How Creator Incentives Change What Gets Published

How Creator Incentives Change What Gets Published

user avatar

by me1afon

20 minutes ago


AI tools helped gather research and write the first draft of this article. I edited it and checked every figure and quotation against the platform's own page, the study's official record or the news report named at the end. The cover illustration is AI-generated. Details at the end.

In the fall of 2018, Jonah Peretti, the chief executive of BuzzFeed, wrote to a senior Facebook official with a complaint. The most divisive things publishers made were going viral, he said, and that created an incentive to make more of them [1].

His example was a BuzzFeed post titled "21 Things That Almost All White People are Guilty of Saying". It drew 13,000 shares and 16,000 comments on Facebook, many of them from people criticising BuzzFeed or arguing with each other. Other work from the same newsroom, from news videos to pieces on self-care and animals, struggled to break through [1]. Peretti wrote that his staff felt "pressure to make bad content or underperform" [1].

He blamed a change Facebook had made to its News Feed earlier that year, meant to promote "meaningful social interactions" between friends and family [1]. Facebook said it wanted to strengthen bonds between users. It ended up teaching publishers that a post which made people argue would travel further than one which informed them.

Nobody at Facebook asked publishers to write angrier posts. The ranking rule asked for comments, and publishers delivered what produced comments. That is the subject of this article: an incentive does not only change how much people publish. It changes what they publish.

This is the seventh part of the series Influence & Incentives, on how other people, rules and rewards shape what we decide, write and publish. The fourth part covered how rewards can push out the reason people acted in the first place. The sixth asked how to measure an author's contribution. This one looks at the output itself: six documented cases where a reward changed the shape of what appeared.

A ranking rule is a pay rule

Creators are paid in more than money. A platform pays them in reach when its feed ranks a post higher, in status when a leaderboard lists their name, and in cash when a revenue share or a token is attached to a count. To the person deciding what to make next, all three work the same way: they signal what the platform will reward.

The social scientist Donald T. Campbell described the general problem in 1979. In his account, the more a numerical indicator is used to make decisions, the more pressure there is to corrupt it, and the more it distorts the very activity it was meant to monitor [8]. It is often called Campbell's law, and it has a better-known cousin in Goodhart's law. Creator platforms are a clean test of it, because the indicators are public, the rewards are quick and millions of people can adjust their work within days.

The cases below are not about cheating. Most of the people involved did nothing against the rules. They noticed what was counted and moved toward it. That movement is the point.

Case 1: rewarding comments rewarded anger

The Facebook change of 2018 gave more weight to posts that drew comments, reactions and reshares. According to The Wall Street Journal, which reviewed internal documents, the company's own researchers found that publishers and political parties were reorienting their posts toward outrage and sensationalism, because that produced the comments and reactions the feed now rewarded [1].

A team of Facebook data scientists wrote in a memo: "Our approach has had unhealthy side effects on important slices of public content, such as politics and news" [1]. The clearest number came from Poland. In an April 2019 internal report, two Facebook researchers wrote that one party's social media team estimated it had shifted its posts "from 50/50 positive/negative to 80% negative", explicitly because of the algorithm change [1]. In Spain, according to the same report, parties said they had learned that harsh attacks on opponents drew the most engagement, and that while they "try not to", in the end "you use what works" [1].

Case 2: from clicks to time watched

YouTube changed its own measure earlier. In March 2012 it changed how it chose suggested videos, and in October 2012 it announced that it had "started adjusting the ranking of videos in YouTube search to reward engaging videos that keep viewers watching" [2]. The goal, the company wrote, was to focus its video discovery features on watch time. Its experiments, it said, had produced "less clicking, more watching" [2].

A ranking based on clicks rewards the title and the thumbnail: anything that makes someone press play. A ranking based on time watched rewards whatever keeps them there. For a creator, that is a different brief. The announcement does not say how video length or style changed afterwards, and I found no YouTube figure on it, so any claim about longer videos would be a guess. What the post does show is a platform deciding to reward a different kind of success, and replacing the count.

Case 3: paid per 30 seconds, cut to 30 seconds

Spotify's royalty rules count a play as a stream once it passes a minimum length. In November 2023 the company described what that rule had produced in one corner of its catalogue: "Some bad actors are shortening whale sound tracks to 30 seconds" [3]. A listener who played two minutes of such recordings generated four royalty-bearing streams instead of one, because each short track counted separately [3].

As Spotify describes it, the trick needed no fake listeners: the incentive simply rewarded cutting the same sound into more pieces. Spotify's answer was to change the rule: from 2024, functional noise recordings, such as white noise, nature sounds and ASMR, must be at least two minutes long to earn royalties, and Spotify said it would work with licensors to value them at a fraction of a music stream [3]. In the same announcement it set a minimum of 1,000 streams in the previous twelve months before a track earns recorded royalties, and said that 99.5% of all streams were of tracks above that line [3].

Case 4: paid per post, flooded with replies

Crypto added a newer kind of incentive: apps that scored people's posts on X about a project and paid out tokens or rewards to the highest scorers. The format became known as "InfoFi". The result was more posts, and especially more replies, written for the score.

