← Back to Companies

CHI Company Investigation · Social Media

YouTubeLong-form video · Shorts · Live · Premium · Music · YouTube TV · San Bruno, California

Sixteen research streams · Verdict: weakly supports · No numerical score assigned

YouTube pays creators better than anything else at its scale.The finding is not in the rate. It is in the boundaries drawn around it.

This investigation was commissioned to test whether YouTube increasingly retains unilateral control over monetization, distribution, enforcement, product and advertising while shifting economic, procedural and dependency risk onto creators and viewers. On the headline bargain the answer is no, and the evidence saying so is YouTube's own: 55% of net ad revenue on long-form, documented from at least 2015 and never publicly reduced, with no arbitration clause and no class-action waiver in the Terms. The hypothesis survived somewhere else entirely — at the perimeter, where the platform decides who is inside the bargain, who adjudicates a dispute, what a pooled figure contains, and what it takes to get a wrong decision reversed.

16Research streams
23Findings survived review
31Improvements credited
0Numerical score assigned
Estimated read 40 min

The centre The headline bargain

On the terms everyone can read, YouTube is the most generous actor in its market.

  • 55% of net ad revenue on long-form watch-page ads — documented as the contractual figure from at least 2015 and never publicly reduced in any period for which primary evidence exists.
  • No arbitration clause and no class-action waiver in the current Terms, against peers who impose both.
  • Thresholds came down in 2023, not up: a new tier at 500 subscribers opened fan funding to smaller channels.
  • Shorts got a revenue share, not a capped bonus fund — an industry first, uncapped, still running.
  • Thirteen relaxations of the ad-suitability rules since the change log opened in 2019, including three first-party admissions of over-restriction.
  • Between 45% and 56% of estimated total revenue goes out to creators, artists and media companies, depending on window and denominator.

The edge Everything that decides whether the bargain is yours

On the terms that decide whether any of that applies to you, YouTube decides alone.

  • Who is inside the Partner Program — and when the bar moves. It has moved four times between 2017 and 2027, in both directions, with notice ranging from thirty-five days to none.
  • What happens in the five days between a copyright claim and a dispute — a boundary untouched by any reform since 2016.
  • Who decides that dispute. The claimant does, at first instance and on appeal. “YouTube doesn't mediate copyright disputes.”
  • What a pooled or “net” figure contains. The Shorts pool and the newly disclosed Premium pool are published as rules, never as figures. There is no audit right.
  • What a wrong decision costs, and what reversing it requires. One appeal, no published resolution time, and a documented pattern of correction that follows attention.
  • Whether you can leave with what you built. Your videos export. The relationship with the people who subscribed to them does not.

Both columns are true at the same time, and neither cancels the other. That is why this page gives YouTube its credit first and at length: a reader who has not absorbed how good the centre of this bargain is will misread everything that follows as an attack on a platform that pays badly. It does not pay badly. It pays better than anyone else at its size — and reserves to itself every question the published rate does not answer.

The test

Control is not hostility. The question is who carries the cost when control is used.

Every platform at YouTube's scale controls eligibility, distribution, enforcement, monetization rules, advertising architecture and product design. Finding that YouTube does so is not a finding at all.

The CHI question is narrower and harder: when YouTube exercises that control, who bears the economic loss, the uncertainty, the procedural burden and the loss of agency? A platform that retains discretion and absorbs the cost of its own errors is not hostile. A platform that retains discretion and routes the cost outward is the thing this index exists to detect. Throughout this page the two are kept apart, and where the evidence only establishes control, the page says so and stops.

Limb A · Unilateral control

A question of documented architecture: Terms, Help pages, program terms, court holdings. Documented, and not seriously contestable. US courts have confirmed there is no external check on YouTube beyond its own contract, and the contract reserves discretion throughout.

A finding on Limb A alone is not a hostility finding. It is the precondition for asking the real question.

Limb B · Risk shifted outward, increasingly

A question of change over time and of who pays when a control is used. Supported for the sub-threshold tail, for Content ID uploaders, for the scope of the AI-training licence and for pooled-construct opacity. Inconclusive for free viewers and for Shorts creators from 2027.

Weakly refuted for established long-form partners since 2023 and for Premium subscribers throughout. No group in the current era reaches the top band.

Creator monetization: the headline shareWeakly refutes55% stable; payouts rising; best published standing rate in the market.
Creator monetization: the boundaryWeakly supportsRight to Monetize; entry thresholds; pooled constructs with no audit right.
Copyright and Content IDWeakly supportsClaimant adjudicates; the five-day revenue boundary falls on the uploader.
Enforcement and appealsWeakly supportsAutomated at first instance; one appeal; no published resolution time.
Audience portabilityWeakly supportsContent exports. The subscriber relationship is not exportable.
Advertising load, free viewersInconclusiveSurfaces expanded since 2018; no ad-load metric exists to measure it.
Premium subscribersWeakly refutesNotice every time; legacy cohorts protected for years; EU opt-in in 2026.
Terms and amendment mechanicsWeakly refutesNotice plus export commitment; no arbitration; no class waiver.
P Primary — YouTube or Google documentation, Alphabet filing, court or regulator record.
S Secondary — reputable trade press or law-firm memo citing a named primary source.
C Claim — creator, community or advocacy account, not independently re-verified.
I Inference — the investigation's own reasoning, labelled as such.

The centre of the bargain

Before the finding, the part of the record that makes the finding hard to dismiss as predetermined.

