Social Media
Snapchat
Snapchat kept the friend graph its competitors abandoned — and then discovered that a private friend graph is something you can sell back to the people inside it.
Every other social assessment in this index describes replacement: a user-directed social graph progressively displaced by an algorithmic feed. Snapchat is the platform where that did not happen, and reading it as another Instagram is the fastest way to get this page wrong. Snaps and Chats are still deleted by default from Snap's servers. There are still no public like counts, no public comments and no follower counts on friend content, and no visible friend list. The app still opens into a camera rather than a feed, and under sworn testimony the setting governing who may contact a user is friends and phone contacts only "and can't be expanded." Established What changed is the commercial layer built on top of that architecture, almost entirely since 2022: advertising placed into the private inbox Snap promised in 2014 it would never touch, a nine-year uncapped personal archive capped and the overage sold, a subscription assembled substantially out of relief from frictions the free product engineers, and a private economy of interpersonal attention — who rewatched you, when they looked, where you rank in their affections — sold to one side of a two-sided relationship. The friend graph is not a legacy surface at Snap. It is the asset being borrowed against.
CFS = CVI − CHI (76 − 58 = +18), which places Snapchat in the value-exceeds-hostility band — alone among the six social platforms in this index, every other one of which sits in the finely balanced band. It carries the lowest CHI in the Social Media category, below Reddit (62), Facebook (64), LinkedIn (67), Instagram (70) and TikTok (74), and a CVI of 76 that is mid-table rather than exceptional. That combination is the assessment: Snapchat is not more valuable than its peers, it is materially less extractive, and the gap is closing from the extraction side. Scores were calculated from the evidence under v2.0 and then sanity-checked against Facebook and TikTok, not copied from them. Overall CHI confidence: B — high on the documentary findings, deliberately lower on three present-tense questions that require a live account and are named rather than guessed at. Overall CVI confidence: B+. Methodology →
The defining contradiction
"We won't put advertisements in your personal communication – things like Snaps or Chats. That would be totally rude."
Snap published that on 16 October 2014, in the first person, on the day it started selling advertising. It has never been retracted and is still on Snap's newsroom. In Q1 2026 Snap told investors that "nearly 75% of U.S. Chat DAU viewed ads in Chat." Snap for Business sells the format to advertisers as "native-style ads" that "appear just like organic Snaps," placed in "Snapchat's most popular tab - the Chat." Fact · Snap's own documents, both ends
CHI does not need to prove deception here, and does not assert it. A dated first-person commitment set against the company's own investor disclosure is a finding on its own terms.
The central distinction · Why Snapchat is not Facebook, Instagram, X or TikTok
Two ways to monetise a social graph. Only one of them takes it away from you.
The thesis that governs the rest of this index — that platforms replaced user-directed social graphs with algorithmically supplied attention — is the correct frame for five of the six companies in this category. Applied to Snapchat it produces the wrong answer, because Snap took a different route to a comparable commercial destination, and the difference is the whole assessment.
- The friend or follow graph stays nominally in placeNobody is unfriended. The graph simply stops determining what appears.
- A recommendation engine takes the surface the graph used to holdMeta's own published figure is that more than half of what people see on Instagram is AI-recommended. Competitor claim · not researched here
- Public metrics supply the performance incentiveFollowers, likes, comments and view counts make the user an audience-facing performer.
- Revenue comes from selling attention to advertisersThe relationship layer is not the product. Reach is.
- The chronological alternative exists as a tab that does not stickPresent on most comparators; reverts to the ranked default.
- The friend graph stays, and stays load-bearingMutual, named, private, with no audience attached. Q1 2026: Snap calls its communications service "our strongest long-term advantage."
- The recommendation product is built beside it, not over itSpotlight passed "more than 40% of total time spent watching content" in Q2 2025 — on its own tab. Friends' Stories were merged into a unified recommender in a 2024 test and the test was withdrawn.
- There is no public performance layer to incentiviseNo follower counts, no public likes or comments on friend content, no visible friend list. That absence has held for fourteen years.
- Revenue increasingly comes from inside the relationshipReciprocity measured and sold; the maintenance ritual priced; the archive capped; the private inbox turned into inventory.
- The chronological alternative for friends' Stories was withdrawn in 2017 and never documented againChat is still recency-ordered. The Stories list is not. Present tense unverified
Snapchat kept your friends and started charging admission to the relationship. This is not a softer finding. It is a different one, and in one respect a harder one: a platform that replaces the graph is selling your attention to advertisers, which everybody understands. A platform that keeps the graph and meters it is selling one friend information about another — a transaction in which the person being measured is not the buyer, is not told, receives nothing, and in Snap's own control documentation has no control at all.
Finding one · Private-Channel Advertising — supported, strongest documentary item in the record
The one surface protected by a written promise is now the one Snap calls its largest reach product.
Finding two · Jurisdictional Two-Tier Protection — the most portable mechanism in this assessment
Snap can keep advertiser messages out of a child's private inbox. Its own product proves the capability across the EU and three additional countries.
