CHICustomer Hostility Index
← Back to Company Comparisons

Social Media Comparison

Five platforms. One power structure.

Facebook · Instagram · X · Snapchat · TikTok

The five primary mass-social platforms in this comparison do not differ primarily in how much power they hold over users. All five control distribution, enforcement, monetization and product defaults to a degree no individual user can match.

The differences appear somewhere else:

what gets monetized, what remains protected, whether user choices persist — and what happens when the platform gets something wrong.

This comparison does not rank five CHI scores. It compares the mechanisms underneath them.

Scores are shown exactly as each platform’s own CHI work publishes them, as secondary context only. They are not re-scored, not combined and not used to order anything on this page. X is covered by a two-regime Special Investigation that deliberately issues no CHI score. Facebook and Instagram are separate platform observations under one parent company, Meta.

01 — The common architecture

Different products. Similar asymmetry.

Facebookbegan with friends
Instagramwith photos
Xwith public conversation
Snapchatwith private messaging
TikTokwith algorithmic discovery

Facebook began with friends. Instagram with photos. X with public conversation. Snapchat with private messaging. TikTok with algorithmic discovery.

Their products still feel different.

Their power structures increasingly do not.

Across all five platforms, the company can determine:

  • what reaches you;
  • what reaches other people;
  • what is eligible for recommendation;
  • what violates policy;
  • whether an account remains accessible;
  • how advertising enters the product;
  • what creators may monetize;
  • which controls exist;
  • and when those controls change.

That does not make every exercise of those powers hostile.

It does mean the mere existence of algorithmic ranking, moderation or advertising tells us very little about which platform treats customers differently.

The useful comparison begins where the similarities end.

Platform power

  • Distribution
  • Enforcement
  • Monetization
  • Defaults

Held by all five. Common to the category.

User remedy

  • Preference
  • Appeal
  • Support
  • Reversal

Where the five diverge. Sections 04 and 05 set out how.

The industry's shared architecture is platform control. The competitive difference is how that control is exercised, monetized and corrected.

02 — What each platform monetizes

Four commercial models. Not four scorecards.

Facebook and Instagram are shown as one panel because the monetization described here is one Meta parent-company policy, not two independent behaviours.

Meta — Facebook + Instagram

Attention + data + status / recourse

Facebook and Instagram remain overwhelmingly free products funded through advertising and behavioral data.

Meta has also built paid layers around verification, status, impersonation protection, creator tools, privacy/ad treatment in some jurisdictions, and human support.

Meta Verified launched with “access to a real person for common account issues” as a paid benefit against a documented history of weak account-recovery support for ordinary users.

One parent-company finding. Counted once, not once per app.

Meta found a way to monetize recourse.

X

Access + status + visibility + eligibility

X moved further than the rest of the peer group in attaching payment to functions that affect how the platform treats the user.

Paid tiers have been associated with verification, reply prioritization, XPro/TweetDeck, creator eligibility, longer posts and video, premium product access, and functionality that had previously been free.

X monetizes status, access and visibility more directly than the other platforms in this peer set.

TikTok

Attention + commerce + virtual transactions

TikTok is commercially aggressive without becoming primarily subscription-driven.

Its monetization concentrates in advertising, Shop, LIVE gifts and Coins, creator transactions, search commercialization, native commercial content and commerce integrated into recommendation.

The completed CHI investigation did not support a broad finding that TikTok had taken meaningful previously free core functionality and moved it behind payment.

TikTok extracts primarily from what users do once inside the product, rather than from progressively charging them to retain access to the product itself.

Snapchat

Protected social core + monetized perimeter

Snapchat still preserves more of the private social relationship than its peers.

Monetization has moved closer to that protected core through Sponsored Snaps in Chat, Snapchat+, Platinum, paid Memories storage, private status features and historically paid control over My AI positioning.

At the same time, fewer than 3% of monthly users pay anything and the audited research found no case where the core friend relationship itself had been placed behind a paywall.

The product remains structurally different from its peers precisely where its newer monetization is beginning to press.

03 — The algorithm is not one thing

Five recommendation models. Five different relationships to the social graph.

