted on by

Storage limits reduced after users accumulate large collections

Storage limits feel abstract until the account already contains ten years of your stuff.

Flickr demonstrates the problem unusually well.

In 2013, Flickr announced that every member would receive 1 terabyte of storage for free. Flickr said that amount could hold hundreds of thousands of photos.

Five years later, after Flickr had changed ownership, the free plan changed dramatically. Free accounts would be limited to 1,000 photos and videos. Users above that limit were told to upgrade to Flickr Pro or download material over the cap. Flickr later extended the deadline before older excess uploads became at risk of deletion.

The earlier 1 TB offer is documented in Flickr’s 2013 account-limits announcement. The later 1,000-item policy appears in Flickr’s official Free and Pro plan update.

The number changed after the collection existed

A 1,000-item limit is not inherently unreasonable for a free photo service.

The problem is chronology.

Someone joining a service under a 1,000-item limit can decide immediately whether that is enough. Someone who joined while a terabyte was available may already have tens of thousands of uploads organized into albums, linked from websites, commented on, favorited, tagged, and embedded elsewhere.

The storage policy changes in one announcement.

The collection does not.

Moving files is only the first layer

Flickr did provide users with ways to download their material, and Pro offered unlimited storage.

But migrating a photo archive is not equivalent to copying a directory from one disk to another.

The user may also care about:

  • album organization,
  • titles and descriptions,
  • comments and favorites,
  • public URLs,
  • embeds on old forum posts or blogs,
  • social connections,
  • and licensing metadata.

A local backup preserves the image bytes. It does not necessarily preserve the public life of the collection.

That is the switching cost.

A free service still has operating costs

Flickr’s decision was not mysterious. Storing and serving huge photo libraries costs money. The company said the revised free tier was intended for sharing a user’s 1,000 best photos and videos, while Pro would provide unlimited storage and other benefits.

A service is not obligated to offer unlimited historical terms forever simply because users liked them.

But users should understand the dependency they create when they treat a commercial platform as permanent storage.

The larger the collection becomes, the more expensive a later policy change is in time, attention, and risk.

That is why generous early storage limits can produce powerful lock-in even without any malicious intent.

The first thousand uploads are easy to move.

The next fifty thousand may come with fifteen years of context attached.

Posted on by

Premium placement sold to participants competing for the same audience

A marketplace can charge sellers for access to customers and then charge them again for better visibility to those same customers.

Amazon’s Sponsored Products system makes the mechanism easy to see.

Amazon explains that Sponsored Products can appear in shopping search results and on product pages. Sellers choose targets and budgets, and generally pay when shoppers click the advertisement. Amazon presents the format as a way to increase visibility and sales. Its own guide explains the system in Sponsored Products best practices.

The interesting part is not that Amazon sells advertising. Lots of websites do.

It is that the advertisers are often merchants already competing inside Amazon’s marketplace.

Visibility itself becomes a product

Imagine ten sellers offering similar products.

They are already competing on price, reviews, shipping, product quality, availability, and ordinary search ranking.

Sponsored placement adds another variable: who is willing to spend money to appear more prominently when the customer is ready to buy?

That creates a second market layered on top of the first one.

The first market sells products to shoppers.

The second sells shopper attention to merchants.

Those markets interact because prominent placement can generate clicks, sales, and data that may influence the seller’s overall performance.

Competition can raise the price of being seen

Paid placement is scarce. Search pages have limited high-visibility positions.

When multiple sellers want the same audience, they can compete for those positions through advertising budgets and bidding strategies.

The result is structurally different from a fixed listing fee. The cost of visibility can rise because competitors also want it.

A successful merchant therefore has to think about two margins: the profit on the product and the cost of acquiring the click.

That can be perfectly rational. If a $2 advertising cost reliably produces a $20 profit, buying visibility is useful.

But the platform benefits from the competition itself.

Paid placement does not prove product quality

This is where the user experience matters.

An advertisement can be highly relevant and still be an advertisement. Its position partly reflects a commercial transaction, not merely an independent judgment that it is the best product.

Amazon labels Sponsored Products, but shoppers still need to distinguish paid prominence from ordinary ranking signals such as price, reviews, and relevance.

For sellers, the same distinction matters in reverse. Better products do not automatically receive the best advertising positions. Advertising performance depends on targeting, bids, budgets, conversion, and other campaign factors.

The extraction question is measurable

Selling premium placement is not automatically abusive. Advertising can help new products get discovered and can improve search when targeting is relevant.

The useful question is how necessary paid placement becomes for ordinary participation.

If merchants can reach customers reasonably well without advertising, sponsored slots are an optional accelerator.