Two separate protections run on two different regulatory perimeters. They are frequently reported as one thing; they are not, and merging them misstates both. Both are taken from Snap's own live ads-privacy page, fetched 2 September 2026.
Snap's words: "If you are under 18 in these countries, Sponsored snaps will only appear outside of your Chat feed." This is a per-cohort exclusion of advertising from a private messaging surface — operating in production, at scale, today.
For those 13–17 accounts Snap does not use inferences about "gender cohort, interests or interests of other people associated with you," does not use activity collected "from your activity on the websites and platforms of our advertisers and partners," and does "not include you in custom audiences."
Finding three · Promissory Product Monetization — supported, with Snap's own attribution closing the gap
Nine years of "save this, it's backed up." Then a cap, then a price.
Finding four · Reciprocity Metering — the mechanism that earns Snapchat its own entry in this index
Snap refused to build a public scoreboard. Then it built a private one and sold the seat.
Snapchat has no follower count, no public like count, no public comments on friend Stories and no visible friend list, and it removed its public Best Friends display in January 2015 and never restored it. That restraint is real, durable, and credited in full further down this page. It is also what makes the metrics saleable: a ranking nobody can see is a ranking somebody will pay to see. Every row below is from Snap's own support documentation.
Fewer than 3% of monthly active users subscribe to anything, so the buyer population is small. The measured population is everyone they are friends with. Strength: Medium-High
Finding five · Paid Remedy — a subscription assembled from relief
A ladder of paid cures for frictions the free product engineers.
Snapchat+ launched on 29 June 2022 at US$3.99/month, in the same quarter Snap told investors that "our revenue growth has substantially slowed," and was described at launch by a Snap product SVP as not expected to be a "material new revenue source." Other Revenue reached $316M in Q2 2026, up 85% year over year.
Finding six · Snap's own object placed above the user's chosen contacts
For at least two years, the only documented way to move the chatbot off the top of your friend list was to pay.
The rebuttal is strong on the chatbot and weak on the placement, for which no user-benefit rationale has ever been offered by anyone at Snap. Snap's public answer to complaints was that if users didn't like it "they don't have to use it." In an application premised on deletion, the one conversation partner pinned above the user's friends is also the one whose content is retained: "All the content shared with My AI is stored until you delete it." Strength: High historically; Low in the present tense
Finding seven · Four Snap documents, three answers
A subscriber paying four times the base price cannot determine from Snap's own pages what they bought.
The question is narrow and it is the one that matters: does the top-priced tier remove advertising from the Chat inbox — the intrusion a subscriber is likeliest to be paying to escape? All four documents below were fetched and verified on 2 September 2026.
Finding eight · Algorithmic ordering of friends' Stories — established historically, unresolved in the present
Taken on 29 November 2017. Partially given back in May 2018. The half users protested is the half Snap kept.
The loss is not that the ranking is malicious. It is that the option was withdrawn and never restored, through the largest user protest the company has faced — and that Snap's answer to a rejected design has never once been to ship the old behaviour as a choice. Strength: High historically; Medium overall pending a live check
How the model works · The spine of the assessment
One asset, six ways to bill it.
Snapchat's commercial layer is not a set of unrelated premium features. It is a coherent conversion of a single asset — a mutual, named, non-public friend graph that no competitor still has — into revenue, along four directions, with two layers laid over the top.
What changed · Six eras, fifteen years
The proposition was not repudiated. It was narrowed to the first screen.
What survives of the 2012 proposition. The camera still opens first. Snaps and Chats are still deleted by default. There are still no public likes, no public comments on friend Stories, no public friend lists and no follower counts on friend accounts. What has gone is the second half of the sentence. "With friends, not strangers" now describes one tab out of five. "Identity tied to now, today" is contradicted by a trillion archived Memories, priced from September 2025. "We won't put advertisements in your personal communication" was true for just under ten years. "We count on editors and artists, not clicks and shares" was true for thirty-four months.
The control surface · Forty-four preferences tested
Where the preference concerns your own data, the control is real, free and durable. Everywhere else, it is absent, priced, or held by somebody else.
This audit asks one narrow question of each stated preference: does a control exist, does it survive, does Snap substitute its own configuration for it, and does exercising it require payment. Where no control exists the cell says so — that is a finding, not a gap. Of the forty-four rows tested, fifteen record no control at all, seven are directly priced, five are priced in the opposite direction — the money buys one person visibility into another — and two exist only inside a regulated perimeter.
Cross-platform comparison · Six differentiating dimensions
Snapchat is more customer-aligned than the cohort on the personal layer, and less so on the commercial one.
