Social Media
Facebook never raised the price. It changed the currency.
Every other company in this index can be examined by looking at a bill. Facebook cannot. It has charged ordinary customers approximately nothing for twenty-two years, and it still does — no mandatory fee has ever been introduced, and the investigation found no previously free core feature moved behind a paywall. Not supported What changed is what Facebook asks for instead. In 2006 the consideration was a profile, a friend list and the clicks you made on Facebook. By 2026 it includes behavior observed across other companies' websites and apps, location signals collected regardless of the location setting, conversations held with an AI assistant, and — the part customers notice least — authority over what appears in their own feed. Facebook's delivered value grew enormously across the same period; that is not in dispute here and it is why CVI is 77. The finding is about the other side of the exchange: the non-monetary price rose, repeatedly, and almost never with the customer's active agreement.
CFS = CVI − CHI (77 − 64 = +13), which places Facebook in the index's finely balanced band: value and hostility are close enough that relatively small changes could alter the relationship. No other assessment on this site currently sits in that band. Assessed under CHI/CVI Methodology v2.0: dimensions score the underlying customer harm, patterns describe the behavior that produced it. CHI 64, CVI 77 and CFS +13 are settled analytical values and are not presented as a range. Overall CHI confidence: A−. Overall CVI confidence: B+. Methodology →
The defining finding
Why can a customer choose a chronological feed, and never once be allowed to keep it?
A chronological, friends-first view of Facebook has existed almost continuously since 2009 — as "Most Recent", then as the Feeds tab from July 2022, then alongside the Friends tab from March 2025. In seventeen years it has never been possible to make any of them the view Facebook opens to. Established Each session returns to the ranked feed, and the customer re-asks. Analytical inference
The choice is real. It just doesn't stick. That is a different, narrower and more defensible accusation than "the algorithm is bad" — and it is the one the evidence supports.
The exchange · 2006 vs 2026
Both sides of the deal got bigger. Only one side was ever renegotiated.
The honest way to read this comparison is not "look how much they take now." It is that the right-hand column grew in both directions at once. Facebook in 2026 returns far more than Facebook in 2006 did. It also asks for categories of consideration that did not exist in the original bargain and were never presented to the customer as a price.
You give 3 categories
- Profile informationName, school, interests — typed in deliberately by the customer.
- Your social graphWho you chose to connect to, visible by default only within your own network.
- ParticipationPosts, photos, messages and the clicks you made on Facebook itself.
You get 4 categories
- Friends
- Photos
- Messages
- A social networkA feed composed entirely of the people you had chosen.
You give 9 categories
- Everything from 2006
- AttentionSold at a density that is now optimized per customer and per session.
- Behavioral dataFull telemetry of what you do inside the product, not only what you post.
- Off-platform activityWhat you do on other companies' websites and apps, via Meta's Pixel and mobile SDK.
- Advertising exposureIn feed, in Stories, in Reels, in Marketplace, in search results.
- Algorithmic agencyAuthority over what appears in your own feed, and over whether your choice about it survives the session.
- Location signalsUsed for advertising even where Location History is switched off. Company disclosure
- Inferred characteristicsAttributes computed about you that you never supplied.
- AI-interaction dataConversations with Meta AI feed advertising personalization in the US, effective 16 December 2025. No opt-out is offered. Company disclosure
You get 10+ categories
- Friends
- Photos, at a scale no other archive matches
- MessengerFree global messaging and calls, end-to-end encrypted by default since December 2023.
- GroupsMeta reported more than 1.8 billion people using Groups monthly. Company claim
- MarketplacePeer-to-peer selling with no listing fee, at a scale that displaced paid classifieds.
- Events
- Entertainment and video
- RecommendationsDiscovery of material outside your own network — genuinely useful to many customers.
- Community and crisis infrastructureSafety Check, fundraising tools, blood-donation matching, disaster maps.
- Substantial free functionalityDelivered identically to the poorest and richest customers on earth.
Pillar one · The surveillance price
The perimeter kept moving away from the customer.