“Algorithm” is not treated here as one generic hostility mechanism. The question is what the recommender did to the relationship the product was originally built on.

Facebook

Algorithmic replacement

Friend/social-graph proposition progressively displaced by ranking and recommendation.

Instagram

Strongest replacement case

Meta disclosed that more than half of the content people see on Instagram is AI-recommended. Friend-content time also fell materially. The Following feed exists but cannot remain the persistent launch default.

Instagram did not merely add recommendations. Recommendations progressively changed what Instagram was.

TikTok

Algorithmic primacy

Recommendation was foundational rather than a later replacement of a friend-first proposition.

X

Algorithmic intervention

The key issue is discretionary control over reach, ranking, links, replies and account distribution.

Much of that power predates Musk.

Snapchat

Algorithmic separation

Chat and friends' Stories remain more separated from Spotlight and Discover. The product still opens to a camera rather than a feed.

A recommendation engine does not have to consume the private social relationship in order for a mass-social platform to function.

04 — When the platform gets it wrong

The power is common. The remedy is not.

Seven remedy dimensions, five platforms. States are deliberately coarse and deliberately not coloured as a scoreboard. Not established means the reviewed CHI material did not establish it — never that it does not exist.

Documented established in the CHI material Partial exists with a material limit Limited thin, automated or indirect Paid made a paid benefit Not established not located in the reviewed material

Scroll sideways to see all five platforms →

Facebook, Instagram, X, Snapchat and TikTok compared across seven remedy dimensions: appeal, human review, published reversal data, human support, paid support, external redress, and timing transparency.
Parent companyMeta · one parent company
Remedy Facebook Instagram X Snapchat TikTok
Appeal exists DocumentedAccount Status appeal routes; the Oversight Board above them. DocumentedAppeals exist; the 2025 wrongful-ban wave met AI-driven appeal loops with no human contact. DocumentedPublic appeals form and written enforcement ladder (Feb 2023). No entitlement to reasons or a timeline. DocumentedComplaint and account-lock appeal routes, reported under the EU DSA. PartialRouted through a single in-app funnel. Reporting/appeal flows subject to a preliminary EU DSA finding (Oct 2025, contested).
Human review LimitedHuman review at Oversight Board level; ordinary recovery ran through automated forms. LimitedAI appeal loops reported through the 2025 ban wave. Not establishedNot established in the reviewed investigation. DocumentedHuman review of Community Guidelines account-lock appeals (EU, H1 2026). Not establishedNot located in the reviewed assessment.
Published reversal data PartialOversight Board decisions are published and overturn Meta regularly. No platform-wide reversal rate located. Not establishedNo reversal rate located. No root cause ever published for the 2025 flags. PartialEU-only, DSA-compelled: 10.8% (Apr–Sep 2024); 25.6% suspensions, 28.7% restricted reach (Oct 2025). DocumentedEU H1 2026: 43% content complaints, 6.2% account complaints reversed. Not establishedNot located in the reviewed assessment.
Human support LimitedLocked-out customers cycled through automated identity forms; video-selfie check added from late 2024. LimitedNo human channel for free customers 2010–2022. Support hub with AI assistant (Dec 2025); Korea service centre (Feb 2026). Still not promised. Not establishedInsufficient comparable evidence in the reviewed investigation. Not establishedInsufficient comparable evidence in the reviewed assessment. LimitedNo phone line, no email, no web form. One in-app “Report a problem” funnel.
Paid support Meta · one shared policy · counted oncePaidMeta Verified (Feb 2023) sold “access to a real person for common account issues”. The Oversight Board later found Meta charged for promised “24/7 access to email or chat agent support” while disabled-account holders received no “meaningful assistance”. Not establishedPaid tiers exist; paid human support not established. Not establishedPaid tiers exist; paid human support not established. Not establishedNot located in the reviewed assessment.
External redress DocumentedOversight Board: externally staffed, Meta-funded, binding on individual content decisions. Not establishedThe Oversight Board is documented in the Facebook assessment; its Instagram coverage is not set out in the Instagram assessment. PartialEU and UK statutory redress (DSA, Online Safety Act). Elsewhere: Texas forum, class waiver, one- or two-year limits. Not establishedEU complaint handling is documented (above); external redress was not assessed in the reviewed assessment. LimitedRemedy diffused: Coin and subscription refunds to Apple or Google; Shop disputes to the seller first.
Timing / SLA transparency Not establishedNo published ordinary-user timeline located. Not establishedNo published ordinary-user timeline located. LimitedAppeals carry no entitlement to a timeline. Terms revisions pre-posted 25–37 days ahead. DocumentedMedian resolution time published: 5.1 days (EU, H1 2026). Not establishedNo general phone/email/web channel and no published ordinary-user SLA located.
43%Reversal on decided content-removal / visibility complaints
6.2%Reversal on account complaints