Snapchat cells are established from the evidence on this page. Every competitor cell is marked [K—V] — knowledge, verify. This investigation researched Snap only, and no competitor claim below was sourced within it; they are carried as orientation, not as findings, and the individual company assessments in this index control.
| Dimension | Snapchat | X | TikTok | |||
|---|---|---|---|---|---|---|
| Where the recommender sits relative to the friend graph | Beside it. Chat and friends' Stories stay separate from Spotlight and Discover; the 2024 merge was tested and withdrawn. Spotlight is over 40% of content time. | Ranked feed replaced the friend timeline; friend posts compete with recommendations inside one feed [K—V] | Ranked since 2016; Meta's own figure is that more than half of what people see is AI-recommended — the corpus's type specimen [K—V] | "For You" is the default; the follow graph is a secondary tab [K—V] | Ranked default over a connection graph; recommendation comparatively weak [K—V] | Feed-native; the follow graph was never primary [K—V] |
| Public status metrics for ordinary users | None. No follower count, no public likes or comments on personal Stories, private friend lists. The status economy is sold privately instead. | Public friend counts, reactions, comments [K—V] | Public followers and comments; hide-likes option [K—V] | Public followers, likes, reposts, views [K—V] | Public connections, followers, reactions [K—V] | Public followers, likes, comments, views [K—V] |
| Advertising inside the one-to-one message surface | Yes — Sponsored Snaps since October 2024, designed to "appear just like organic Snaps," reaching nearly 75% of US Chat DAU, excluded for under-18s only in the EU, Norway, Switzerland and Israel. | Messenger inbox ads exist, in a separate app, not formatted as messages from a person [K—V] | DM ad tests reported; not established as shipped [K—V] | Promoted DMs historically; not a current default surface [K—V] | Sponsored InMail is message-shaped, but LinkedIn is not a private-friend messenger [K—V] | None in DMs [K—V] |
| Paid removal of platform-created friction, and paid status | Both, systematically. Streak Restore, the My AI unpin, ad removal at the top tier only, Memories storage — plus Solar System, rewatch, timestamp and peek status signals sold to one side of a friendship. | Ad-free subscription in the EU only; no paid status-signal product [K—V] | Ad-free subscription in the EU only [K—V] | Premium cuts ads and adds analytics and reach-boost — the nearest analogue [K—V] | Premium sells profile-view data and InMail [K—V] | No ad-free tier; analytics free [K—V] |
| Reach and persistence of a non-personalised option | EEA, UK and Switzerland only, despite an August 2023 promise of global rollout. Persists where it exists — it removes profiling, not recommendation. | Chronological tab exists but reverts [K—V] | "Following" tab exists but reverts [K—V] | "Following" tab exists; persistence inconsistent [K—V] | "Recent" sort exists but reverts [K—V] | Following tab exists; not sticky [K—V] |
| Default persistence of user content | Deleted by default after viewing or 24 hours — unique in this cohort. Narrowed by Memories, saved chats, My AI and metadata; on reporting and Snap's own 2019 conference material, end-to-end encryption is understood to cover one-to-one image and video Snaps only [T2/T3 — no Snap primary security document obtained]. | Persistent; Messenger E2EE default [K—V] | Persistent; Vanish mode optional [K—V] | Persistent [K—V] | Persistent [K—V] | Persistent [K—V] |
Snapchat's case for Snapchat · Why CHI is not higher
The counter-evidence is architectural, dated, commercially costly, and mostly sworn or contractual.
This section is why the verdict is MIXED rather than hostile, and why the CHI is the lowest in the Social Media category. If this page leaves you thinking Snapchat is simply an extraction machine, it has failed — several of these decisions cost Snap real money and it has held them for a decade.
Deletion by default, still the architecture. The privacy policy effective 21 September 2026: Snaps and Chats "will be deleted by default from our servers within 24 hours after we detect they've been opened by all recipients or have expired." Across nearly two trillion Snaps a year, this destroys the exact substrate every competitor treats as its core asset. No commercial rationale is discoverable for it. Fact
A limit Snap removed from its own discretion. Under sworn testimony to the Senate Judiciary Committee on 31 January 2024: default "Contact Me" settings are "friends and phone contacts only for all accounts, and can't be expanded." A messenger that forbids stranger contact forgoes the network-expansion loop that drives DAU on every follow-graph platform. Sworn
No public comparison economy on the personal layer, for fourteen years. No follower counts, "no public likes or comments" on friend Stories, private friend lists. Public metrics are cheap, self-reinforcing and directly monetisable, and Snap declined to build them on friend content — then deleted the one it had shipped, in January 2015, and has not restored it in eleven years.
The app opens into a camera, not a feed. Sworn, and consistent across nine years of shipped designs including the abandoned three-tab test. The first thing the product asks for is production, not consumption. It is a statement about the front door rather than about the whole house — Spotlight is one swipe away — but no comparator makes it at all.