On 15 January 2026, Nikita Bier, X's head of product, announced that the platform would "no longer allow apps that reward users for posting on X", saying they had produced "a tremendous amount of AI slop & reply spam" [4][5]. X revoked the apps' access to its API [4]. Kaito, the best-known of them, said it would end its Yaps programme and its incentivised leaderboards, and its founder said that tighter eligibility rules and filters had not fully stopped low-quality posts and spam [4][5].

The research notes behind this article list several ranking services of this kind, and their scoring formulas were mostly not public, partly to make them harder to game. Hiding the formula did not change the basic signal. If a post about a project earns points, people will write posts about that project, and many of them will be written for the points.

Case 5: paid per vote, votes went on sale

Steemit, a blockchain social network, paid cryptocurrency to the authors and curators of popular posts. A 2019 study of 539 million operations by 1.12 million users found that more than 16% of cryptocurrency transfers on the platform went to curators the researchers suspected were bots, and it found an underlying supply network for those bots [6]. When votes carry money, a market in votes is a predictable result. Cheating of this kind is the subject of a separate article in this series; here it matters as the far end of the same line, where the counted signal stops describing the work at all.

Case 6: a new format, a new crowd of posters

Incentives do not only push in bad directions. The clearest example of a positive shift in the research notes is Farcaster, a crypto social network. When the network launched Frames, small interactive apps inside posts, in January 2024, the number of accounts posting each day rose from about 2,400 the day before the launch to 19,100 about a week later, and the number of accounts reacting rose from about 2,700 to 25,000 [7].

A new format with fresh attention gave many more people a reason to publish. The same numbers also show the limit of a count: they say how many accounts posted, not whether the posts were any good, or whether the people stayed once the novelty passed.

What incentives change, and how to see it

Across the six cases, an incentive changed at least one of four things.

What changes What it looks like Case
Tone Posts tuned for reaction: anger, alarm, argument Facebook, 2018
Form Work cut, padded or restructured to fit the counted unit Spotify noise tracks; YouTube's move away from the click
Volume More posts, replies or votes made for the score, not the reader InfoFi apps on X; Steemit
Who publishes New people drawn in by a format, or pushed out by a payment floor Farcaster Frames; Spotify's 1,000-stream minimum

Two things in that table deserve emphasis. First, the shift usually happens without anyone breaking a rule, which is why rule enforcement alone rarely fixes it. Second, where platforms did respond, they changed the incentive rather than asking creators to behave better: YouTube changed the measure, Spotify added a floor, X cut off the apps. Facebook is the counter-example. According to the Journal, Mark Zuckerberg resisted some proposed fixes because he worried they would hurt engagement, and only in 2021 did Facebook announce tests that put less emphasis on how likely someone is to comment on or share political content [1]. A distorted incentive rarely corrects itself; someone has to decide to change it.

Five questions before you design or join a creator programme

These apply to anyone who runs a rewards programme, a leaderboard or a revenue share, and to anyone deciding whether to take part in one.

  1. What exactly is counted? Name the unit: a click, a minute, a comment, a post, a vote. Then imagine the cheapest piece of work that maximises it. That is what you will get more of.
  2. Can the unit be split or repeated? If one long piece can be cut into ten short ones, or one opinion posted as ten replies, expect it to be.
  3. Does the count reward reaction or understanding? Comments and shares are easiest to get with conflict. Measures of reading or return visits are harder to provoke.
  4. Who stops publishing? Every incentive also tells some people their work no longer pays. Ask who those people are before you launch.
  5. How will you notice, and what will you change? Decide in advance which sign would tell you the incentive is distorting the work, and be ready to change the rule rather than blame the people following it.

A worked example (fictional)

The project, the leaderboard and all figures below are invented to show the questions in use.

A new crypto wallet wants attention before launch. It sets up a leaderboard: every post on X that mentions the wallet earns points, and the top 500 accounts each month share a pool of future tokens.

Run the questions. The counted unit is a mention, so the cheapest winning post is a one-line reply containing the wallet's name. That unit can be repeated endlessly, across every popular thread. The count rewards visibility, not understanding, so few posts will explain what the wallet actually does. People who write careful reviews will find that a single long thread scores the same as one short reply, and many will stop bothering. Within weeks the wallet's name is everywhere and almost none of the posts help anyone decide whether to use it.

A version with the same budget: pay a few independent writers a fixed fee for in-depth reviews, published with clear disclosure, and let readers reward the useful ones by reading and sharing them. It produces far fewer posts. Each of them can be judged on its own merits.

What this article cannot tell you

The six cases come from different platforms, years and kinds of evidence. The Facebook figures come from internal documents as reported by The Wall Street Journal; I could not read the documents themselves, and the Polish party's 80% figure is that party's own estimate. YouTube's announcement reports direction, not size, and does not describe how content changed. Spotify's statements about bad actors are the company's own account. The InfoFi account rests on news reports of X's announcement and Kaito's response. None of the cases is a controlled experiment, so they show a pattern, not a measured effect size. Nothing here measures what Dapp.Expert readers or authors would do under any particular reward. This is not legal, financial or investment advice.