This is not the claim that Facebook sells customer data. It does not, in the ordinary commercial sense — advertisers buy access to audiences and Meta performs the matching, and no regulator has found a literal sale. Not supported The finding is narrower and more durable: over twenty years the information Facebook collects moved progressively farther from anything the customer deliberately did on Facebook, and the controls for it consistently arrived years after the collection did.
The pattern underneath is a sequence, not a scandal. Collection ships quietly or on by default; a researcher, journalist or regulator finds it; a partial control appears afterwards. Web tracking began in 2010 and got its viewing tool in 2020. The local-socket channel ran for nine months and stopped only after publication. The one durable exception runs the other way and deserves saying plainly: Meta shut down its face-recognition system in November 2021 and deleted more than a billion facial templates. Company disclosure
Cambridge Analytica is the most famous thing that has ever happened to Facebook, and it is not the centre of this assessment. Treated as a scandal it produces outrage and no analysis. Treated as evidence about architecture it produces something precise.
Between 2010 and April 2015, an app a customer installed could request extensive data about that customer's friends. A researcher's personality quiz obtained roughly 270,000 direct consents and, through those friend permissions, data associated with as many as 87 million people — Facebook's own upper-bound estimate, which the company could not verify precisely because the logging to do so did not exist. Company disclosure Facebook closed the friend-data permissions in 2014–15, before the scandal broke; it learned of the transfer in December 2015 and suspended the firm in March 2018. The sanctions that followed — including a US$5 billion regulatory penalty and a US$725 million class settlement — are the largest in the history of customer data. Settlement · no admission
The 87 million did not fail to read something. 99.7% of them had no interaction with the app at all. Their exposure was decided by other people's choices, under defaults they had never seen.
Consent TransitivityA privacy architecture in which one person's consent or permissions can expose, or determine the treatment of, another person's data without that person's direct consent. Facebook's friend-permissions API is the canonical case; contact-book uploads that build records about non-customers are the same mechanism, still operating. Recorded here as a CHI mechanism identified by this investigation — not yet a standalone Lexicon pattern. CHI mechanism
Pillar two · The agency inversion
Facebook was built around the people you chose. It now runs on what Facebook chooses for you.
Nearly all of the evidence for this section comes from Meta itself — earnings calls, court filings and sworn testimony — which is why it carries the highest confidence on the page. The three measures below are not one statistic. They use different denominators and are shown separately for that reason; combining them into a single line would be more persuasive and less true.
Posts in the Facebook feed recommended by AI from accounts the customer does not follow.
Stated by Mark Zuckerberg on Meta's Q2 2022 and Q1 2024 earnings calls, with the 2022 figure accompanied by a stated intention that it would "more than double". Company disclosure. Meta has not published a Facebook-specific figure since.
Time on Facebook spent with content from friends.
Meta's own figures, advanced in its defense at the FTC antitrust trial in 2025 and reported by multiple outlets from the trial record. Court filing / trial evidence. This investigation did not obtain the underlying docket exhibit; the figure is used as Meta's stated position, which is the only capacity in which it matters here.
Mark Zuckerberg, testifying under oath, April 2025.
In November 2025 the court found for Meta, holding it does not hold monopoly power and observing that its focus "has shifted from connecting users to friends and family to providing algorithm-generated recommendations". The FTC has appealed; the case is not final.
Read the second and third panels together and the strangeness becomes clear. The same transformation that Facebook presented to customers in 2018 as a return to "meaningful social interactions" between friends was presented to a federal court in 2025 as evidence that friends were no longer what Facebook is for. Both statements were made by the same company about the same product. Only one of them was made to the people using it.
New CHI pattern · Adopted
Preference Amnesia
The Facebook investigation produced one finding the existing Lexicon could not describe. It is not that the customer has no choice, and it is not that the default is wrong. It is that the choice is real, is offered, is honoured for the length of one session, and then quietly reverts to the configuration the company prefers — permanently placing the burden of re-asking on the customer.
Preference Amnesia: a product allows customers to change a default or express a preference, but repeatedly resets, forgets, or refuses to persist that choice, requiring the customer to reassert it while restoring the company's preferred configuration.