Snapchat · EU DSA disclosure, H1 2026Snap disclosed a 43% reversal rate on decided content-removal/visibility complaints, a 6.2% reversal rate on account complaints, resolution times, and human review of Community Guidelines account-lock appeals.

This is not evidence that Snap’s first-line moderation is superior. The customer-aligned finding is that Snap publishes the error/reversal information.

Company disclosure

How to read this matrix. It records whether a remedy is established in the completed CHI work, not how well it performs. X’s EU reversal figures and Snap’s EU figures are both regulatory disclosures and are EU-scoped; neither is a global number. Meta’s paid-support finding is a single parent-company policy and is shown once, across both Meta columns.

The platforms differ less in whether they can make consequential unilateral decisions than in whether the customer can understand, challenge and reverse those decisions.

05 — Where user choice stops

A control is not necessarily control.

Preference Amnesia, as a lifecycle.

  1. User choosesA setting that appears to return agency.
  2. Control appliesIt works — for a session, a period, or while the signal is repeated.
  3. Control expires / resets / retrainsBy app open, by timer, or by model decay.
  4. Platform default returnsThe company’s preferred configuration is back.

↺ The customer re-asks. The cycle repeats.

InstagramFollowing

Chronological followed-accounts feed. Resets to algorithmic Home on every app open; cannot be set as the default.

Company documentation
InstagramRecommendation snooze

The only control over interleaved suggestions — capped at 30 days by design.

Company documentation
Instagram · FacebookRecommendation reset

Real controls that sit outside the default and do not bind it. Facebook’s friends-first views have never been the view the app opens to.

CHI finding
TikTokNot Interested

Cuts content sharply while applied; one independent audit found feeds “dominated by such content again” once users stop.

Independent measurement
TikTokTopic weighting

Manage Topics settings “won’t eliminate any topics entirely, but can influence how often they’re recommended.”

Company disclosure

Evidence boundary. Facebook is the Lexicon’s type specimen for Preference Amnesia and Instagram the corroborating case. The TikTok examples show decay and reweighting — the “retrains” step above — but TikTok’s own assessment records Preference Amnesia as not established under the Lexicon’s strict test, because decay in a live ranking model is not the same as a stored preference being reset, and TikTok never promised persistence. They are shown here as the same lifecycle, not as the same finding.

Social platforms increasingly offer settings that appear to return agency to the user.

But many do not permanently control the core default.

Recommendation reset can mean start the machine over, not turn the machine off.

The platform remembers what helps it personalize. It is less reliable at remembering when the customer asks it not to.

Platforms often permit local preference correction without surrendering control of the default.

06 — The outliers

Eight findings that are not shared across the peer set.

Each is an outlier within this reviewed five-platform corpus only. The coloured rule marks direction, not rank: adverse to the customer, customer-aligned, or structural.

Adverse outlierCustomer-aligned outlierStructural outlier

X

Paid organic visibility

Outlier
X is the only platform in this reviewed peer set with clear evidence that ordinary subscription tiers explicitly purchase preferential organic reply visibility.
Why it matters
This is not paid verification. Labelled ads sell placement; here payment entered the ranking of organic replies — a “reply boost” graduated by price tier from Oct 2023, still described on X’s Help Center.
Evidence confidence
Company documentation High for reply prioritisation. A paid For You boost in 2025–26 is unresolved.