A CHI investigation that cannot state the strongest version of the company's position has not finished its work. On the variable YouTube controls most directly — what it pays and on what terms — the evidence runs against the hypothesis, and it runs against it hard.

55%Of net long-form ad revenue“YouTube will pay them 55% of net revenues from ads displayed or streamed on their public videos on their content Watch Page.” Documented as the contractual figure from at least 2015; no public reduction in any period with primary evidence.P
0Arbitration clausesThree independent retrievals of the current Terms found no arbitration clause and no class-action waiver. Disputes are litigated in the courts of Santa Clara County. Among scaled consumer platforms this is unusual, and it favours the creator.P
500Subscribers, from June 2023An expanded lower tier — 500 subscribers, three uploads in 90 days, 3,000 watch hours or 3M Shorts views — opened fan funding to channels far below the full bar. The threshold moved down.P
45%Shorts pool share, uncappedFrom February 2023 Shorts moved from a discretionary $100m fund to a monthly revenue share. No competitor matched it. The share has not been cut since.P
13Ad-suitability relaxations since 2019Against roughly fifteen restrictions, with relaxation dominating since 2023 — and three occasions on which YouTube said in its own words that its guidelines “had become too restrictive.”P
>$100bnPaid out, four-year windowTo creators, artists and media companies. Computed against estimated total revenue that is between 45% and 56% — a range, not a point, because Alphabet does not disclose the components.PI

What the alternatives publish

The comparison below is not a claim that YouTube is the best platform for a creator on every dimension. It is narrower: among scaled video platforms, YouTube is the one that publishes a standing percentage and has kept it. Several competitors publish no percentage at all, and several have restructured their creator programmes twice in four years. Competitor instability is itself a switching cost, and it belongs in YouTube's column.

Published standing creator shares across scaled platformsP platform documentation S trade press
PlatformPublished standing shareStability of the programme
YouTube55% long-form net ad revenue; 45% of the Shorts pool; 70% on commerce after tax and store feesLong-form share documented from at least 2015 and not publicly reduced
Twitch50/50 on subscriptions, with a 70/30 tier available on qualifying termsCut for partners in 2022; partially reversed in 2024
Spotify (podcast partner programme)50%, from January 2025, in a limited set of marketsNew; no long record
Rumble60%, with long exclusive agency options attachedExclusivity terms materially change the comparison
TikTokNo percentage publishedCreator fund replaced twice
Meta (Facebook / Instagram)No percentage publishedReels Play bonuses ended March 2023
XNo percentage publishedCreator payouts restructured twice

Read this table for what it is. It measures disclosure and stability of a published rate, not total creator outcomes, which depend on audience, niche, sponsorship and geography. It also does not measure the things this page goes on to examine, none of which appear in a revenue-share percentage.

Who gets inside

The share has been stable for a decade. The question of who is entitled to it has been reopened four times.

The Partner Program is the door. Everything in the previous section — the 55%, the Premium allocation, the Shorts pool, access to a human being at Creator Support — is on the far side of it. YouTube sets the height of that door, and has moved it in both directions.

Partner Program entry requirements · 2017 → 2027P

6 April 2017 · raised
Different unit — not to scale 10,000 lifetime views Announced with immediate effect on existing members: “Starting today, we will no longer serve ads on YPP videos until the channel reaches 10k lifetime views.” No notice period.
16 January 2018 · raised
4,000 watch hours 1,000 subscribers + 4,000 watch hours Thirty-five days' notice, and existing members below the new bar were removed. YouTube's characterisation: 99% of those affected were earning under $100 a year, 90% under $2.50 in the previous month.
13 June 2023 · lowered
3,000 watch hours 500 subscribers, expanded tier Three uploads in 90 days plus 3,000 watch hours or 3M Shorts views. Fan funding only — no ad or Premium revenue — but the bar moved down for the first time, and it counts.
Current full tier
4,000 watch hours 1,000 subs + 4,000 hours or 10M Shorts views Qualified watch hours over twelve months, or qualified Shorts views over 90 days. Review typically takes about a month. Membership lapses after six months or more of inactivity.
1 February 2027 · raised
8,000 watch hours 1,000 subs + 8,000 hours or 20M Shorts views Announced 10 August 2026. For new applicants the watch-hour bar doubles and the Shorts bar doubles. Existing members keep their place in the programme.

What the chart shows and what it does not. Bars are drawn to a single scale — the watch-hour requirement, indexed to the 8,000-hour bar that applies from 2027. The 2017 gate was expressed in lifetime views, a different unit, so it carries no bar at all; the alternative Shorts routes and the fan-funding tier are likewise not commensurable with watch hours and are stated in words instead. The direction of travel is not one-way — 2023 moved the bar down — and existing partners have been grandfathered on entry every time since 2018. The finding here is not that thresholds exist. It is that the class of creators outside the programme is defined by YouTube, has been redefined four times, and, since November 2020, is monetized anyway.

Mechanism one

Two channels can upload the same video. Only one of them is entitled to be paid for it.

This is the cleanest structural asymmetry in the entire record, and it is not hidden: it is in the Terms of Service, in YouTube's own words, and it has been there since 18 November 2020.

