How Can Publishing Platforms Protect Quality in an Age of Mass AI Generation?

—A proposal to attach refundable deposits to works and “likes”

The easier AI makes it to create content, the harder it may become to choose what to read or watch.

Videos, blog posts, novels. More works do not give their audiences more time. Carelessly posted content can bury works that their creators truly want people to see. That is the problem on my mind.

In the previous article, I considered attaching refundable trust deposits to emails and calls to make mass unsolicited communication harder.

As that conversation continued, another thought occurred to me. Could the same idea apply to publishing works and evaluating them?

Paying a Deposit When Publishing a Work

Imagine a fiction platform that asks you to deposit ¥10,000 for 30 days when you publish a novel. These amounts and periods are hypothetical settings to explain the mechanism.

The money is returned after the agreed period. It is not a fee permanently collected from the author.

In this article, I also refer to this refundable deposit as collateral.

During that period, the author cannot use the ¥10,000 for anything else. They temporarily allocate their own funds to publishing the work.

Personally, I could imagine choosing to set aside ¥10,000, or even ¥100,000, for a time if it helped people see a work I had put real effort into.

Still, this is not a small burden for everyone. I want to return to that point later.

What would happen with mass posting?

If each work requires a separate ¥10,000 deposit and 1,000 works are awaiting refunds, the funds required would total ¥10 million. Even if someone can generate all those works, their funds available for publication are limited.

The actual amount required is the deposit per work × the number of works whose deposits have not yet been returned. By posting over time and reusing refunded money, someone could publish many works with less capital.

What this method directly constrains, then, is the speed of posting large numbers of works within a given period.

Giving Authors a Reason to Choose What to Publish

With such a system, authors might think about which works to release before posting everything they can make.

From many generated candidates, they could select one they want to show, even if it needs revision. They would decide which work deserves their limited available collateral.

I think that selection matters for AI-assisted works too. Even as AI increases production capacity, authors still choose what to put into the world.

The aim here is not to sort works solely by how they were generated. It is to introduce a burden at publication that gives authors a reason to choose.

But depositing money does not necessarily make a work better. Someone with enough capital could still publish a large volume of careless work. If the profits exceed the burden of locking up funds, there would still be an incentive to continue mass production.

Even so, could this be one way to make nearly unlimited posting harder?

Could “Likes” Carry Collateral Too?

Another idea I found interesting during the conversation was to ask the people evaluating works to put up collateral as well.

For example, someone could deposit ¥1,000 for 30 days to recommend a work.

This differs from a tip paid to the author. After the agreed period, the money goes back to the person who made the recommendation.

In addition to an ordinary “like”:

I am setting aside ¥1,000 for a time to recommend this work.

They could make that statement of intent.

A click saying “I enjoyed this” gains an additional action: temporarily making one's own funds unavailable. To me, that seems to lend weight to the recommendation's credibility.

The author's collateral for publication, and the reader's collateral for recommendation. Two kinds of action become visible around a work.

However, a deposit does not reveal a person's inner feelings. What we can observe is that they left a specified sum unused for a specified period.

I wonder whether that fact could be one indicator when considering how committed someone is.

Do Not Rank Works by Money Alone

What concerns me is turning this directly into a ranking where works with larger deposits rise to the top.

That would give too great an advantage to authors and supporters with more money.

Even with the same ¥1 million total, one person depositing ¥1 million conveys different information from 1,000 people depositing ¥1,000 each.

On a work's page, I would want to show the author's deposit, the number of recommenders, the total recommendation deposits, and the distribution of amounts per person separately.

If these are used for ranking, measures such as capping the influence of one person's large deposit would be needed so that money does not become endlessly powerful.

Head counts are not automatically trustworthy either. One person might create many accounts, or an author might recommend their own work under another name.

Even with collateral, questions remain about who is recommending a work and their relationship to its author.

There Is No Point in Excluding Committed Newcomers

We cannot universally say, “If you are serious, you can spare ¥10,000.”

People can put their heart into a work while having little money available. Some cannot lock up funds needed for daily life. The size of a deposit cannot stand in for the strength of someone's enthusiasm.

For example, a platform could allow a few submissions for free and require collateral as posting volume increases. For recommendations, it could retain an ordinary route for leaving comments and ratings.