Snap Map shipped with the protective default and the opt-out in the first release, on 21 June 2017, and Snap then publicly and correctly rebutted a wave of inaccurate police and press warnings that it broadcast children's locations. This page does not use the 2017 panic against Snap, because the panic was false. Fact
The Simple Snapchat retreat, April 2025. Snap proposed merging friends' Stories into "our first-ever unified recommendation system," ran it as a staged test on roughly 25 million of some 450 million daily users, published results unfavourable on iOS, and killed it because "our most engaged Snapchatters consistently demonstrated a preference for a five-tab layout." The design raised content consumption — the metric Snap reports as its growth engine — and Snap withdrew it anyway. This fails as an adverse finding and is recorded here as a user win.
A teen-safety stack that is dated, escalating and produces numbers Snap publishes. Family Center (August 2022), stranger-contact warnings producing "more than 12 million blocks" by mid-2024, mutual-friend thresholds, device-level blocking, a strike system, and screen-time insights in January 2026. Snap is among the five companies producing over 75% of all NCMEC CyberTips globally. Adoption is small and is stated rather than omitted: Spiegel's 2024 Senate answers report roughly 200,000 parents and 400,000 linked teens against "more than 20 million teenagers" in the US — one to two percent. The timing is consistently downstream of hearings and regulators; the delivery is credited here and the reactivity is scored above.
Snap publishes its own error rate. Its EU DSA report for H1 2026 discloses a 43% reversal rate on decided content-removal complaints, a 6.2% reversal rate on account complaints, median resolution of 5.1 days, and human review of every Guidelines account-lock appeal. No large platform compared here publishes more. The same numbers indict the architecture — 95.7% of 85.6 million measures automated, one appeal only — and both readings are true.
Fewer than 3% of monthly users pay anything. Camera, Chat, friends' Stories, Snap Map and Memories under 5GB are free and no core functionality has ever been paywalled. The base subscription has been $3.99 since 29 June 2022 and no price increase is established. Costly compliance is also on the record: Snap absorbed roughly 415,000 locked accounts under Australia's under-16 law after saying it "strongly disagree[d]," walked away from a Perplexity arrangement worth up to $400M of contracted revenue rather than force a rollout, released Android dark mode free after seven weeks of complaint, and endorsed KOSA first among major platforms.
Deductions applied, and why CVI is 76 rather than higher. The only independent longitudinal satisfaction measure does not say users love it: ACSI has Snapchat at 71 on debut in 2019, 68 and last place in its category in 2024, and 72 in 2026 against a 75 industry average — seven years, never above mid-table, below YouTube, TikTok, Instagram, LinkedIn and X, and level with Facebook and Reddit. Trust, Safety & Reliability sits at 9/15 because a self-declared birthdate remains the only universal age gate, there is no human support channel outside a single asynchronous appeal, and the European Commission opened formal Article 28 child-protection proceedings on 26 March 2026 (undecided, no fine).
CVI 76/100 against CHI 58/100 produces CFS +18 — value currently exceeds hostility, the only social platform in this index in that band. Overall CVI confidence: B+.
Score breakdown
Why CHI is 58 — the lowest in this category — and why it is not lower.
58 needs a reason that is not "Snapchat feels nicer." The reason is structural and it is legible in the dimension split: Snapchat scores materially lower than its peers on Information & Privacy and Behavioral Manipulation, because its defaults dispose rather than accumulate and it has never built the public comparison economy that drives the category — while still carrying substantial Revenue Extraction and Trust & Transparency penalties, because it monetises a relationship layer nobody else has and because its own dated words contradict its own current conduct unusually cleanly.
Calibration. Scores were computed from the evidence under Methodology v2.0 and then sanity-checked against the existing Facebook (64 / 77 / +13) and TikTok (74 / 81 / +7) assessments, not derived from them. Snapchat produces the widest divergence between hostility and value of any social platform in the corpus, which is what the evidence predicts: the personal layer is the strongest in the category and the commercial layer is the most inventive. CHI 58, CVI 76 and CFS +18 are settled values, not a range. The provisional research band was "Concerning, at the lower end," and the computed score lands there.
Associated patterns · Final state
The patterns materially relevant to this verdict.
Primary
Supported
Qualified, partial, and not supported
Falsification · What the evidence does NOT support
This investigation corrected its own starting assumptions more than it confirmed them.
Twelve of the brief's original premises were disproved outright by the evidence and are recorded in a binding do-not-publish list rather than softened. Publishing the failures is the price of being believed about the findings above — and several of the loudest claims about Snapchat are not merely unproven but a distortion of something duller and more interesting.
Location sharing has been off by default, friends-only, with Ghost Mode, since the 21 June 2017 launch, and Snap's live support page still says so. In June 2017 police forces and press told parents the opposite; Snap corrected them, accurately. This page does not use the 2017 panic against Snap, because the panic was false. One narrow question remains open: whether the first-open selector pre-selects "My Friends," which requires a live account to settle.
Contradicted by Snap's own disclosed deletion defaults — and, tellingly, by the opposite allegation in state litigation, which is that Snap retains too little. A dossier cannot plead both. The accurate finding is narrower and is scored above: deletion is real and its exceptions are specific — Memories, saved chats, Stories, My AI conversations and communications metadata, for which Snap's current law-enforcement guide states no retention period at all.