You grant to YouTube the right to monetize your Content on the Service … This Agreement does not entitle you to any payments. YouTube Terms of Service, “Right to Monetize,” effective 18 November 2020, retained in the 15 December 2023 Terms P

Read the two halves together. The first grants YouTube the right to run ads against any upload on the service. The second disclaims any corresponding obligation to share what those ads earn. For a channel inside the Partner Program the gap is closed elsewhere, by the programme terms that set the 55%. For a channel outside it, nothing closes the gap. At the time the clause took effect, YouTube told non-partner creators directly: “Since you're not currently in YPP, you won't receive a share of the revenue from these ads.”S

Identical video. Identical ad inventory. Identical viewer. The only variable is the uploader's contractual status.

Same content · different entitlement

Inside the Partner Program

Meets the current threshold and has accepted the module terms
YouTube may run ads against the uploadYes
Creator receives a share of that revenue55% net, long-form
Creator can turn monetization off per videoYes
Access to Creator Support chat and emailYes
Per-video revenue reporting in StudioYes

Outside the Partner Program

Below the threshold, not yet reviewed, or removed from the programme
YouTube may run ads against the uploadYes
Creator receives a share of that revenueNone
Creator can opt out of monetization for that classNo
Access to Creator Support chat and emailNo documented channel
Per-video revenue reporting in StudioNot applicable

The precise finding, and its limits. This is not theft, it is not unlawful, and YouTube does not “take creator revenue” — a non-partner channel was never owed a share, and the clause says so in advance and in plain language. The finding is class-separated monetization and entitlement: a class of uploader that YouTube monetizes, that receives nothing, that cannot opt out, and whose boundary YouTube itself has drawn and redrawn four times between 2017 and 2027 — including, from February 2027, drawing it further out for new applicants.

The counter-argument, stated fairly. Free hosting, transcoding, storage, bandwidth, search, recommendation and abuse-prevention for an unlimited number of uploads is an expensive service delivered at no charge, and advertising is how it is funded. A rule that says “we fund the free tier from ads, and a revenue share is a benefit of the programme you have not yet joined” is coherent and disclosed. What the defence does not answer is the opt-out: a creator who would rather carry no advertising at all, and forgo nothing because they are earning nothing, has no route to that outcome.

Mechanism two

A copyright claim starts a clock the uploader is not told to watch. It runs for five days.

Content ID is the largest rights-management system on the internet: 2.50 billion claims in 2025, against roughly 4,450 active claimants. Around 0.5% are disputed. Where a dispute is filed, the uploader prevails in 67.42% of cases, and in about 75% of those that reach appeal. More than 90% of claims monetize a video rather than block it. On the numbers, the system mostly works, and mostly leaves content online.PS

The finding is not in those numbers. It is in a timing rule that has sat unchanged since 2016 and that decides, before anyone has adjudicated anything, which party keeps the money earned while the question is open.

Revenue during a Content ID disputeScale: 0–30 days from the claim date · scrolls sideways on a narrow screen · verified against YouTube Help, 12 September 2026P
Day 0 · claim filed Day 5 · boundary Day 30 · claimant deadline
Dispute filed on day 2Inside the five-day window
Held from the claim date — every cent is preserved pending the outcome
Dispute filed on day 12Outside the five-day window
Released to the claimant Held from the dispute date
If the dispute is timely

“If you dispute a copyright claim within 5 days, any revenue from the video will be held, starting with the first day the claim was placed.” If the claim is later released, the held revenue is the uploader's. The escrow does exactly what an escrow should.

If the dispute is late

“If you dispute a copyright claim after 5 days from the original claim date, we'll start holding revenue the date the dispute is made” — and “after 5 days from the claim date, any held revenue is released to the claimant.” That earlier money is not recovered if the claim later falls away.

Why this is the sharpest procedural finding in the record. The escrow was itself a 2016 reform, introduced under pressure, and it is real: disputed revenue genuinely is preserved. But the boundary it draws allocates an unrecoverable loss on the basis of the uploader's reaction time, not on the basis of who turns out to be right. A creator who is on holiday, who misses a notification, or who spends a week working out whether a claim is worth contesting, forfeits that week's earnings on a claim that may be withdrawn the moment it is challenged. Nine years of reform — timestamps on manual claims, the short-use rule, the Copyright Match Tool, Erase Song, annual transparency reporting — have not touched it.

Mechanism two, continued

When an uploader disputes a claim, the party that decides the dispute is the party that made it.

The sequence is documented on YouTube's own pages. A claimant files. The uploader may dispute. The claimant then has 30 days to respond and may release the claim, reinstate it, or escalate to a copyright removal request. If the uploader appeals a reinstatement, the claimant has seven days to respond to that too, and may issue a legal notification that removes the video and lands a copyright strike. Three copyright strikes in 90 days puts the account, and any associated channels, at risk of termination.

YouTube doesn't mediate copyright disputes. YouTube Help, copyright dispute process P

That sentence is the architecture. It is also legally rational: under the DMCA safe harbour, a platform that adjudicates ownership takes on exposure it is not required to carry, and US courts have repeatedly confirmed that the neutral posture is lawful. One appellate court has held there is no affirmative cause of action against YouTube for how it applies its own repeat-infringer policy; another court observed that a takedown notice “is miles away from substantive proof of copyright ownership or infringement.”S

But consistent with the statute is not the same as compelled by it. Nothing in the DMCA requires that the party with a commercial interest in the outcome be the first-instance decision-maker on a private contractual claims system that operates far beyond the statute's takedown machinery. That allocation is a design choice, and the costs of it are asymmetric in a way the numbers make visible.