Such a design might constrain mass posting while leaving room for new authors and readers to participate.

On the other hand, repeatedly claiming free allowances through many accounts would weaken that constraint. Accessibility and resistance to abuse would have to coexist.

“I Did Not Enjoy It” Should Not Be a Reason for Forfeiture

People have different tastes in novels and videos.

I would not want a system that confiscates an author's deposit merely because readers did not enjoy the work, or takes away a reader's collateral because their recommended work failed to become popular.

That would disadvantage unusual styles and works that are not yet understood.

The basic proposal is to lock funds for a fixed period. Even if penalties for violations were added, their conditions and decision procedures would have to be defined separately from preferences about works.

A deposit system does not eliminate the need to evaluate quality or operate and moderate a publishing platform.

Using Collateral to Give Actions Weight Has Precedents

When I asked AI to investigate this idea more broadly, similar mechanisms already existed. Their shared feature is putting up one's own funds to participate, make a claim, or support a chosen action.

But their purposes, refund timing, and conditions for losing the money differ.

For example, the Tokyo District Court provides instructions for returning real estate auction bid deposits to designated bank accounts. Although this is a different use from a publishing platform, traditional transactions also attach refundable money to an intention to participate. Tokyo District Court guidance

The following examples are closer to this proposal.

Requiring Collateral from Both Submitters and Challengers

Kleros Curate is a mechanism for registering items on lists with specified eligibility conditions. Its official documentation describes a deposit paid by the registrant and returned when registration is accepted. Challengers also put up deposits, and disputes enter an adjudication process. The losing side may forfeit its deposit. Official Kleros documentation

I read this as a design that asks both the person submitting information and the person criticizing it to take responsibility for their claims.

However, the decision concerns whether an item meets the list's registration criteria. It cannot simply be transferred to judgments about whether a novel is entertaining or a video is appealing. Kleros's adjudication mechanism is also distinct from a national court.

In my proposal, readers deposit collateral to recommend a work, rather than to challenge it. Both designs involve deposits on two sides, but they ask different things of evaluators.

Research on Deposits Backing an Article's Accuracy

A 2025 study examines a proposal in which publishers put up a “Veracity Bond” backing the accuracy of articles, while challengers also provide collateral. One side can lose its deposit depending on the adjudication result.

The researchers conducted three scenario experiments in which participants read descriptions of hypothetical media systems. In Study 3, with 192 participants, a publisher's bond strengthened the impression that the author was responsibly engaged. In contrast, requiring a challenger to put up collateral weakened perceptions of the system's fairness. Research paper by Barbosa and colleagues

This was not an experiment demonstrating improved quality or reduced spam on an actual publishing platform. Still, I see it as research close to this idea: committing one's own money can influence the audience's judgment.

Recommenders and challengers occupy different positions, so these results cannot be directly applied to my proposal. Even so, we need to consider how people perceive an entry requirement involving collateral, and whether it weakens the voices of those with fewer funds.

A Patent Case Involving Deposits for Email Spam Prevention

Bonded Sender, the spam prevention service discussed in the previous article, also appeared in a related court case.

The 2012 decision in CBT Flint Partners v. Return Path in the U.S. District Court for the Northern District of Georgia describes a system where senders posted deposits, complaints about unwanted sending triggered deductions, and deposits were returned in the absence of abuse. The court held that this deposit did not constitute an “advertising fee” within the meaning of the patent at issue. Reproduced court opinion, 870 F. Supp. 2d 1369

This was a patent infringement dispute, not a demonstration of the deposit approach's effectiveness. According to the opinion, the successor service removed the deposit requirement in 2006.

Within the limited context of interpreting patent terminology, this is an example of a refundable deposit being treated differently from a permanently collected fee. The money I want to attach to posting and recommending would also be returned after a fixed period. I want the burden of its temporary unavailability to become a reason to choose one's actions.

An Experiment on Changing Behavior and a Mechanism Connecting Support to Works

Outside online trust, there is a smoking cessation experiment conducted in the Philippines called CARES. Participants deposited money for six months; passing a test meant getting it back, while failure meant it was donated. A 2010 paper reports that 11% of those offered the program participated, and the offered group's six-month test pass rate was about three percentage points higher than the control group's. Research paper by Giné and colleagues

Depositing money to continue an action one has chosen is a way to constrain one's future behavior, known as a commitment device. But this experiment involved potential forfeiture and an intervention that included collection visits. Its effect cannot be attributed solely to depositing funds, nor can we assume the same effect on the quality of submissions.