The clinical literature will not carry it. The only peer-reviewed study located (n=2,483) found correlations with problematic smartphone use and FOMO that the authors themselves called "weak." The internal documents that would support a stronger claim are unauthenticated allegations quoted at third hand from an attorney-general press release, which Snap calls "patently false" and "cherry-picked." The mechanic and its monetisation are documentary and are scored. The clinical claim is not, and does not appear in this page's own voice.
Snap did fall about 7.2% on 22 February 2018. But Citi had downgraded Snap from Neutral to Sell two days earlier, citing "the poor reception among users of Snapchat's recent major app redesign," and contemporaneous reporting notes analysts had already been tracking redesign-linked engagement declines. Causation is not established and is not asserted here. The documented driver is the redesign — which is a finding on this page in its own right.
Snap ran it as a bounded test, published metrics unfavourable on iOS, and abandoned the three-tab design in April 2025 because its most engaged users preferred the old layout. A company single-mindedly maximising content time does not kill a design that raised content time. Published as hostility this misdescribes the record. It survives only as evidence of intent — Snap did propose it, and did name support for "our advertising business over the long-term" among its aims — and because Spotlight's placement beside the camera was retained.
The FTC charged six counts of deception in May 2014 and the Commission approved a final order 5–0 on 31 December 2014, imposing a privacy programme under independent monitoring for twenty years — to approximately 2034 — with no monetary penalty and no admission or denial of liability by Snap. It is of record and it is cited on this page as such. It is not an adjudicated finding of wrongdoing and must not be described as one.
Every internal-document quotation in the public record reaches this investigation through an attorney-general press release, not through the complaint text; no investigator successfully parsed the underlying pleadings. Snap's position, on the record, is that the allegations are "patently false" and that the states "repeatedly mischaracterize" its documents and "cherry-picked" employee-concern material. The New Mexico court's April 2025 denial of Snap's motion to dismiss establishes justiciability only. Snap has never had a liability verdict entered against it on minors' issues, and no finding on this page rests on any of this material. Allegation · New Mexico and Utah · unresolved
There was no $9.99 Platinum test in 2024 — no such price exists at any date; the US launch was a reported $15.99 in February 2025. The Memories cap was September 2025, not 2024. Streak Restore's free and paid versions launched together on 1 March 2023, not free in July 2023 and paid in 2024. No Snap policy sets biometric retention at "up to a year" — the April 2025 policy never uses the word "biometric," and the stated periods for face, hand and voice data are three years from last use, two in Colorado, and within 90 days of an annual review finding storage unnecessary. Snap's arbitration clause contains no mass-arbitration batching protocol in either the April 2025 or September 2026 Terms. And the widely attributed Spiegel line "Facebook is not a place for people to hang out" appears in no located source and is not attributed to him here.
Limitations · The three unresolved questions that change something
Nobody on this investigation holds a current Snapchat account, and the page says so wherever it matters.
Almost nothing important about Snap's documentary record is missing. Almost everything about its current in-app behaviour is. Every claim about what the app does today rests on Snap's own documentation, its regulatory filings, its investor disclosures or third-party reporting. The evidence-gap appendix runs to five categories; only the three that would change a finding are published here.
Two structural limits are load-bearing and are stated rather than worked around. Web archives were unreachable throughout, so no historical version of any Snap page could be dated directly and all version-dating rests on press reports and Snap's own effective-date stamps — which is why several dates on this page are given as periods rather than days. And no research stream successfully parsed the underlying New Mexico or Utah pleadings, so every internal-document quotation in the public record reaches this assessment at third hand through an attorney-general press release. That material is quarantined, labelled, and no finding rests on it.
Evidence & methodology
- Company documentation — the promiseAdvertising on Snapchat — Snap Newsroom, 16 October 2014Source of "We won't put advertisements in your personal communication – things like Snaps or Chats. That would be totally rude." Volunteered in the first person at the moment of monetisation, never retracted, re-verified 2 September 2026.
- Company documentation — the practiceSponsored Snaps — Snap for BusinessThe advertiser-facing design premise: "native-style ads" that "appear just like organic Snaps," placed in "Snapchat's most popular tab - the Chat," with no push notification. The resemblance is the selling proposition, in Snap's own words.
- Company documentation — the single best exhibitSnap and Ads / Ads & Your Privacy — values.snap.comBoth jurisdictional perimeters, on one live page: "If you are under 18 in these countries, Sponsored snaps will only appear outside of your Chat feed" (EU, Norway, Switzerland, Israel); and the separate teen exemption from interest inference, activity-based ads and custom audiences (EEA, UK, Norway, Switzerland). Fetched 2 September 2026. Snap states the differential and offers no rationale.