What a false positive costs each party
 Approved claimantUploader
Cost of filing a claim that turns out to be wrongEffectively nil. Claims are matched automatically against a reference; no published sanction attaches to a claim that is simply released.Revenue diverted from the claim date, unrecoverable after day five; time; the risk of a strike if an appeal fails.
Who decides the disputeThe claimant, at first instance and again on appeal.Nobody the uploader can appeal to inside YouTube. “YouTube doesn't mediate.”
What happens if the uploader escalates to a counter-notificationReceives the uploader's legal name, physical address and telephone number.Must disclose them. This is required by the DMCA, not by YouTube, and is not a YouTube design finding.
Abuse rate publishedNone for Content ID partners; no sanction count published either.Published for the public webform: more than 6% of removal requests are a likely false assertion of ownership.P

YouTube's defence of Content ID, stated at full strength

It deserves to be, because most public criticism of Content ID is aimed at a system that does not exist. Content ID keeps videos online in over 90% of claims rather than taking them down. A dispute mechanism exists; an appeal stage was added in 2012; disputed revenue is escrowed. Manual claims carry timestamps since 2019, very short or unintentional music uses can no longer be manually monetized, and claimants who abuse manual claiming lose the tool. The Copyright Match Tool extended detection to millions of ordinary channels that could never have qualified for Content ID. Erase Song lets creators strip claimed audio. YouTube has sued an individual who used false strikes to extort creators, obtaining a $25,000 settlement, an injunction and a written apology in October 2019. And YouTube publishes copyright transparency reporting in a depth no peer platform matches — which is the only reason the dispute and win-rate figures quoted above exist at all.

The narrow finding that survives all of that is not “Content ID is abusive.” It is that the platform gives approved claimants substantial procedural power, leaves the claimant as first-instance adjudicator, publishes no abuse rate or sanction count for that tier, and imposes a five-day revenue boundary whose downside falls entirely on the uploader. That is enough, and the investigation claims nothing beyond it.

Not claimed. That creators are generally pressured out of disputing — fewer than 1% of claims are disputed, but nothing in the record measures why, and uploaders who do dispute win about two-thirds of the time. That YouTube is financially indifferent to who wins: Content ID partner revenue-share terms are not public, so the claim cannot be made in either direction.

Mechanism three

One announcement in August 2026 made three different powers visible at once. They are not the same power.

The changes take effect on 1 February 2027 and have been collapsed in most coverage into a single story about YouTube tightening the screws. They should be kept apart, because two of them are ordinary programme administration and the third is not a change at all — it is a disclosure that does not reconcile with an existing one.

One · Entry

The bar doubles, for new applicants only

From 1 February 2027 a new applicant needs 1,000 subscribers plus 8,000 qualified watch hours in 365 days, or 20M qualified Shorts views in 90 days — double the current requirement on both routes.

Existing members are grandfathered on programme entry and keep their place regardless of the new thresholds.

Widens the unshared class

Two · Shorts floor

A monthly earnings floor that is not grandfathered

To earn from the Shorts Creator Pool in a given month, a creator must maintain 10M qualified Shorts views over the trailing 90 days. This applies to incumbents as well as applicants.

Falling below it does not remove anyone from the Partner Program and does not affect other earnings. It withdraws one revenue stream for that month.

The one term not grandfathered

Three · Premium pool

A percentage appears where there had only been a word

The 2027 materials disclose a creator pool of 30% of net subscription revenue for Premium and 60% for Premium Lite, split between long-form and Shorts.

A current Help page still says the “majority of the revenue will go to our partners,” with no percentage. Press reporting from 2015 put the figure at 55%.

Unreconciled disclosure gap

On the Premium pool, precisely what can and cannot be said

This page does not say that YouTube reduced the Premium creator share from 55% to 30%, and no reader should take that from it. The word “net” is undefined in the 2027 materials: nobody outside YouTube knows what is deducted before the 30% is struck, so 30% of an unknown base cannot be compared with 55% of a different unknown base. It may be a reduction. It may be the same money described differently. It may be more.

What can be established, and was re-checked on 12 September 2026, is that all three statements are currently live and none of them bridges to the others: the Help page promising “the majority,” the 2015 press figure of 55%, and the 2027 pool of 30% of net. The finding is the absence of a reconciliation, not the presence of a cut. That distinction is the entire difference between a disclosure finding and an accusation, and the investigation makes only the first.

Also not claimed. That incumbent creators are being removed from the Partner Program — they are not. That the programme is being de-monetized — it is not; YouTube's own framing is that it expects to pay more to creators in 2027 than in 2026. The narrow finding on the floor is that an incumbent can remain a partner in good standing and still earn nothing from the Shorts pool in a month they fall below a threshold that did not exist when they joined.

Mechanism four

Your content travels. The people who came for it do not.

This is the quietest section on the page and, in combination with the one after it, the most consequential. It is also the section where it is easiest to overclaim, so the wording is deliberately narrow: YouTube does not “own” a creator's subscribers in any legal sense. The audience relationship is platform-bound.