If unpaid visibility becomes too weak to sustain a business, the ad system starts functioning more like a toll lane built inside a road the merchants already depend on.

That boundary is where premium placement becomes part of the platform-lock-in story.

Posted on by

Paid verification as a change in the meaning of platform status

A symbol can stay visually identical while changing what it means.

The blue checkmark on Twitter, now X, is a strong example.

Under the legacy verification system, Twitter described the badge as a signal that an account of public interest was authentic. Applicants had to satisfy criteria built around authenticity, notability, and activity.

X began winding that system down on April 1, 2023. Its current help documentation says accounts verified under the old criteria did not retain the blue check unless they subscribed to X Premium. The blue badge now means the account has an active Premium subscription and meets eligibility requirements.

X documents both systems in its legacy verification policy and its current page explaining the blue checkmark.

The old badge answered a different question

Legacy verification was not a universal certificate of honesty.

It did not mean every statement from the account was true, and it did not mean Twitter endorsed the person. It primarily helped users distinguish notable authentic accounts from impersonators.

That made the badge useful shorthand when a celebrity, journalist, government figure, company, or other public account appeared in a fast-moving feed.

The newer badge answers something else.

X says the check indicates an eligible account subscribed to Premium. It explicitly notes that the blue check itself does not mean the account has been ID verified.

Same shape. Different data.

Buying access changes the trust calculation

Subscription-based verification can still provide useful signals. X reviews accounts for eligibility and can remove badges for rule violations or certain profile changes.

But users can no longer safely apply the old interpretation to the new symbol.

A blue check by itself does not establish that the account is notable under the former criteria, nor does it establish identity verification.

That means a user evaluating an unfamiliar account has to inspect more evidence:

  • the handle and account history,
  • linked official websites,
  • organization affiliations,
  • other verification labels,
  • and the content itself.

The icon does less interpretive work than it once did.

Status systems become confusing when their semantics move

Platforms are allowed to redesign verification. The old system had its own problems, including opaque decisions about who qualified and who did not.

Paid access can also broaden participation in features that were once restricted to a relatively small group.

The problem appears when users continue reading the old meaning into the new badge.

Verification symbols work because people learn a convention. Once the convention changes, trust does not automatically update at the same speed.

That makes paid verification more than a pricing story.

It is a semantic change to a piece of interface language millions of people had already learned how to read.

Posted on by

Mandatory payment processing and the removal of cheaper alternatives

Payment processing looks like plumbing until the marketplace owns the pipe.

eBay’s transition to managed payments is a useful example because the company did not merely add another checkout option. It made its own managed payment system a prerequisite for selling.

eBay says it began managing payments for sellers in 2018 and would require all sellers to register for the system in 2021. Its current user agreement still states that using managed payments is a prerequisite for using eBay as a seller. The company explains the model in eBay is managing payments and its User Agreement.

Centralized payments do solve real problems

There are obvious advantages to one payment layer.

The marketplace can present more payment methods to buyers, handle disputes in one system, collect fees automatically, standardize payouts, and reduce the number of separate accounts a seller has to manage.

eBay describes the result as simpler selling and a more flexible checkout.

That is real value.

The tradeoff is that sellers lose control over an important part of the transaction stack.

Choice disappears even when checkout improves

Before a platform controls payments end to end, a merchant may be able to choose among processors based on rates, payout timing, fraud tools, account history, or other business needs.

Once the marketplace requires its own processing layer, that comparison stops mattering for transactions on the platform.

The seller cannot say, “Processor B is cheaper for my business, so I will use that instead” if the marketplace itself determines the payment route.

That does not prove the mandatory option is more expensive in every case. Payment pricing is messy, and the answer depends on transaction size, category fees, country, processor terms, and which older arrangement is being compared.

The important change is bargaining power.

The platform now controls two toll booths

A marketplace already controls access to its buyers.

When it also controls payment processing, it controls another layer between the sale and the seller’s bank account.

That creates efficiencies, but it also makes future pricing changes harder to avoid. A merchant cannot keep the marketplace while swapping only the processor if the two are contractually tied together.

For an established seller, leaving may mean abandoning reviews, listing history, saved searches, customer habits, and marketplace traffic simply to regain choice over payments.

That is why mandatory processing belongs in discussions of platform lock-in.

The problem is not that integrated payments are inherently bad. They can be substantially easier for both buyer and seller.

The warning sign is structural: when a platform converts an optional service into required infrastructure, sellers stop comparing that service on an open market.

The platform comparison becomes much larger and uglier: accept the entire stack, or move the business.

Posted on by

Seller fees that increase after merchants become established

A marketplace fee is easy to understand when a seller joins.