An example closer to supporting creative works is Subsocial's Content Staking, which came up in the conversation with AI. Official documentation describes locking tokens and rewarding users and creators through actions such as “likes.” Official Subsocial documentation

This connects support and rewards through a shared funding pool. It differs from placing separate deposits on individual works, but it is a precedent linking evaluation and locked funds.

Looking across these examples, collateral has at least three uses: encouraging action, creating a loss for violations, and expressing an intention to an audience.

The central aim of my proposal is to make mass posting harder through funds locked per work, and to make authors' and recommenders' choices visible. When borrowing from precedents, I want to distinguish which function I am adopting.

This Could Lead to Another View of Influencers

During the conversation, I wondered whether this might also connect to a new way of thinking about influencers.

How many followers does someone have? How many views? Alongside those figures, we could look at how many people supported the works they recommended by locking their own funds for a time.

It might provide another angle on the influence of a recommendation.

However, adding up deposits over several years would count the same money repeatedly. Any display would need to separate funds currently locked from the historical record.

If a recommender receives rewards or financial backing, I would also want those relationships disclosed. Even a collateral-backed evaluation is not free from advertising or conflicts of interest.

I Want a Place Where the Decision to Choose a Work Is Visible

What I wonder about building is a publishing platform where the decisions of authors choosing what to publish and readers choosing what to recommend become visible.

Refundable collateral would not automatically guarantee quality. Nor would it precisely measure commitment itself.

Still, it might constrain careless mass posting and mass rating, and give people a reason to select the works they bring.

Could that lead to a community or platform that maintains a certain level of quality? That is my hypothesis here.

I am extending the Proof of Commitment considered in the previous article to works and recommendations. I use this name to explain my concept. The underlying idea of giving actions weight through collateral already has precedents.

I want you to see this work. What am I willing to put up for that?

A place where that action also informs judgment may have value precisely because AI is increasing the number of works people can create.

About the Use of AI

The concept emerged from my conversations with ChatGPT. I used ChatGPT to organize the article and research related official documentation, papers, and court cases. I also gave Claude the article and a fact memo containing verified information to review the logic and wording. This article draws on descriptions of mechanisms and research findings; it does not verify the services' current operating status or test the effectiveness of my proposal in practice.

---

The Exchanges Between Me and AI That Led to This Article

These are readable summaries of actual conversations, preserving their substance while editing the wording.

Me: I think the email deposit idea could also apply to videos, blogs, and fiction platforms. For a work I put real effort into, I could choose to deposit ¥10,000 for a while. But that would be a major burden for someone carelessly posting in bulk. Could we build a platform like that?

AI: Requiring refundable collateral for each submission would increase the funds needed as posting volume grows. However, the design should consider newcomers who cannot provide those funds, balancing the burden on mass posting with ease of participation.

Me: If I say, “I want you to see this; I put my heart into it,” personally I could accept having around ¥100,000 tied up. And if a like came with ¥1,000 in collateral, wouldn't that give it more weight? It would become a question of how much money the evaluator sets aside to recommend a work.

AI: We can consider an author's publication collateral separately from a reader's evaluation collateral. What is observable is the act of locking funds for a fixed period, rather than an inner state. Ranking by money alone would also make financial resources too influential.

Me: So it could be an indicator when considering the commitment of both creators and evaluators. In an age flooded with mass-generated content, I think we might build a community or platform that maintains a certain level of quality.

AI: Collateral itself does not guarantee quality, but it can be considered as a proposal to impose economic constraints on mass posting. The concept uses the actions of authors and readers as two sources of information for choosing works.

Me: Please investigate the idea of requiring collateral more broadly. If there are similar examples or court cases, I would like to incorporate them into this article.

AI: There are mechanisms requiring deposits from both registrants and challengers, and research on bonds backing article accuracy. There is also a patent case involving deposits for spam prevention. However, the existence of precedents and an improvement in quality on the proposed platform need to be considered separately.