Exports

Via Google Takeout and the data-portability schemaP
  • Uploaded videos, in original format or MP4
  • Playlists and channel settings
  • Video metadata: titles, descriptions, tags
  • The creator's own channel subscriptions — the channels they follow
  • The creator's own comments and community posts

Does not export

No resource exists in the schema for any of these
  • Subscriber identities — including subscribers who have set their subscriptions public
  • Any means of contacting subscribers off-platform: no email, no SMS, no messaging export
  • Comments received on the creator's videos
  • The audience relationship itself, in any portable form
  • Analytics and view history, which is what sponsorship rates are priced against

Why this is structure rather than hostility on its own — and when it stops being only structure. The privacy defence is real and it is strong: subscriptions are private by default, and a platform should not hand over a list of people who did not agree to be handed over. It does not explain why public subscribers, or contactability with explicit consent, are not offered. And the asymmetry becomes load-bearing the moment it is combined with the next section, because termination removes an asset the creator was never able to copy. A creator can rebuild a catalogue. They cannot re-acquire an audience they were never permitted to hold.

Comparison, used carefully. Substack tells writers that what they publish is theirs to own and exports posts and the subscriber list; Patreon creators hold patron contact data. These are not substitutes for YouTube and are not offered as ones — they are different products with different economics. They are cited only to show that non-portability of the audience is a design choice available to be made differently, not a technical inevitability. Across the platforms in this vertical, none exports the follower relationship; that makes it a baseline of the category rather than a way of telling one platform from another.

One further item belongs here because it goes to what a subscription means in the first place. Recommendation, not subscription, is the primary distribution channel: YouTube's last published figure, from 2018, put 70% of watch time as recommendation-driven, and a YouTube growth executive said in September 2026 that the subscriptions feed is “quite small by comparison” with click-through from subscriber sources “about maybe a 10% or less.”S And on 21 April 2026 the “All” notification setting stopped delivering mobile push to subscribers who have not watched the channel in roughly a month and have ignored recent alerts — disclosed, bounded, justified by user behaviour, and still a divergence between a label and what it does.

Mechanism five

Enforcement is automated at first instance, appealable once, and demonstrably faster to correct when someone is watching.

Five hundred hours of video arrive every minute. No human-first moderation system is possible at that scale, and the investigation does not treat automation as a finding. What it examines is the architecture around the automation: how many chances a creator gets, how long it takes, who reviews, and what the record shows about which cases get corrected.

The enforcement architecture, in YouTube's own documentationP
StageWhat is documented
Detection96.1% of removed videos were first flagged automatically in Q4 2023; 96.4% in H1 2024.
NoticeCategory-level. Reasons are given unless doing so would compromise an investigation or the integrity, operation or security of the service.
Appeal“You may appeal each strike only once.” One termination appeal, available within a year. For ad-suitability decisions: “After your one appeal, the reviewer's decision is final.”
ReviewYouTube states appeals go to human review. Ad-suitability appeals “can take up to 7 days.” No published resolution time exists for termination appeals.
EscalationNone inside the published process. Creator Support chat and email are a Partner Program benefit; there is no documented human contact channel for a viewer, a non-partner creator or a Premium subscriber.
ConsequenceTermination extends to “any other YouTube channels” and to channels in which the terminated person is repeatedly or prominently featured. Channels are attributes of a Google Account, which is the unit of enforcement.

Video reinstatements as a share of appeals · YouTube's own filings to the California Attorney GeneralP

Q4 2019 22.0% of appealed videos reinstated
109,000 appeals · 24,000 reinstated
Q4 2023 8.2% of appealed videos reinstated
355,975 appeals · 29,360 reinstated
H1 2024 11.0% of appealed videos reinstated
745,707 appeals · 82,190 reinstated

These are three data points, not a trend, and they are deliberately not drawn as a line. The ratio fell and then partly recovered. It has at least two readings the data cannot separate: first-pass accuracy improved, so fewer appeals deserved to succeed; or appeal review became less generous. The 2019 and 2023–24 figures may not rest on identical definitions. The series also stops: YouTube's filing of 31 March 2026 omits these metrics, citing a 2024 federal appellate decision on the statute that compelled them. That leaves H1 2024 as the last primary reinstatement data in existence for the platform — and it is worth stating plainly that this data exists at all only because YouTube published figures that made its own falling ratio visible.

What the November 2025 termination wave established, and what it did not

In early November 2025 a number of established channels were terminated in quick succession, several for association with unrelated struck channels, and several were restored within a day or two of the cases circulating publicly. On 13 November YouTube responded. The statement is the single most useful document in this stream, because it is the company describing its own correction process without being asked a leading question: YouTube said it had “manually reviewed hundreds of the cases circulating online”, that it upheld most of the terminations, and that only a small number were overturned.S

Read that carefully, in both directions. It confirms that circulation triggered manual review — attention is doing work in the correction pipeline that the published appeal process does not describe. It equally confirms that most of those terminations survived a second look, which is not the picture of a system flailing. Five days earlier, TeamYouTube had stated that “appeals are manually reviewed so it can take time to get a response,” against creators reporting rejections within minutes.S

The finding, at its narrowest

In the publicised cases this investigation examined, correction frequently followed public attention rather than the ordinary appeal path, and YouTube has confirmed that cases circulating online triggered manual review.

The mechanism that produced several of the false positives — termination for being “linked” to a separately struck channel — is documented and structural, not a one-off.

What is expressly not claimed

That ordinary appeals never work: the record contains at least one documented case resolved by ordinary appeal, and there is no data at all on the non-public population, which is almost certainly the overwhelming majority.

That attention is required for correction; that appeals are fake; that AI rejects appeals; that a system fault caused the wave. None of these is supported, and YouTube denies the third.

