Blocking the Exit Doesn't Protect Your Platform. It Protects Mediocre Creators.

Based on the research ofSheng and Chen, "Cross-Platform Links: Blessing or Curse?," MIS Quarterly, 2026

The ban is the obvious move. It is not the optimal one

Every platform operator inherits the same instinct: if a user might leave, close the door. A link in a creator's bio that points to a competing app is, on its face, a leak in the wall. Conventional platform-governance wisdom treats cross-platform links as pure downside, a direct, measurable transfer of attention and ad revenue to a rival.

Jianchao Sheng and Jianqing Chen built a game-theoretic model to test whether that intuition survives contact with how creators actually behave, and it does not, at least not universally. Their setup has two competing digital-content platforms, each hosting a creator who chooses how much effort to put into content quality, and a population of users who differ in how well their home platform's content fits their taste. A cross-platform link lets a user follow the creator to the rival platform when the fit at home is bad enough.

Two effects now pull in opposite directions. The consumption-loss effect is the one everyone already worries about: users with high misfit use the link to leave, and the platform loses the advertising revenue those viewing minutes would have generated. The quality-enhancement effect is the one the ban ignores: once a creator knows the worst-fitting viewers can walk, the creator has a reason to raise quality, both to win new viewers who can now find them through the link and to hold onto the viewers who remain. Better content raises consumption among everyone who stays, and that gain is pure upside the platform would not have captured under a ban.

Which effect wins depends on two numbers: the unit cost of misfit (how painful it is for a mismatched user to stay versus leave) and the average advertising rate (how much revenue a minute of attention is worth). The authors show that when misfit cost is high, both platforms allow the link in equilibrium, because the quality lift from disciplining the creator outweighs the users who would have left anyway. When both misfit cost and the ad rate are low, both platforms ban it, because there is little quality upside to capture and the consumption that would be lost is relatively valuable. In the middle range, where misfit cost is low but the ad rate is high, either allowing or banning can be the equilibrium outcome, meaning the same market can rationally tip either way depending on how the platforms coordinate. Strikingly, the model also shows that both platforms banning the link is its own trap: the link benefits the creator and users regardless, and mutual banning leaves both platforms' revenues worse off than mutual allowing would, a prisoner's dilemma that neither platform can escape unilaterally.

TikTok already built the experiment

This is not a hypothetical. TikTok's own help center documents the mechanism the model assumes. A user can "add a link to a website on your TikTok profile if you have at least 1,000 followers, or if you have a Verified Business Account," and the stated purpose is explicit: "this allows creators to share external websites with their audience, such as their portfolio or other social channels." The platform even tightened the mechanism recently while keeping the substance intact: TikTok "no longer supports links that open or log you into other social media apps," but creators "can still add links to your profile that direct people to those platforms' websites," meaning a creator can still point an audience at a YouTube channel, just through a web link rather than a native deep-link. The gate is not "no linking out." It is "earn the right to link out by first building an audience."

That gate matters, because it is exactly the kind of screening the model's "regular creator" case assumes: a platform does not hand the exit option to every account, only to creators who have already demonstrated enough pull that the quality-enhancement effect has something to bite on. TikTok pairs that gate with the other lever in the model, the advertising rate. In May 2022, the company launched TikTok Pulse, a program that shares ad revenue with creators and media publishers who have 100,000 or more followers, explicitly described at the time as similar to YouTube's existing creator payment program. YouTube's own current eligibility rules for its Partner Program require a channel to clear thresholds, such as 1,000 subscribers plus either 4,000 watch hours in the past year or 10 million Shorts views in the past 90 days, before ad revenue sharing turns on at all. Both platforms are, in effect, paying only for content that has already proven it can hold an audience, which is the same quality bar the cross-platform link is meant to enforce for free.

Put those two policies next to each other and TikTok's choice stops looking like an oversight. It is tightening the ad-rate lever (pay creators more for staying good) while leaving the misfit-exit lever open (let the worst-fitting viewers go find what they want elsewhere) rather than trying to seal both doors at once.

The exit you leave open is not a leak in your wall. It might be the only thing still forcing your creators to be good.

When platforms panic and wall the garden anyway

The counterexample is just as instructive, because it shows what the naive ban actually costs beyond the model's own variables. In December 2022, Twitter rolled out a policy banning users from promoting their accounts on Facebook, Instagram, Mastodon, Truth Social, Tribel, Nostr, and Post, including removing links or usernames for those platforms from bios, with repeat violations triggering permanent suspension. The backlash was immediate and came from inside the building: co-founder Jack Dorsey called the policy nonsensical, Stanford Internet Observatory director Alex Stamos called it "the clearest declaration of weakness I've ever seen from a major US tech platform, and a transparent declaration of anticompetitive intent," and even Elon Musk's own earlier tweets, which argued the test of a healthy system is whether it needs a wall to keep people from leaving, were being quoted back at him within hours. Twitter deleted the policy page the same night without a formal explanation.

Instagram took the opposite bet a few months later. In April 2023, it rolled out support for up to five links in a bio, for all account types, after years of restricting users to a single link, a limitation that had let third-party tools like Linktree build an entire business around the gap. TechCrunch tested the rollout and reported no trouble linking to "competing platforms, like TikTok or YouTube," and noted the feature followed complaints about TikTok's own tighter restriction, which at the time still reserved clickable links mostly for business accounts. Two platforms, watching each other, each recalibrating how open to be, which is precisely the strategic interaction the model describes rather than a one-sided policy choice either company could make in isolation.

The rule: don't confuse the wall with the value

The design question a platform operator should be asking is not "should we allow links to competitors." It is "where do we sit on misfit cost and advertising rate, and how much quality is our creator actually capable of adding if we make the exit credible." A platform with low misfit (most users are already well matched) and a thin ad rate has little to gain from opening the door; the model says ban it. A platform where mismatch is common and painful has the opposite answer, because the creator's quality response is doing real work, recovering consumption from the users who stay and even pulling in new ones who can now find exactly what they want through the link.

The deeper warning is for the middle case, and for industry-wide policy. Sheng and Chen's result that mutual banning can be a prisoner's dilemma, worse for both platforms' revenues than mutual allowing, is the kind of finding that should worry regulators as much as executives: a market where every platform independently chooses to wall off its creators can land everyone, including users and creators, in a worse equilibrium than one where cross-platform links are simply normal. Before a platform tightens its outbound-link policy on reflex, the model suggests a cheaper question than a ban: what happens to the creator's incentive to be good, if staying no longer requires it.