X

Forced monetization

Outlier
X shows the strongest documented conversion of previously free functionality into paid access.
Why it matters
The pattern includes sale of status, SMS two-factor authentication behind payment, paid eligibility to earn, read limits tied to payment status, and formerly free product functionality such as XPro/TweetDeck moving into paid tiers.
Evidence confidence
Company terms & disclosure High.

X

Customer-to-adversary conversion

Outlier
X has a documented adversarial pattern toward commercial counterparties and adjacent ecosystem participants that is not established at comparable depth across the other four in the reviewed corpus.
Why it matters
Advertisers whose remedy was to stop buying became defendants in a federal antitrust suit. This does not claim peers never litigate.
Evidence confidence
Court record High for advertisers; unresolved or absent for subscribers, creators and developers.

Meta · parent company

Monetized human support

Outlier
Meta is the clearest case of turning access to a human for ordinary account problems into an explicit paid product benefit.
Why it matters
One parent-company finding. Facebook and Instagram are not counted separately here.
Evidence confidence
Company disclosure Oversight Board Launch terms are Meta’s own; the Board’s finding on disabled accounts followed.

Snapchat

Private-graph preservation

Outlier
Snapchat is the clearest structural counter-model to recommendation swallowing the private social relationship.
Why it matters
The recommender is built beside the friend graph, not over it: no public likes, comments or follower counts on friend content, and an app that opens to a camera.
Evidence confidence
Company documentation Established. Snap has tested unifying friends’ Stories into its recommender.

Snapchat

Default deletion

Outlier
Snapchat is the clear peer-set outlier in making deletion rather than indefinite retention the default for ordinary private friend content.
Why it matters
Snaps and Chats are still deleted by default from Snap’s servers. Retention is the user’s opt-in, not the platform’s default.
Evidence confidence
Company documentation Established.

Snapchat

Moderation reversal transparency

Outlier
Snap publishes unusually granular moderation reversal and appeal data.
Why it matters
The publication of the error rate is the customer-aligned behavior — not the error rate itself.
Evidence confidence
Company disclosure EU DSA report, H1 2026.

TikTok

Free-access monetization model

Outlier
TikTok is the clearest peer-set case of aggressive monetization without a broad pattern of converting meaningful formerly free core functionality into paid access.
Why it matters
Paywalling of core functionality was tested and not established, while LIVE economics and commerce in discovery sit at the ceiling of its findings.
Evidence confidence
CHI finding Bounded negative on paywalling; strongly supported on commerce.

07 — Industry norm vs genuine outlier

Common does not mean customer-friendly.

Industry norm

  • Algorithmic ranking
  • Broad account enforcement
  • Advertising
  • Behavioral data use
  • Weak ordinary-user support
  • Creator eligibility discretion

Not established as industry norm

  • Paid organic visibility X
  • Systematic free-to-paid functionality conversion X
  • Default deletion of private friend content Snapchat
  • Detailed moderation reversal disclosure Snapchat
  • Monetized human account support Meta
  • Customer-to-adversary conversion X

Common does not mean customer-friendly. It means non-unique.

Peer prevalence changes how CHI interprets uniqueness and relative severity; it does not erase customer harm.

08 — Full five-platform peer matrix

Fourteen dimensions. Five platforms. No numbers.

A reference matrix, drawn only from the completed CHI work on each platform. Each cell gives the mechanism, how it is implemented, and the evidence status. There are no severity values and no winners. Where the reviewed material does not establish something, the cell says Not established — never “none”. Shared Meta policies span both Meta columns and count once.