Additional CHI condition. The forgotten preference must systematically benefit the company or advance the company's preferred behavior. A setting that fails to save in both directions is a bug. A setting that only fails in the direction that costs the company money is a pattern.
Why it is distinct. Algorithmic Replacement describes what the feed became. Choice Illusion describes options that were never real. Default selection describes where a customer starts. Preference Amnesia describes none of those: the option is real, the customer used it successfully, and the product declined to remember. Facebook is the type specimen — from approximately 2009 to 2026, customers could reach chronological and friends-first views but could never make one the persistent default.
“You already fucking asked me.”The plain-English translation, in the register this index was founded in. It is a small cost, paid an unlimited number of times — which is exactly why it irritates customers out of all proportion to its apparent size. Open the full Lexicon entry →
One qualification, kept deliberately. The exact reset behaviour has varied by platform and release across seventeen years, and Meta has never documented it. What is established is the outcome: no persistent chronological or friends-only default has ever been available. Where the evidence is a long trail of consistent customer-facing documentation rather than a single company statement, this page treats the outcome as established and the mechanism as observed, not as disclosed. Analytical inference
Pillar three · The support vacuum
Three billion customers, and for most of them nobody to ask.
This is the section of the Facebook assessment with the sharpest individual consequences. A Facebook account is not a subscription that can be re-created; it holds two decades of photographs, the groups a person belongs to, their marketplace history and, for many small sellers, their livelihood. The sequence below is the finding — not any single step in it.
Meta no longer discloses Facebook-only user figures; the last published numbers were around 3.07 billion monthly. Support at this scale is a genuinely unprecedented problem and this assessment says so.
Reporting documented locked-out customers cycling through automated identity forms with no response: "I sent these forms in morning, noon and night, multiple times a day. Nobody got back to me, not once." Some bought a $299 Oculus headset — a Meta hardware product with a real support team — purely to reach a human who could escalate their Facebook account, then returned it. Customer reports
A subscription at $11.99 per month on the web and $14.99 in-app, bundling a verified badge, impersonation monitoring and account support. Roughly three months earlier Meta had cut about 11,000 roles. Meta's stated rationale for the product is authenticity and impersonation protection, and this investigation found no evidence that support was degraded in order to sell it — no internal document, no admission, nothing that would survive contest.
The letter documented account-takeover complaints rising more than 1,000% in one state's records between 2019 and 2023 (73 complaints to 783, with 128 in January 2024 alone) and quoted customers directly: "I have attempted to contact Facebook but there is no customer support and the methods online are all dead-ends." Regulator letter It was a letter, not an enforcement action, and no public Meta response was located.
From roughly April 2025, large numbers of accounts and thousands of groups were suspended, some flagged for offences the customers had no connection to. Meta acknowledged a technical error affecting groups and did not confirm a cause for the individual bans; attribution to automated moderation remains an allegation. The pattern journalists documented is the relevant part: accounts reinstated only after a reporter asked. Paying Meta Verified subscribers reported the support they had bought made no difference.
The Board found a "lack of transparency and consistency" in Meta's two-track approach to disabling accounts, noting "the difference between what leads to one type of violation or the other isn't clear or well-documented", and recommended Meta disclose the role AI plays in enforcement. On the paid tier it found that Meta charges for Meta Verified with promised "24/7 access to email or chat agent support" while failing to provide users with disabled accounts any sort of "meaningful assistance". Oversight Board finding
Advertising creep · Supported
The ads did not just multiply. They became personal in a second sense.
The News Feed carried no paid advertising for its first five and a half years. Ads entered the feed in January 2012 and then entered every surface Facebook subsequently built: mobile feed, autoplay video, mid-roll, Stories, Marketplace, search results, Reels. No major surface, once monetized, has been permanently de-monetized.