- Company announcementIntroducing Memories Storage Plans — Snap Newsroom, 26 September 2025The 5GB cap, the 100GB / 250GB / 5TB tiers, "12 months of temporary Memories storage" above the limit, "more than 1 trillion" saved Memories, and free export. Snap publishes no price on this or any other page; the reported US figures are secondary-sourced and attributed as such throughout.
- Company announcementIntroducing Memories — Snap Newsroom, 6 July 2016The invitation: "a new way to save Snaps and Stories on Snapchat… a personal collection," opt-in, user-initiated, free and uncapped. The nine-year baseline against which the 2025 cap is measured.
- Company announcementIntroducing the New Snapchat — Snap Newsroom, 29 November 2017"The new Friends page to the left of the camera displays your friends based on the way you communicate with them." The disclosed ranking signal — the user's own communication behaviour — is also the strongest rebuttal to the finding it supports.
- Company announcementNew features and transparency measures for Snapchatters in the EU — Snap Newsroom, August 2023The dated global commitment: the controls "will initially be available to Snapchatters in the EU before rolling out to our global community over the coming months." Three years later Snap's own support pages confine the personalisation opt-out to the EEA, the UK and Switzerland.
- Company announcementKeep The 🔥 Going — Snap Newsroom, 1 March 2023Streak Restore: "restore one Streak for free with just one tap," with additional restores purchasable in the same release. Establishes that free and paid restore launched together, correcting a widely repeated two-stage chronology.
- Company documentationWhat is Snapchat+? — Snap Support article 7121577610900"The Platinum Monthly Plan for Snapchat+ lets you enjoy Snapchat+ with no Sponsored Snaps, and no Story or Lens ads!" One of the four Snap documents that give three answers to the same question. Fetched 2 September 2026.
- Company documentationSnapchat Platinum — Snap Support article 18520949571988"No Story or Lens ads"; "You may still see sponsored places and My AI responses." The phrase "Sponsored Snaps" does not appear. The companion Chat-ads article (31709629111060) agrees with this one and not with the page above.
- Company documentationFriend Solar System — Snap Support article 8098318922516The subscriber sees "which planet you are in their Solar System." The page confirms it is "off by default for first time subscribers" — a genuine default reversal that must never be written as "Snap turned it off for everyone," because whether existing subscribers were switched off is not established.
- Company documentationPrivacy Policy, effective 21 September 2026 — values.snap.comBoth halves of the privacy finding in one document: "Snaps and Chats sent in Snapchat will be deleted by default from our servers within 24 hours"; and "if there's a dog in the photo… inform us that you're into dogs so we can surface you fun dog videos and dog food ads." Alongside the affirmative carve-out: "We do not use the private content and communications you send to people who are your friends to personalize your experience, make recommendations, or show you ads."
- Company documentationPrivacy by Product: Spotlight — values.snap.comSnap's own statement of the ranking objective, unusually candid: "if you can't stop watching dance challenges, we will show you more dance-related content."
- Company documentationTerms of Service, effective 21 September 2026 — snap.com§12 "Your content in Memories might become unavailable for any number of reasons, including… a decision on our end to terminate your account"; §16 termination "for any other reason"; §19 the $100 liability ceiling; §20 individual arbitration with a 30-day opt-out. Tested against the April 2025 text and containing no mass-arbitration batching protocol, disproving a premise this investigation carried in.
- SEC filingSnap Inc. Form 10-K, FY2025 — filed 5 February 2026Advertising approximately 87% of FY2025 revenue; DAU as "a critical measure of our user engagement"; Snapchat+ offering "special insights into their friendships"; Platinum as "an ad-free experience"; and the risk factor naming the trade-off directly — harm may follow if "we do not provide a compelling user experience because of the decisions we make regarding the type and frequency of advertisements that we display or the structure and design of our products." The "shown in chronological order" sentence was read in full against accession 0001564408-26-000013 and modifies the photos within a Story, not the Stories list.
- SEC filingSnap Inc. Form 10-K, FY2018 — filed 6 February 2019Snap's own filed causal statement on the redesign: DAUs "declined primarily due to changes in the design of our application and continued performance issues with the Android version." A company attributing a year of user loss to a design change users had asked, about 1.25 million times, to have withdrawn.
- Investor disclosureSnap Inc. Q1 2026 Investor Letter — 6 May 2026"Nearly 75% of U.S. Chat DAU viewed ads in Chat"; "Memories Storage was an important driver of this acceleration… a larger than anticipated share of new subscribers acquired through Memories are choosing higher ARPU subscription offerings"; and the description of Snap's communications service as "our strongest long-term advantage."
- Investor disclosureSnap Inc. Q1 2025 Investor Letter — 29 April 2025The Simple Snapchat retreat, carried as counter-evidence: "Our most engaged Snapchatters consistently demonstrated a preference for a five-tab layout" — and, in the same paragraph, the element that was retained: Spotlight "now placed directly to the right of the Camera."