How this article was made

Task What the AI did What a person did How it was checked
Research Collected material on creator programmes, metrics and author rankings into research notes Chose the question and the cases to rely on Each figure and quotation checked against the original page or official record [1]–[8]
Draft Wrote the first draft from the notes Edited the text, the structure and the worked example A separate review compared figures and claims with the sources
Images Generated the cover photograph-style illustration; built the diagrams from a code template Chose the illustration and the content of the diagrams Labelled as AI-generated in the disclosure line and alt text
Not done by AI — The decision to publish and responsibility for the text —

Conflicts of interest: none. The author has no commercial relationship with the platforms, companies or researchers named; the wallet and leaderboard in the example are fictional.

Responsible for this article: Dmitriy Lastovka, author. Corrections: contact form.

Sources

Labels: official (the platform's own announcement), report (a news report), paper (a study, read via its official abstract or record).

  1. Keach Hagey, Jeff Horwitz, "Facebook Tried to Make Its Platform a Healthier Place. It Got Angrier Instead.", The Wall Street Journal, 15 September 2021. https://www.wsj.com/articles/facebook-algorithm-change-zuckerberg-11631654215 Report, read in full via the copy entered into the record of the US House Energy and Commerce subcommittee hearing of 1 December 2021: https://www.congress.gov/117/meeting/house/114268/documents/HHRG-117-IF16-20211201-SD012.pdf
  2. YouTube, "YouTube search, now optimized for time watched", YouTube Official Blog, 12 October 2012. https://blog.youtube/news-and-events/youtube-search-now-optimized-for-time Official.
  3. Spotify, "Modernizing Our Royalty System to Drive an Additional $1 Billion toward Emerging and Professional Artists", Spotify for Artists, 21 November 2023 (updated: live from 1 April 2024). https://artists.spotify.com/en/blog/modernizing-our-royalty-system Official.
  4. Jason Shubnell, "Kaito token plummets after X revises API policies to ban InfoFi crypto projects", The Block, 15 January 2026. https://www.theblock.co/post/385805 Report.
  5. André Beganski, "Elon Musk's X Bans Access to 'InfoFi' Crypto Projects Amid 'AI Slop' Backlash", Decrypt, 15 January 2026. https://decrypt.co/354736/elon-musks-x-bans-access-infofi-crypto-projects-ai-slop-backlash Report.
  6. Chao Li, Balaji Palanisamy, "Incentivized Blockchain-based Social Media Platforms: A Case Study of Steemit", ACM WebSci 2019, arXiv:1904.07310. https://arxiv.org/abs/1904.07310 Paper (abstract).
  7. James Hunt, "Daily active users on Farcaster surge to all-time high, driven by Frames launch", The Block, 5 February 2024. https://www.theblock.co/post/275971 Report.
  8. Donald T. Campbell, "Assessing the impact of planned social change", Evaluation and Program Planning 2(1) (1979), 67–90. https://doi.org/10.1016/0149-7189(79)90048-X Paper (official record; the idea is summarised, not quoted).

Tier I

Sector: #18291

Sealed Hiding Place Room

Resource Cache

Resource Cache

Tier I

Requires 25% Tier Progress to Claim
Meme Cache

Meme Cache

Tier I

Requires 50% Tier Progress to Claim
Equipment Cache

Equipment Cache

Tier I

Requires 75% Tier Progress to Claim

After collecting, hiding places will be stored in your inventory and can be opened with Keys.

Other articles

Measuring an Author’s Contribution Beyond Output

The easiest number about an author is how much they published, and it says the least. Five layers of contribution, and what each one can and cannot show.

user avatarme1afon

Why Persuasion Principles Are Not Universal Laws

Some classic persuasion effects still hold after 45 years; others shrank to almost nothing when large teams repeated them.

user avatarme1afon

When Rewards Crowd Out Motivation

A fine for late parents doubled lateness. Research on incentives shows when a reward replaces the reason people acted, and when it helps.

user avatarme1afon

Scarcity Claims That Readers Can Verify

Scarcity really does make things look more valuable, which is why it is so often faked. What a scarcity claim needs before a reader can check it.

user avatarme1afon

Offering Value Without Creating a Debt

A small gift can make people feel they owe something back. Field experiments, drug-company meals and crypto airdrops show when giving quietly creates a debt.

user avatarme1afon

When Social Proof Makes a Bad Decision Easier

Following others is often a sensible shortcut. Research on conformity, online ratings and crypto manipulation shows when it misleads. Five questions to ask.

user avatarme1afon

Important disclaimer: The information presented on the Dapp.Expert portal is intended solely for informational purposes and does not constitute an investment recommendation or a guide to action in the field of cryptocurrencies. The Dapp.Expert team is not responsible for any potential losses or missed profits associated with the use of materials published on the site. Before making investment decisions in cryptocurrencies, we recommend consulting a qualified financial advisor.