Company disclosure the company’s own documents, terms or statements Adjudicated a court, regulator decision or binding review body Preliminary regulator preliminary, contested, not final Litigation allegation filed or unsealed, not adjudicated Independent measurement audit, study or survey Reported journalism or customer accounts CHI inference CHI’s reasoning about those facts

Scroll sideways — the dimension column stays fixed →

Facebook, Instagram, X, Snapchat and TikTok compared across fourteen dimensions. Qualitative only; no scores.
Parent companyMeta · one parent company
Dimension FacebookAlgorithmic replacement InstagramStrongest replacement case XAlgorithmic intervention SnapchatAlgorithmic separation TikTokAlgorithmic primacy
1. Account Enforcement Automated enforcementWrongful-suspension reporting; accounts reinstated only after a reporter asked. Attribution to automated moderation remains an allegation.Reported Mass wrongful bansJun–Aug 2025 wave, including false child-exploitation flags; no root cause published.Reported Termination discretion“Any or no reason” (2016), reworded “no reason at our convenience” (2023). Reinstatements by owner poll, no written criteria located.Company terms Automated at scaleEU H1 2026: 85.6m moderation measures, 95.7% automated, 81.4% self-reported accuracy.Company disclosure Not establishedInsufficient comparable evidence in the assessment. Branded-content auto-enforcement from Aug 2025, with For You ineligibility as sanction.Company documentation
2. Appeals / Remedies Appeal + review bodyOversight Board found in June 2026 that account bans lack due process and transparency.Oversight Board Automated appeal loopsAI-driven loops, 2025. Preliminary EU finding on “dark patterns” in reporting and appeals (Oct 2025), contested.Preliminary regulator Remedy AsymmetryForm-based appeals, no right to reasons or timeline. Four terms revisions narrowing user remedies (2023–26); EU/UK statutory redress.Company terms Published reversal43% of decided content complaints and 6.2% of account complaints reversed (EU, H1 2026).Company disclosure In-app onlyOne “Report a problem” funnel. Same Oct 2025 preliminary EU finding on reporting/appeal flows, contested.Preliminary regulator
3. Customer Support Automated recoveryIdentity forms with no response; some bought a $299 Oculus headset to reach hardware support. Video-selfie check from late 2024.Reported Support vacuumNo human channel for free customers 2010–2022. Dec 2025: “support hasn’t always met expectations.” Still not promised.Company statement Not establishedInsufficient comparable evidence in the reviewed investigation. Not establishedInsufficient comparable evidence in the reviewed assessment. Single funnelNo phone, email or web form; hacked-account guidance loops back to the same path.Company documentation
4. Contractual Discretion Silent consideration changeTerms published throughout, but no customer was shown a moment at which the consideration changed.CHI inference Terms retained2012 revolt forced retraction of the ad-likeness clause; data-sharing and arbitration provisions stayed.Company terms Contractual DiscretionVisibility clause word-for-word since 2016. Texas forum with X’s asymmetric right to sue; class waiver; creator payouts changeable “for any reason”.Company terms Not establishedInsufficient comparable evidence in the reviewed assessment. Currency discretionLIVE Coin conversion reserved to TikTok’s “absolute and sole discretion”.Company terms
5. Status, Verification & Visibility Meta · shared policy · counted oncePaid verification + recourseMeta Verified (Feb 2023): badge, impersonation monitoring and support at $11.99 web / $14.99 in-app. Adds; does not confiscate.Company disclosure Pay-for-VisibilityStatus put on sale; “reply boost” graduated by tier. EU found the purchasable checkmark “deceives users” (€120m, Dec 2025).Adjudicated Private status, soldSolar System rank and Snapscore multiplier via Snapchat+. No public likes or follower counts on friend content.Company documentation Transactional statusLIVE gifts, with Diamonds framed as recognition of “popularity”. Paid verification not established.Company documentation
6. Paid Access & Feature Gating Paywall Creep: not supportedNo mandatory charge in twenty-two years; no previously free core feature moved behind payment.CHI inference Price unchanged$0. Ad-free subscription only in the EU, under enforcement (€5.99–€7.99).Company disclosure Forced MonetizationStatus, SMS 2FA, eligibility to earn and read limits became tied to payment. Formerly free XPro/TweetDeck moved into paid tiers, while feature gating also extended to post length and Articles.Company terms Monetized perimeterSnapchat+, Platinum, Memories storage capped at 5GB free, My AI removal once gated to Snapchat+. Fewer than 3% pay.Company documentation Not establishedPaywalling of core functionality tested and not established.CHI inference