- Investor disclosureSnap Inc. Q4 2024 Investor Letter — 4 February 2025Sponsored Snaps delivering "more than 50 million impressions on average in the US to Snapchatters 18 or older, making it our largest single day reach product." The phrase Snap uses for the surface it promised in 2014 to leave alone is "incremental advertising inventory."
- Investor disclosureSnap Inc. Q2 2026 Prepared Remarks — 3 August 2026493M DAU, 971M MAU, Other Revenue $316M (+85% YoY), and the counterweight that constrains this entire assessment: "Less than 3% of our monthly active users are paying subscribers."
- Sworn testimonyWritten testimony of Evan Spiegel — US Senate Judiciary Committee, 31 January 2024Snap's fullest sworn account of its own defaults, and the source of three counter-evidence findings: "We designed Snapchat to open into the camera, instead of a content feed"; "default 'Contact Me' settings are set to friends and phone contacts only for all accounts, and can't be expanded"; "there are no public likes or comments." Made in the same twelve months Snap shipped a unified recommendation test over friends' Stories and advertising into the private inbox.
- Regulatory submissionSnap Inc. response: Protecting children from harms online — Ofcom, 26 July 2024Snap's written cost-benefit case against per-service age assurance, filed with a regulator in its own name: per-check cost and "significant user drop-off," plus the assertion that "our recommender systems are not a material risk factor for this type of harm." The public analogue of material that elsewhere reaches this assessment only as unadjudicated allegation.
- RegulatoryTech firms commit to stronger anti-grooming measures — Ofcom, 21 May 2026Snap "agreed to make significant and long-overdue changes by adopting all the recommended grooming prevention measures," with UK age checks over summer 2026. Carried in both directions: the commitment is real, and Ofcom's own word for it is "long-overdue."
- Regulatory — open, undecidedCommission opens formal DSA proceedings on Snapchat and child protection — European Commission, 26 March 2026Opened with the Dutch Digital Services Coordinator; names under-13 access, default settings, moderation and reporting, and "dark pattern" design including disappearing messages, streaks and the discovery tab — the first regulatory proceeding anywhere to name the streak mechanic. The opening of proceedings does not prejudge the outcome: there are no preliminary findings, no decision and no fine as of 31 August 2026.
- Regulatory — resolved without admissionSnapchat settles FTC charges that promises of disappearing messages were false — FTC, 8 May 2014Six counts of deception; the final order approved 5–0 on 31 December 2014, imposing a privacy programme under independent monitoring for twenty years — to approximately 2034 — with no monetary penalty and no admission or denial of liability. Cited here as of record, and never as an adjudicated finding.
- Company transparency reportingSnap Transparency and DSA reporting — values.snap.comThe disclosure that makes the enforcement finding possible and also rebuts it. EU DSA H1 2026: 85,588,403 moderation measures, 95.7% automated, 81.4% self-reported accuracy, and 43% of decided content-removal complaints reversed, with human review of every Guidelines account-lock appeal. Global H2 2025: 471,892 appeals, 31,951 reinstatements.
- Independent measurementConstant Companion: A Week in the Life of a Young Person's Smartphone Use — Common Sense Media, 26 September 2023The only independent passive-sensing measurement in the corpus: a median 19.6 Snapchat notifications a day to teenagers, the highest of any app. The same report's youth advisers defend Snapchat on the point advocacy groups attack — "it's a person every time" — and both halves are carried here.
- Satisfaction indexAmerican Customer Satisfaction Index — social mediaThe independent measure that constrains the CVI. Snapchat debuted at 71 in 2019, fell to 68 and last place in its category in 2024, and stands at 72 against a 75 industry average in 2026 — below YouTube, TikTok, Instagram, LinkedIn and X, and level with Facebook and Reddit. The 2023 figure appears as both 70 and 71 in ACSI's own sources and 2025 is absent; no unbroken series is published here.
- Independent researchTeens, Social Media and AI Chatbots 2025 — Pew Research Center, 9 December 202555% of US teens use Snapchat, 46% daily, 12% "almost constantly," shares little changed since 2022. The population to which the jurisdictional advertising finding applies.
This page is a synthesis of a frozen evidence package: twelve parallel research streams of approximately 102,000 words, followed by three adversarial passes over them — one written as Snap's general counsel, one as plaintiffs' counsel, and one fact-and-consistency audit that resolved cross-stream contradictions and produced a binding do-not-publish list. Roughly 326 distinct sources are registered, of which 153 are documents Snap itself authored. Evidence is current to 2 September 2026.
The do-not-publish process is the reason to trust the rest. Twelve of the investigation's own starting premises were disproved by the evidence and struck rather than softened, including a Platinum price that exists at no date, a Memories cap dated a year early, a two-stage Streak Restore chronology that never happened, a biometric retention period that appears in no Snap policy, an arbitration clause that does not exist in either current Terms version, and a widely circulated Spiegel quotation that no source supports. Four further claims that circulate constantly — that Snap Map broadcasts location by default, that a celebrity tweet erased $1.3 billion, that Snapchat keeps everything, and that Memories will be deleted on 21 September 2026 — are demonstrably false and are published as false.