The harder question comes later: what happens after the shop has spent years becoming valuable inside that marketplace?

Etsy supplied a clean example in 2022. The company announced that its transaction fee would rise from 5% to 6.5% beginning April 11, 2022. Etsy said the additional revenue would support marketing, customer support, seller tools, and marketplace safety.

The announcement remains in Etsy’s Seller Handbook, and Etsy separately confirmed that the new 6.5% rate applied to the total order amount.

A percentage increase compounds with the business

For a shop doing $1,000 in fee-applicable sales, the difference between 5% and 6.5% is $15.

At $10,000, it is $150.

At $100,000, it is $1,500.

That does not mean the fee increase was unjustified. Etsy argued that spending more on buyer acquisition, support, and enforcement would improve the marketplace for sellers. A larger marketplace can provide more value than a cheaper one with no customers.

But percentage fees have an interesting property: the more successful the merchant becomes, the more expensive the platform becomes in absolute dollars.

The seller is not starting from zero somewhere else

An established Etsy shop may have years of reviews, favorites, search placement, returning buyers, listing history, and links from elsewhere on the web.

Those assets are not completely portable.

A seller can open an independent storefront or move to another marketplace, but the new location does not inherit the old reputation graph. Customers must be taught where to go. Search visibility has to be rebuilt. Reviews may remain behind.

That makes a fee increase different for an established merchant than for someone deciding whether to open a shop today.

The new seller can simply reject the deal.

The established seller has to price the cost of leaving.

The correct comparison is fee versus supplied value

It is tempting to look at any marketplace fee increase and call it extraction.

That is too simple.

If a platform raises fees and delivers proportionally better fraud protection, support, advertising, conversion, payment handling, or buyer traffic, a seller may still come out ahead.

The useful question is whether the incremental value keeps pace with the incremental cost.

Etsy said its 2022 increase would fund more marketing, a support-team expansion, enforcement work, and seller tools. Sellers could then judge whether those improvements were worth another 1.5 percentage points on transactions.

That is the broader lock-in pattern to watch.

A platform first becomes useful enough for businesses to build around it. The businesses accumulate platform-specific history. Only then does changing the fee become especially powerful, because the alternative is no longer merely “use a competitor.”

It is “rebuild part of the business somewhere else.”

Posted on by

Organic business reach replaced by paid promotion

A business can spend years collecting followers and still discover that “follows this Page” does not mean “will see this Page.”

Facebook documented that distinction clearly in 2014.

The company announced that, beginning in January 2015, people would see fewer Page posts that it classified as overly promotional. Facebook said users had complained about posts that mainly pushed purchases, app installs, contests, or recycled advertising creative. Its announcement warned that Pages posting this kind of material should expect their organic distribution to fall significantly over time.

Facebook’s own explanation is preserved in Reducing Overly Promotional Page Posts in News Feed.

A follower list is not a mailing list

This is the important mechanical difference.

If 20,000 people join an email list, the sender controls the act of sending. Delivery can still fail, filters can intervene, and people can unsubscribe, but the platform does not rank every message against unrelated content before deciding whether the subscriber gets a chance to see it.

A social-media follower relationship works differently.

The platform owns the feed. It decides which followed accounts appear, in what order, and how frequently. A business may accumulate a large audience while having no guaranteed path to that audience.

That is not necessarily deception. Feed ranking is unavoidable once the volume of available content exceeds what a person can reasonably read.

But it changes what an audience asset actually is.

Paid promotion fills the gap

Facebook’s 2014 announcement explicitly said the change would not increase the number of ads people saw. The stated goal was better News Feed quality, not forcing every business to buy advertising.

That distinction matters.

A decline in unpaid reach is not proof that a platform secretly suppressed a Page in order to sell ads. Competition for feed space, ranking changes, user behavior, content quality, and posting frequency all affect distribution.

Still, once reliable organic distribution weakens, paid promotion becomes the obvious way to regain predictable reach.

The business now has two separate relationships with the platform:

  1. it creates content and gathers followers there;
  2. it may then pay the same platform to put that content in front of more people.

The lock-in is the audience itself

Leaving is not as simple as opening another account elsewhere.

Followers do not automatically move. Years of comments, reviews, posts, social proof, and customer habits remain attached to the old platform.

That makes reduced organic reach economically important even when no conspiracy is involved.

The useful measurement is not “Facebook killed organic reach” as a slogan. It is how much of an established audience a Page can reach without payment, how that changes over time, and what a business must spend to recover comparable visibility.

A million followers look like ownership on a dashboard.

Operationally, they are rented access.