Meta has described the mechanism itself. On the Q4 2025 earnings call the CFO stated that impressions grew 18%, "driven primarily by engagement and user growth and to a lesser degree, ad load optimizations" — and, separately, that "in the second half of 2025, our initiatives on Facebook to redistribute ads across users and sessions delivered a nearly 4x larger revenue impact than Facebook ad load increases." Company disclosure
There is no authoritative long-term series for how many posts in a Facebook feed are advertising, and this page does not state one. A measurement in 2014 put ads at roughly 3% of desktop feed posts; analysts in 2016 estimated around one in ten; widely repeated modern figures could not be traced to a rigorous study. The direction is well evidenced. The number is not, and inventing precision here would be the easiest way to lose the argument.
The disclosure trend is documented separately: from technical measures in 2016 and 2019 that made ads harder for software to distinguish from organic posts, to March 2026, when Meta replaced the "Sponsored" label with a visibly smaller "Ad" tag.
The second sense of "personal" is what makes this more than ordinary monetization. Redistributing ads across customers and sessions means ad load is no longer a property of the product — it is a property of you. Two people opening the same app on the same morning are not necessarily being shown the same quantity of advertising, and neither is told which one of them is the heavier payer. There is nothing unlawful in that, and it is the logical endpoint of performance advertising. It is also the point at which "how much does this cost me?" stops having a published answer.
Promise reversal · One exhibit
The strongest promise-reversal evidence in this index is not a marketing slogan. It is a court order.
Facebook has made many statements about privacy over twenty years and it would be easy — and weak — to assemble a chronology of them. One exhibit is stronger than ten. The one below has the quality the others lack: the earlier representation was not a promise to the public, it was an obligation to a regulator, and the later enforcement action was expressly predicated on its breach.
29 November 2011
The commitment
Facebook settled an eight-count US Federal Trade Commission complaint alleging it had deceived customers about privacy — including that information customers had designated as private, such as their friends list, had been made public. Settlement · no admission of liability.
The resulting consent order bound the company: no misrepresenting privacy, express consent required before overriding customers' privacy preferences, a comprehensive privacy programme, and independent assessments for twenty years.
24 July 2019
The US$5,000,000,000 penalty
The FTC imposed a five-billion-dollar civil penalty — at the time roughly twenty times the largest privacy penalty ever levied anywhere — whose stated predicate was violation of the 2012 order, alongside fresh allegations: settings that omitted that friends' apps could still reach a customer's data; whitelisted apps retaining friend-data access after the announced cut-off; misrepresented facial-recognition defaults; and phone numbers collected for two-factor security used for advertising.
Also a settlement, also without admission. The FTC vote was 3–2, with both dissenters arguing the terms were too lenient. Facebook has not been found liable by a court on these counts.
The qualification matters and is not a technicality: both the 2011 and 2019 actions were resolved by consent, without admission of liability. What makes this exhibit strong is not a finding of guilt. It is that the second action's own stated basis was the breach of commitments made in the first — a documented sequence of promise, obligation, and enforcement, entirely within the regulator's record. Scored under Trust & Transparency 11/15.
Falsification · What we tried to prove and couldn't
Facebook has the strongest counter-file of any company in this index.
Facebook is the most disliked company CHI has assessed, which makes this section the most important one on the page. Five hypotheses that would have made the case dramatically stronger were tested and did not survive. Each is recorded here in the form it failed.
Facebook's case for Facebook
Why CVI is 77.
Assessed independently of the hostility investigation. If this page only convinces you that Facebook extracts, it has failed, because the same twenty years produced one of the largest transfers of free capability to ordinary people in commercial history.
Twenty-two years with no mandatory monetary price, delivered identically to the poorest and the richest customers on earth. No subscription business in this index achieves that, and no priced product can. Fact
Economic research repeatedly finds customers value Facebook far above its price: a real-money randomised experiment published in 2020 found the median participant required roughly $100 to give it up for four weeks. The same study also found wellbeing improved while they were away — both halves are reported here. Peer-reviewed
Groups: Meta has reported more than 1.8 billion people using them monthly, with tens of millions of admins. Support communities, local life and mutual aid at a scale with no real substitute. Company claim
Marketplace: peer-to-peer buying and selling with no listing fee, at a scale that displaced paid classified advertising for millions of households, and a documented reason younger customers keep accounts open at all.