Evidentiary status is preserved and never upgraded for readability. Every internal-document quotation in the public record reaches this assessment at third hand through an attorney-general press release; it is quarantined, labelled ALLEGATION with the alleging party named, carried with Snap's rebuttal — "patently false," "repeatedly mischaracterizes," "cherry-picked" — and no finding on this page rests on any of it. The New Mexico court's denial of Snap's motion to dismiss establishes justiciability only. Three JCCP and MDL bellwether matters settled confidentially before verdict; no amount is published, no settlement motive is inferred, and a settlement without admission is not described as a concession. The 2014 FTC matter was resolved by consent order without admission of liability. The European Commission's Article 28 proceedings are open and undecided.
Reported prices are attributed rather than asserted, because Snap publishes no price for Platinum or for Memories storage on any of its own pages — the $3.99 base subscription, announced by Snap itself in June 2022, is the exception. The Q4 2025 subscriber figure is published only with its definitional caveat, never bare. Deletion behaviour at the end of the Memories grace period is stated as unestablished rather than guessed. The clinical characterisation of Snapstreaks is omitted rather than hedged, because the peer-reviewed literature will not carry it. And three present-tense questions — the chronological Stories view, the current My AI unpin gate, and Platinum's treatment of inbox advertising — are named as unresolved rather than resolved against Snap. CHI 58/100, CVI 76/100 and CFS +18 are settled values under Methodology v2.0. They are not a range.
Final verdict
MIXED — THE FRIEND GRAPH SURVIVED. IT IS NOW THE PRODUCT LINE.
Snapchat is, on the evidence, the most structurally customer-aligned product of the major social platforms this index has examined — and it is being monetised in ways none of them have attempted. Both halves of that sentence are documentary, and neither cancels the other.
The aligned half is not a public-relations artefact. Snap's server-side default is to delete friend-to-friend messages rather than retain them, across nearly two trillion Snaps a year, destroying the exact substrate every competitor treats as its core asset. There are no public like counts, no comments and no follower counts on friend content, and no visible friend list; the public Best Friends display was withdrawn in January 2015 and has never come back. The application opens to a camera. Under oath, the setting governing who may contact a user is friends and phone contacts only "and can't be expanded" — a constraint no follow-graph platform could adopt without abandoning its growth model. Snap Map has been off by default since launch, and the 2017 panic that it broadcast users' locations was misinformation Snap correctly rebutted. When Snap tested merging friends' Stories into a recommendation engine, its own instrumentation said its heaviest users hated it, and it stopped. Fewer than three percent of monthly users pay anything.
The unaligned half is equally documentary, and it is concentrated almost entirely in the last four years. A dated first-person promise never to advertise in personal communication, set against "nearly 75% of U.S. Chat DAU viewed ads in Chat." A nine-year uncapped archive capped, with the CEO telling investors the cap was "a big driver of the subscriber growth." A chatbot placed above every human conversation, whose removal was purchasable for at least two years. Advertising excluded from minors' private inboxes in the EU, Norway, Switzerland and Israel, and nowhere else, with no stated reason. A private economy of interpersonal attention — rewatches, timestamps, peeks, rank inside a friend's affections — sold to one side of a relationship, with the other side holding no notice and no control. A daily reciprocal obligation whose failure has a price. And the ordering of friends' Stories, taken in 2017 through the largest user protest in the company's history, and not documented as returning since.
Two things fix the verdict, because in both cases Snap's own conduct narrows the available defences. The first is jurisdictional: Snap's product proves that excluding advertising from a child's private inbox is technically and operationally feasible at cohort level, and Snap applies it only inside the documented protected perimeters. Whether the carve-out preceded or followed regulatory pressure could not be established, because web archives were unreachable; what is documentary is the coverage, and the absence of any stated rationale. The second is that Snap keeps telling investors why. This is not an assessment built on inference about motive. Snap called the inbox "incremental advertising inventory." It attributed subscriber growth to the Memories cap. It described My AI's value as "the intense signal we get." The product changes hardest to explain as user improvements become immediately legible as inventory creation — and Snap explains them that way itself, usually in the paragraph above the announcement.
So the finding is narrow and specific, and it is not the one most people expect. Not that an algorithm replaced your friends — Snap built that product, measured it, and withdrew it, which no other platform in this corpus has done. Not that Snapchat is addictive — the literature will not carry it and this page does not assert it. The finding is that a company which kept the relationship layer everyone else abandoned discovered that the relationship layer was the most monetisable thing it owned, and has spent four years metering it, pricing the frictions inside it, selling the measurements to one party, and placing advertising into the room it once promised to leave alone. The user has not been deprived of the product they came for. They are being progressively charged for the parts of it that were free, watched by other users who paid, and advertised to in the room they were told was private.
Nobody took your friends away.
So why is there a price on knowing them?