7. Advertising / Ad Load Personalized ad loadAds redistributed “across users and sessions” — a larger revenue impact than ad-load increases (Q4 2025).Company disclosure Advertising CreepAds across Feed, Stories, Explore, Reels, search and profiles, with ad load managed per customer.Company disclosure Advertiser eligibilityFrom Apr 2023, buy a checkmark or spend $1,000 a month to run ads. Twitter-era promoted products were labelled.Company disclosure Ads in ChatSponsored Snaps from Oct 2024; “nearly 75% of U.S. Chat DAU viewed ads in Chat” (Q1 2026). Under-18s are excluded in the EU, Norway, Switzerland and Israel; no global exclusion is established in the reviewed material.Company disclosure Commercial densificationAds, Shop in the feed (2023), sponsored search (2023–26). All paid placements marked “Sponsored” or “Ad”.Company disclosure
8. Creator Monetization Not establishedInsufficient comparable evidence in the reviewed assessment. Program volatilityNo-listing-fee storefronts and branded-content tools, offset by the 2022–23 commerce and bonus whipsaw.CHI inference Pay to be eligible to earnRevenue sharing required a subscription (Jul 2023); payouts on paying-user engagement under an undisclosed formula (Nov 2024).Company terms Not establishedInsufficient comparable evidence in the reviewed assessment. Virtual-currency economyCoins → Gifts → Diamonds; roughly 50% of face value reaches the creator.CHI inference
9. Algorithmic Opacity Undocumented resetsReset behaviour has varied across seventeen years and Meta has never documented it.CHI inference Recommendation shareMore than half of content seen is AI-recommended; friends ~7% of viewing time (court record).Company disclosure Partly open codeRanking code published (2023, 2026); live values and some files not disclosed. The 4.0 / 2.0 multipliers are code defaults, not measured reach.Company disclosure Converging rankers“Unifying the ranking models between Spotlight and Stories to a single backend stack” (Q1 2024).Company disclosure Retention objectiveInternal “Algo 101” (published by the New York Times, 2021): the “ultimate goal is to add daily active users.” Nine controls, none for commerce.Reported · company document
10. Reach / Distribution Control Algorithmic ReplacementFriends-first views exist but have never been the view Facebook opens to.CHI finding Default never cededFollowing and Favorites cannot be set as default; every app open returns to algorithmic Home.Company documentation Discretionary reachLink deprioritisation confirmed by owner post (Nov 2024); rival links blocked or delayed (2022–23). Restricted-reach labels since Apr 2023.Company statement Graph kept separateSpotlight on its own tab; Chat and friends’ Stories outside it; app opens to a camera.Company documentation Recommender as homeFor You is the first feed; Following secondary from day one.Company documentation
11. Hidden Degradation Discovery-driven controlsCollection ships quietly or on by default; a partial control follows discovery (web tracking 2010, viewing tool 2020).CHI finding Cumulative, not concealedThe transfer was announced in fragments; no single step was hidden.CHI finding Present under both regimesTwitter’s 2FA-data misuse ($150m, 2022) is the largest single hidden harm in the record; X improved disclosure of rule-based limits.Regulatory order Not establishedInsufficient comparable evidence in the reviewed assessment. Safeguard scored on coverage60-minute teen limit: daily use ~108.5 → ~107 minutes; success defined by media coverage.Unsealed filing · disputed
12. Data / Privacy Controls Consent TransitivityAverage participant’s data sent to Meta by ~2,230 companies (self-selected sample).Independent measurement Consent through termsDPC fines €405m (2022) and €390m incl. €180m for Instagram (2023); EU €200m over pay-or-consent (2025). All under appeal. E2EE removed from DMs, 2026.Adjudicated · under appeal Uncompensated licenceAI-training licence “with no compensation” (Nov 2024 terms). Grok EU processing suspended in Irish proceedings (Aug 2024).Company terms Deletion by defaultSnaps and Chats deleted from servers by default. My AI content and location used to personalize ads.Company documentation Non-refusable targetingBehavioural ad targeting non-refusable since 2021. Non-personalised feed toggle EU and Kazakhstan only.Company documentation
13. Moderation Transparency Published review bodyOversight Board decisions bind individual content and regularly overturn Meta.Oversight Board Contested disclosurePreliminary EU findings on reporting/appeal flows and researcher data access (Oct 2025), contested.Preliminary regulator Disclosure without constraintRestricted-reach labels, EU reversal rates, “Under the Hood” label pilot (Aug 2026). Trust & Safety Council dissolved (Dec 2022).Company disclosure Granular disclosureMeasures, automation share, self-reported accuracy, reversal rates and resolution times (EU, H1 2026).Company disclosure Not establishedInsufficient comparable evidence in the assessment. Same Oct 2025 preliminary EU finding, contested.Preliminary regulator
14. Youth / Age Controls Not establishedInsufficient comparable evidence in the Facebook assessment. Teen AccountsPrivate by default, messaging limits, sleep mode, supervision (Sep 2024). Preliminary EU finding on under-13s (Apr 2026), contested.Preliminary regulator Not establishedNot assessed in the reviewed investigation. Youth carve-outsNo Sponsored Snaps in Chat for under-18s in the EU, Norway, Switzerland and Israel; Family Center switch for My AI. ~415,000 accounts locked under Australia’s under-16 law.Company documentation Default limit, disputed effect60-minute daily limit on by default for under-18s (Mar 2023); LIVE gifting age-gated at 18.Company disclosure