The same asymmetry shows up in external redress. Under the EU's Digital Services Act, users can take content disputes to a certified out-of-court body. The Appeals Centre Europe reported that through August 2025 it had “received no content from the company” and had decided just 29 of more than 340 eligible YouTube disputes submitted to it, while equivalent agreements with other large platforms had been concluded far more quickly.S Whether YouTube has since reached an agreement could not be established, and if it has, the finding narrows accordingly under the index's principle of crediting fixes in full.

Finally, the remedy where a policy itself is later withdrawn. Creators terminated under COVID-19 and election-integrity rules that YouTube has since retired can, from October 2025, apply after a one-year wait for a new channel. Copyright and Creator Responsibility terminations are excluded. The original channel, its subscribers, its catalogue and its earnings history are not restored.P The programme is a genuine improvement on nothing at all. It is also re-application rather than reinstatement, and the cost of YouTube changing its mind sits with the creator.

Falsification

We tried to prove the rest of it. Here is everything that failed, and why it failed.

A hostility finding is only worth reading if the investigation can show what it looked for and did not find. Most of the widely circulated criticisms of YouTube are on this list. Several of them are the criticisms this investigation expected to confirm.

Hypothesis tested
What the evidence actually shows
Result
Core long-form creator economics have deteriorated
The 55% net share is documented from at least 2015 and has not been publicly reduced in any period for which primary evidence exists. Cumulative payouts have risen at every published milestone. The single most-repeated criticism of YouTube is the one the record most clearly contradicts.
Does not hold
Premium subscribers are treated hostilely
Every price rise carried at least thirty days' notice. The original $9.99 cohort was held at that price for nearly eight years, then given three further months. The 2026 EU increases required subscribers to affirmatively accept. Features were added throughout, and a cheaper Lite tier exists. Family pricing has outrun both inflation and the individual plan — that is a fact about pricing, not a hostility finding.
Weakly refutes
YouTube ran a bait-and-switch on the creator economy
Eighteen candidate “switches” were tested. The pattern holds for three of them and partially for two more. Of the rest, five were externally compelled, several were ordinary offers or outright counter-evidence, and one was retracted within a day of being noticed. It does not hold for the core long-form Partner Program. The better description of the overall sequence is escalation under external pressure, clarification after backlash, and quiet retirement later — a pattern of inconsistency, not a designed switch, and no finding of intent is made anywhere in this investigation.
Does not hold platform-wide
Measured advertising load has increased
The structural ad surface plainly expanded: mid-rolls from ten minutes to eight in 2020, ads between Shorts from 2023, thirty-second connected-TV non-skippables, pause ads, and the removal of per-format creator controls in 2023. But no independent ads-per-hour time series exists and YouTube has never published a viewer-side ad-load metric. The expansion is a ledger. The increase is not a measurement, and this page does not assert it as one.
Unmeasurable
Shorts cannibalise long-form and its economics
No public, independent, causal study exists in either direction. A 2023 newspaper report documented internal concern, sourced anonymously — which is evidence of concern, not of cannibalisation. Shorts revenue and watch-time share are not disclosed by Alphabet. The most that can be said is that Shorts yields are far lower per view: roughly 3–14% of long-form RPM in one 274-channel study.
Not established
Demonetization discriminates against particular creator communities
A 2019 creator-run test found context-blind and inconsistent classifier responses to certain title terms; YouTube attributed this to classifier error. The litigation raising the claim was dismissed on legal grounds without any factual adjudication of the underlying allegations. Inconsistency in a classifier is documented. Discrimination is not, and this page does not allege it.
Not established
YouTube suppresses or shadowbans specific creators
No evidence of intentional suppression of specific creators was located and none is alleged. The defensible statement is far narrower: distribution is mediated by a system YouTube controls, describes only in general terms, changes continuously and does not itemise — and no Help page tells a creator that a particular video has been classified into a reduced-recommendation tier.
Not established
YouTube's interface is built on dark patterns
Applying the FTC, OECD and EU taxonomies strictly to eleven candidate practices, exactly one clears the bar: the cookie-refusal path found non-compliant by the French regulator in December 2021 — and Google remediated it EEA-wide in April 2022, beyond what the order required. Cancellation flows, autoplay, the ad-block interstitial and the sign-in gate do not meet the standard. The term is not used on this page for anything else.
One instance, remediated
YouTube uses AI to reject appeals
YouTube states that appeals go to human review. Creators have reported rejections arriving within minutes, and YouTube has acknowledged automated decisions where its systems have a high degree of confidence that content is violative. Those are two different things, and the record does not bridge them. The claim is not supported.
Not supported
The reinstatement rate has collapsed
The reinstated-to-appealed ratio fell from 22.0% to 8.2% and then recovered to 11.0%. Three data points, possibly non-comparable definitions, two available readings, and a series that ends in H1 2024. That is a gap in transparency. It is not a demonstrated collapse, and calling it one would be the same error this page criticises elsewhere.
Not supportable as a trend
Creators are pressured not to dispute copyright claims
Fewer than 1% of Content ID claims are disputed, and the reasons are not measured anywhere. Uploaders who do dispute prevail about two-thirds of the time, and about three-quarters of the time on appeal — win rates that sit awkwardly with a pressure thesis.
Not established
Meaningful human support requires a network or manager
A single-source creator assertion; not corroborated. What is documented is stratification: Creator Support chat and email are a Partner Program benefit, and viewers, non-partner creators and Premium subscribers have no documented human contact channel at all.
Not established

Why publish a list of things that did not survive? Because the five findings this page does make are only worth anything if the reader can see what was tested alongside them. An index that reported every hypothesis it started with would be an opinion. Adjacent antitrust findings against Google in search, Android, Chrome and app distribution were also excluded throughout — none of them names YouTube, and none is scored here.