Messenger: free global messaging and calling, end-to-end encrypted by default since December 2023 — shipped against sustained public opposition and at a measurable cost to Meta's own detection statistics. Fact
Civic and crisis infrastructure: Safety Check and Crisis Response since 2014, blood-donation matching with more than 100 million sign-ups reported, disaster maps used by relief agencies, and fundraising tools that moved billions of dollars to charities with no platform fee. Company claim
Accessibility and reliability at a scale few organisations operate: automatic image descriptions for blind customers from 2016, and a service that stays available to billions of people daily.
The Oversight Board: an externally-staffed body Meta funds, whose decisions on individual content bind the company and which overturns it regularly — including, in June 2026, over Meta's own account-ban process. No comparable platform has built one. Counted as genuine innovation in customer protection. Fact
Customers can and do leave: US teen usage fell from 71% to roughly a third between 2014–15 and 2022. Whatever else is true, this is not a captive market. Survey research
Deductions applied, and why CVI is not higher: the account-support failure sits inside Trust, Safety & Reliability (10/15); recommendation-driven low-quality and AI-generated content degrades the core experience; and native charitable and local-news tooling has been wound down, making some delivered value historical rather than current.
CVI 77/100 against CHI 64/100 produces CFS +13 — the finely balanced band. Facebook delivers a great deal. It is the narrowness of the gap, for a product with no price, that this assessment is about. Overall CVI confidence: B+.
Methodology · Resolved by this investigation
Fixed under pressure.
Facebook forced CHI to settle a question every large-company assessment eventually hits: what happens to a company's score when it fixes something only because a regulator, a court, a journalist or a competitor made it? Almost every meaningful improvement in Facebook's data and consent practices is traceable to external compulsion. A framework that ignores this flatters the company; a framework that punishes it stops measuring the customer's actual experience.
This rule is applied consistently across this page and is being carried into other assessments where the same question arises. It is the reason Facebook's CHI is 64 rather than higher: several historical practices scored here were genuinely ended, and the page credits the endings while keeping the record.
Associated patterns · Final state
The patterns materially relevant to this verdict.
Primary
Supported
Qualified and partial
Preference Amnesia is a new Lexicon entry established by this investigation. Consent Transitivity is recorded as a CHI mechanism and does not yet have a Lexicon entry. CHI Lexicon →
Evidence & methodology
- Company disclosureMeta Q4 2025 earnings call transcriptSource of the verbatim "ad load optimizations" and "redistribute ads across users and sessions" statements. Verified directly against the transcript for this page.
- Oversight BoardMeta's Oversight Board says account bans lack due process, transparency — 4 June 2026Reporting of the Board's findings on ban transparency, AI disclosure, and the failure to provide disabled-account holders with any sort of "meaningful assistance" despite Meta Verified's promised support. Wording verified for this page.
- RegulatoryFTC imposes $5 billion penalty and sweeping new privacy restrictions on Facebook — 24 July 2019The promise-reversal exhibit. Settlement without admission of liability; stated predicate is violation of the 2012 order.
- RegulatoryFacebook settles FTC charges that it deceived consumers by failing to keep privacy promises — 29 November 2011The eight-count complaint and consent order that the 2019 penalty was predicated upon.
- Regulator correspondenceMultistate letter to Meta on account takeovers — 41 attorneys general, 5 March 2024Primary source for the complaint-volume figures and the quoted customer statements about the absence of support. A letter, not an enforcement action.
- InvestigativeEach Facebook user is monitored by thousands of companies — Consumer Reports / The Markup, 17 January 2024Source of the ~2,230-companies figure, from 709 volunteers' own downloaded archives. Self-selected sample; treated as an order of magnitude.
- AcademicMeta pauses mobile port tracking tech — 3 June 2025The local-socket web-to-app linkage identified by IMDEA Networks, Radboud University and KU Leuven, and Meta's decision to pause it days after disclosure.
- Court judgmentA$20m penalty for Meta companies over Onavo — ACCC, 26 July 2023Federal Court of Australia penalty for conduct liable to mislead customers about the use of their data.