Posted on by

Paid tiers that gradually inherit advertisements

Paying for a service does not permanently define what the service contains.

Amazon Prime Video made that unusually clear.

In September 2023, Amazon announced that Prime Video movies and shows would begin carrying “limited advertisements” in several countries starting in early 2024. The company said the change would help fund continued investment in content. It also said the price of Prime membership itself would not change in 2024.

There was, however, a new way to keep the old viewing experience: U.S. Prime members could pay an additional $2.99 per month for an ad-free option.

Amazon’s announcement is still available in its Prime Video advertising update.

The subscription stayed paid while the experience changed

This is different from a free service introducing advertising.

Prime members were already paying for a bundle that included Prime Video. After the change, the ordinary paid tier gained advertisements and the ad-free version became an extra charge.

That matters because the customer’s mental model is often based on the service they originally joined. A person may reasonably think, “I already pay for this,” even when the contract allows the provider to change features later.

The platform’s view is different. A subscription buys the current package under the current terms, not a frozen copy of the product forever.

Both statements can be true at once.

Advertising becomes a second price

The monetary price of Prime did not have to rise for the cost of watching Prime Video to increase.

A viewer could now pay in one of two ways:

  • tolerate advertising interruptions, or
  • add another monthly charge to remove them.

That is why advertisements inside paid services are useful to examine separately from ordinary price increases. The platform can increase revenue without changing the headline membership price.

The user sees a feature that used to be part of the normal experience separated out and sold back as an upgrade.

Not every added ad is automatically platform decay

Streaming video is expensive. Content production, licensing, delivery infrastructure, and sports rights cost real money. Amazon explicitly tied the advertising change to continued investment.

So the existence of ads is not proof that a service has become worse overall.

The useful question is narrower: what changed for an existing paying customer?

In this case the answer is easy to measure. Prime Video went from a paid service where ordinary on-demand viewing was generally ad-free to a paid service where limited advertising became the default, while avoiding those ads required another monthly payment.

That is the mechanism worth watching across subscription services.

The price on the front door can stay exactly where it was while part of the old experience quietly moves behind a second door.

Posted on by

Increasing advertising density inside a previously useful service

An advertisement does not have to cost money to cost the user something.

It can cost twelve seconds.

Then another twenty.

Then the interruption required to remember where the video left off.

That is why advertising density is better measured in attention and task interruption than in rectangles on a page.

YouTube provides a documented example of the available ad surface expanding over time.

Mid-roll inventory became available on shorter videos

Before July 2020, YouTube’s mid-roll ads were limited to videos longer than ten minutes.

TeamYouTube announced that the threshold would drop to eight minutes. Eligible existing videos could have mid-rolls enabled, and new eligible uploads would have the feature enabled by default unless creators changed their preferences. See TeamYouTube’s 2020 explanation of the mid-roll change.

The practical result was simple:

A larger share of the video catalog became eligible to contain advertising inside the program rather than only around it.

That is an increase in possible ad density even though not every available slot necessarily serves an ad.

The machinery kept gaining more places to work

YouTube’s current ad documentation lists pre-roll, post-roll, skippable, non-skippable, bumper, mid-roll, feed, Shorts, and other formats. Long-form videos can also receive ad pods, meaning two back-to-back video advertisements in one break. See YouTube’s advertising formats documentation.

In 2025, YouTube announced another mid-roll change: for existing monetized videos that already had manually placed mid-roll slots, the company would add automatic slots by default unless creators opted out. YouTube said combining manual and automatic opportunities could increase creator earnings. See TeamYouTube’s 2025 mid-roll update.

Again, an ad slot is not a guarantee that an ad appears.

But more eligible videos and more eligible slots create more opportunities to sell attention.

The user pays in interruption time

Suppose a 20-minute tutorial once contained one interruption and later contains three.

The monetary price to the viewer can remain exactly zero.

The experience still changed.

Useful measures include:

  • ad minutes per hour watched,
  • interruptions per video,
  • number of consecutive ads,
  • time before content begins,
  • screen area occupied by ads,
  • clicks required to dismiss or skip them,
  • and whether paid removal is available.

Those measures are more meaningful than simply declaring a site “full of ads.”

More ads can fund something real

Advertising pays creators, infrastructure bills, licensing costs, moderation teams, and the platform itself.

More inventory is not automatically evidence of platform decay.

The Enshittification question is about the exchange.

If advertising increases, what improved for the user in return?

If nothing improved, the platform may be extracting more attention from an audience whose reasons for staying were built earlier.

That is the measurable form of the complaint.

Not there are ads now.

But the same task now requires surrendering more of your time to the platform’s revenue machinery than it used to.