The other side

The strongest case for YouTube, put as its own counsel would put it.

This section is not a courtesy. Every finding above was run against it, and several did not survive the encounter.

The economics are the best in the market, and they have held

YouTube runs the largest ad-supported creator economy in existence and pays out roughly half of total revenue. The 55% long-form share is the best published standing rate among scaled video platforms and has never been cut — while one competitor ended its bonus programme, another cut its partners, and two more restructured twice each in the same period.

The contract is unusually favourable

No arbitration clause. No class-action waiver. Disputes go to court. Material changes carry advance notice and an express commitment to let creators export their content. The one clause in YouTube's history that read as an unreviewable termination right was clarified within twenty-four hours of being noticed, never invoked, and has since been deleted entirely.

Automation is a consequence of scale, and everything is appealable

Five hundred hours of video arrive every minute; first-instance review cannot be human. Every action is appealable and terminations are appealable for a year. YouTube published appeal and reinstatement data that exposed its own falling ratio — a disclosure no competitor volunteered and which is the sole reason that figure can be cited against it.

Demonetization costs YouTube too

Ad suitability is an advertiser constraint, not a platform preference: when a video is demonetized, YouTube forgoes its 45% alongside the creator's 55%. The change log shows thirteen relaxations, three explicit admissions that the rules had become too restrictive, an automatic second review of limited-ads decisions since 2025, and a proactive re-review in which YouTube went back through past decisions and reversed them without the creator asking.

Content ID exceeds what the law requires

It keeps content online in over 90% of claims, offers dispute and appeal stages, escrows disputed revenue, timestamps manual claims and strips claiming rights from abusers. YouTube cannot adjudicate ownership without jeopardising safe harbour, and courts have confirmed the neutral posture is lawful. Its copyright transparency reporting is more detailed than any peer publishes.

Much of what looks hostile is required by someone else

Counter-notification disclosure is mandated by the DMCA. Tax withholding is mandated by the Internal Revenue Code. The iOS price differential is a platform commission. The children's-privacy self-designation regime was a regulator's chosen remedy design, not YouTube's. None of these belongs in a hostility column, and none of them is scored here.

No regulator or court has found against YouTube on this question

No court or regulator has found YouTube unfair or deceptive toward creators or viewers. Every US creator viewpoint-discrimination suit has failed. Three years as a designated very large platform in the EU have produced requests for information and no formal proceeding. The single regulator-established consumer-design finding concerned a cookie banner and was remediated.

The investigation's own tags concede the weak points

Prevalence is unquantified throughout: there is no creator-side error rate, no ads-per-hour series, no survey of dependency, and the case register is selected by newsworthiness and therefore cannot measure how often the ordinary process works. The investigation labels every one of these gaps, and none of its findings rests on an unverified item.

The closing submission, and the honest answer to it. A dominant platform with imperfect processes is not a hostile one — that is correct, and it is why the verdict on this page is not SUPPORTS and why three stakeholder groups come out the other way. What the defence does not reach is the specific set of places where the cost of an unresolved question lands on the creator by default: the five days before a dispute, the class that is monetized without a share, the pool whose base is not published, the reversal that arrives with attention. Those are not imperfections in a process. They are allocations, and they run one way.

What changed

YouTube fixes things. The lag between the harm and the fix is the variable worth watching.

Thirty-one dated improvements were logged. Under this index's Fixed Under Pressure principle, a remediation is credited in full for the current customer experience, and the trigger that produced it is recorded separately rather than used to discount it. A fix that happened because of a backlash is still a fix.

Voluntary After backlash Regulatory or political Competitive
Oct 2012Appeal stage added to Content ID disputes — until then the claimant's first decision was the last one.Backlash
Nov 2015Fair-use legal defence programme, funding up to $1m in costs for selected counter-notified creators. No public activity since launch.Voluntary
2016Escrow of disputed Content ID revenue — the reform that created the five-day boundary examined above, and the last one to touch it.Backlash
Jul 2018 →Copyright Match Tool, extended to more than three million channels by 2023 — detection for creators who could never qualify for Content ID.Voluntary
Feb 2019Warning tier and a uniform strike ladder, replacing inconsistent penalties for a first violation.Voluntary
Jul–Sep 2019Timestamps required on manual copyright claims; no manual monetization of very short or unintentional music use; repeat abusers lose manual claiming.Backlash
Dec 2021 →Copyright Transparency Report, published annually and in a depth no peer platform matches. Most of the copyright figures on this page exist because of it.Voluntary
Apr 2022Equal “Reject all” and “Accept all” on the cookie banner, after the French regulator's December 2021 decision — rolled out across the EEA and the UK, beyond what the order required.Regulatory
2022–23The “no longer commercially viable” termination clause deleted — clarified within a day of being noticed in 2019, never invoked, and gone from the current Terms.Backlash
Feb 2023Shorts revenue share replaces the Shorts Fund — an industry first: an uncapped share rather than a discretionary bonus pool.Competitive
Jun 2023Expanded Partner Program tier at 500 subscribers — the entry bar moving down for the first time since 2012.Competitive
Aug 2023Warnings expire after 90 days on completing policy training, justified by YouTube on the basis that an unintentional strike “can financially impact” a creator's bottom line.Voluntary
Dec 2023Current Terms: advance notice plus an export commitment; no arbitration; no class waiver.Voluntary
Mar 2025Automatic second review of videos rated limited-or-no-ads, moving part of the burden of appealing off the creator.Voluntary
Mar 2023 → Sep 2026The profanity rule softened, then abandoned — and past decisions reversed without being asked. A November 2022 update had been applied to back catalogues “regardless of when they were uploaded.” It was softened in March 2023, dropped in July 2025, and YouTube then identified uploads demonetized solely under it, re-reviewed them and restored their monetization ratings with no creator action. The best single instance of remediation in the record — and no back-payment for the intervening period was stated.Backlash, then voluntary
Oct 2025“Second Chance” pilot for creators terminated under since-retired COVID-19 and election-integrity policies: a new channel after one year, not a restoration.Political pressure
2025–26Likeness detection extended from partners to all adult users, and third-party AI-training controls added — though those controls default to off, and cover named third parties rather than Google's own training.Voluntary