How to read this matrix. The mechanism line names what the platform does; it is not a rating. A longer or more specific cell reflects deeper underlying research, not greater hostility — see Method & limits below. Scores do not appear in the matrix, and no row or column total exists.

09 — Method & limits

What this comparison can and cannot carry.

Research depth

The source investigations are not equally deep.

X has a substantially larger underlying research package than several peers.

More findings therefore do not automatically mean greater hostility.

This comparison normalizes by mechanism rather than raw finding count.

Meta double-counting

Facebook and Instagram remain separate platform observations because their product architecture and customer relationships differ.

Shared Meta parent-company policies do not count as two independent behaviors in aggregate interpretation.

Bounded negatives

Absence of evidence is not evidence of absence.

Language used on this page where appropriate:

  • not established
  • not located
  • insufficient comparable evidence

Evidence status

Distinctions are preserved among company disclosure, adjudicated finding, preliminary regulator finding, litigation allegation, independent measurement and CHI inference.

They are not flattened.

Company disclosure Adjudicated finding Preliminary regulator finding Litigation allegation Independent measurement CHI inference

Sources: every statement on this page is drawn from CHI’s completed work on the five platforms — the Facebook, Instagram, Snapchat and TikTok assessments, the X / Twitter Special Investigation (research cutoff 2 September 2026) and, for X product history, the Elon Musk Special Investigation — and from the CHI Lexicon entries for Preference Amnesia and Algorithmic Replacement. Primary sources are cited on those pages. No new research was conducted for this comparison, and no platform is re-scored. LinkedIn, Reddit and YouTube hold their own CHI work and are outside this formal peer comparison.

Final finding

The platforms are more alike in power than they are in restraint.

No single platform dominates every hostility mechanism.

The five share a common architecture of company control over distribution, enforcement, monetization and defaults.

Their differences become visible in what happens next.

Meta combines attention and data extraction with increasingly monetized status and recourse.

X converts more access, visibility and eligibility into paid products.

TikTok extracts primarily through attention, commerce and transactions while preserving broad free access.

Snapchat keeps more of the private social relationship outside the recommendation engine while increasingly monetizing its perimeter.

And across the category, the deepest common weakness is not that platforms possess extraordinary power.

It is that users often have much less power to correct the platform than the platform has to act on the user.

Social media's common architecture is platform control. Its real competitive difference is what gets monetized, what stays protected, and whether the customer gets a remedy.

CHI AssessmentFacebook → CHI AssessmentInstagram → Special InvestigationX / Twitter: What Actually Changed → CHI AssessmentSnapchat → CHI AssessmentTikTok → CHI LexiconPreference Amnesia → CHI LexiconAlgorithmic Replacement → All comparisonsCompany Comparisons → How CHI worksMethodology →