- Company disclosureAn update on our use of face recognition — Meta, 2 November 2021Shutdown of the Facebook face-recognition system and deletion of more than one billion facial templates. Counted as a genuine reversal.
- Company disclosureLaunching default end-to-end encryption on Messenger — Meta, 6 December 2023The clearest customer-favorable act in the record; shipped against public objection from law enforcement and child-safety organisations.
- Court judgmentMeta wins FTC antitrust trial — 18 November 2025Judgment for Meta, including the finding that its apps "have continuously improved" while remaining free. Under appeal by the FTC; not final.
- Peer-reviewedThe Welfare Effects of Social Media — American Economic Review, 2020Randomised deactivation experiment: large stated valuations of Facebook alongside measured wellbeing improvements while deactivated. Both halves are used on this page.
This page is a synthesis of a completed evidence dossier of approximately 16,500 words covering six parallel research streams, each instructed to attempt falsification as well as confirmation.
Evidentiary status is preserved throughout and is not upgraded for readability. Allegations — including claims contained in unsealed litigation filings — are labeled as allegations and are never presented as findings. Settlements resolved without admission of liability are identified as such every time they appear. Several matters described here remain live as of August 2026, including the FTC's appeal of the November 2025 antitrust judgment, the European Commission's preliminary Digital Services Act findings of October 2025, appeals against European competition and data-protection decisions, and ongoing product-liability litigation in the United States; none of these are described as concluded.
Three claims used prominently on this page were re-verified against primary or near-primary sources during the build: the Q4 2025 ad-load and ad-redistribution statements (verified verbatim against Meta's transcript), the Oversight Board's June 2026 wording on Meta Verified support (corrected during verification to the Board's actual phrase, "meaningful assistance"), and the 17% friend-content figure (corroborated across independent reports of Meta's trial evidence, and attributed here as Meta's own stated position rather than as an independently audited measurement).
Claims the dossier could not sufficiently verify were omitted rather than softened. These include any specific modern figure for advertising density in the Facebook feed, any figure for Meta Verified subscriber adoption, and the total number of customers affected by the 2025 suspension waves. Where Meta is the only source for a figure, it is labeled a company claim. CHI 64/100, CVI 77/100 and CFS +13 are settled analytical values and are not presented as a range.
Final verdict
CONCERNING — THE RELATIONSHIP IS FINELY BALANCED
Your data
The surveillance price
What Facebook knows moved outward from what you typed, to what you clicked, to what you do on other companies' websites and apps, to what you say to an assistant. Information & Privacy: 13/15.
Your attention
The advertising price
Sold from every surface, at a load Meta now optimizes per customer and per session — which means the quantity you personally pay is no longer a published figure. Scored across Revenue Extraction and Behavioral Manipulation.
Your agency
The control price
Authority over your own feed, transferred by default and never durably returnable — the choice exists, and has never been allowed to stick. Customer Restriction: 14/20.
Facebook is not the most extractive company in this index and it is not close. Its monetary price is zero, it has never raised it, and the accusation that usually carries a hostility case here — that the customer is paying more for less — is simply not available. On the site's own scoring model, the twenty-five points reserved for Revenue Extraction produce Facebook's lowest proportional score. The product delivers a very great deal, to more people, for free, than almost anything ever built.
What the evidence establishes is a different kind of increase. The consideration a customer hands over in 2026 — behavior observed off the platform, location signals collected against a setting that says otherwise, conversations with an assistant, and the authority to decide what appears in their own feed — is far larger than the consideration handed over in 2006, and at no point was any of it presented as a price. Facebook's improvements are real and this page credits them in full, including the ones it was forced into. But the pattern underneath is consistent enough to name: value flowed to customers freely, and agency came back only under pressure.
CFS +13 places Facebook in the index's finely balanced band — value and hostility close enough that relatively small changes could tip the relationship either way. That is a remarkable place for a free product to end up. It is also, on the twenty-year trajectory documented above, a fragile one.
Facebook never raised the price.
So why does it cost so much more than it used to?