What the ledger actually shows. YouTube corrects, repeatedly, and more quickly since 2023 than before it. The number worth watching is not whether a fix arrives but how long it takes: five months for the profanity rule to soften, seven years for manual claims to require timestamps, and nothing at all, in nine years, for the five-day escrow boundary. The lag is the measurement.

The verdict

Weakly supports — and the reasons it is not the band above or the band below are both worth stating.

The index defines this band as: meaningful structural problems exist, but the evidence is mixed, improving, confined to specific customer groups, or partially justified. All four qualifiers apply here, which is unusual, and is why the band fits rather than merely being available.

Why not the band above — why this is not SUPPORTS

No stakeholder group in the current era reaches it. The headline creator share has not been cut. Entry thresholds moved down in 2023. The Terms contain no arbitration clause and no class waiver. Ad-suitability rules have been relaxed thirteen times since 2019, three times with YouTube saying outright that it had gone too far, and once with past decisions reversed proactively.

The clearest surviving findings are also the narrowest: a five-day timing rule, a class-separated entitlement, an unreconciled disclosure, a correction pipeline that runs faster under attention. They are real and they are bounded. Evidence of intent is absent everywhere and no intent finding is made.

Why not the band below — why this is not INCONCLUSIVE or WEAKLY REFUTES

Because the findings that survive are documented in YouTube's own words, not inferred: a contractual right to monetize content it is not obliged to pay for; a five-day boundary that hands unrecoverable revenue to a claimant on the basis of the uploader's reaction time; a Premium pool figure that contradicts a live Help page nobody has reconciled; a floor that takes a revenue stream away from incumbents who never agreed to it.

And because of what the boundaries are attached to. A creator whose channel is terminated in error loses an audience they were structurally prevented from copying, appeals once, waits without a published deadline, and — on YouTube's own account of November 2025 — may find that being noticed is what triggers a second look. That combination is more than inconclusive.

The split matters more than the band. Anyone reading this page as “YouTube is hostile to creators” has read it wrong: for established long-form partners since 2023, the evidence runs the other way, and so it does for Premium subscribers throughout and for Shorts creators since the revenue share began. The finding concentrates where it always concentrated — on the creators outside the programme, on uploaders inside the copyright system, on anyone whose case is decided by an automated process and then reviewed once, and on EU users who tried to use an external route in 2024 and 2025 and found the door slow to open.

Confidence. High on structure and mechanics, which rest on primary documents. High on the economic figures, moderate on ratios built from them, because the windows and denominators are assumptions and are stated as such. Moderate on the enforcement series: three paired data points, possibly non-comparable definitions, ending in H1 2024. Low on prevalence — there is no creator-side error rate, no ad-load series and no quantification of community complaint anywhere in the public record. Low on intent, where no finding is made at all. Items that could not be re-verified before publication were either omitted or are marked in the text; none of them carries a conclusion here.

Current state · September 2026

Generous at the centre. Unilateral at the edges.

YouTube has built the most generous and most stable creator economics in its market. That sentence is not a hedge attached to a criticism; it is the first finding of this investigation and it survived every attempt to knock it down. If a creator asks which scaled video platform publishes a real revenue share and has kept it, the answer is this one, and it is not close.

It has then surrounded those economics with an enforcement, copyright, distribution and audience architecture in which every unresolved question is resolved by YouTube — disclosed as a rule but rarely as a figure, and corrected most reliably when someone happens to be watching. Not by conspiracy, and not by design in any sense this investigation can evidence. By default: because somebody has to hold the decision, and the party holding it has no counterparty able to make it explain itself.

The 2027 changes make the geometry unusually easy to see. The same announcement grandfathers incumbents into the programme, doubles the bar for everyone arriving next, imposes one floor that incumbents did not agree to, and discloses a percentage that nobody has reconciled with the promise still sitting on the Help page. Generosity and unilateralism in a single document, neither cancelling the other.

That is the whole finding. Not that YouTube pays badly — it does not. Not that the rules are unfair — mostly they are not. Only that the further a creator sits from the centre of the bargain, the more of the platform's uncertainty they are quietly holding, and the fewer of them there are to notice.

The Social Media vertical

Every company here is assessed against the same question. YouTube is the one that answers it best at the centre and alone at the edge.

YouTube is the first build-out of the Social Media vertical to test creator economics as the primary lens. The dimensions it surfaced — whether a standing share is published, whether the follower relationship exports, how many appeals exist and who reviews them, whether an out-of-court body gets a reply — are the ones on which the platforms in this sector actually differ.