What we found
The FTC says social media platforms don't always thoroughly vet the ads you see or the advertisers behind them, so an ad that looks real in your feed may not be. According to an FTC data spotlight cited in the alert, in 2025 people reported losses of more than $95 million to scams that started when they ordered something after seeing an ad on social media. The FTC adds that because many people don't report scams, real losses are likely much higher. The alert describes the pattern as anyone being able to pay to place ads in your feed, including scammers impersonating real brands, selling knock-off products, or touting a program to teach you how to invest and beat the stock market. It also notes advertisers can target you with personalized ads based on your online activity, such as your interests and shopping habits. Before buying from an ad or post, search the company's name plus "scam" or "complaint" as the FTC recommends. If you spot or lose money to a fake or scammy ad, the FTC asks you to report it at ReportFraud.ftc.gov. We could not establish from this material how any individual platform reviews ads, so treat any specific ad as unchecked until you look into the seller yourself.
- The only evidence is a single official FTC consumer alert, with no independent corroboration in the harvested set.
- The loss figure is self-described as based on consumer reports and is stated by the source itself to understate real losses.
- The advisory describes a general pattern rather than a specific campaign, seller or ad that could be examined.
- Reviewed by 3 models, 2 from independent houses.
THE RULE
Shoppers who buy from an eye-catching social media ad without checking the seller risk paying for goods that never arrive or that are knock-offs, with reported losses in this category exceeding $95 million in 2025.
What we don’t know
- Which platforms or advertisers the reported losses involved.
- How many reports the $95 million figure represents, and the typical loss per person.
- What vetting, if any, individual platforms actually apply to ads and advertisers.
- Which real brands scammers were impersonating in these ads.
- Whether any of the money reported lost was recovered.
The bench — who voted
3 INDEPENDENT AI MODELS REVIEWED THIS. ALL 3: MODERATE RISK.
The card names a count. Here are the seats behind it, with what each one said.
ANTHROPICclaude-opus-5 · MODERATE
Several statements the finding attributes to the FTC alert are absent from the single evidence row: that many people don't report so real losses are likely much higher; that anyone can pay to place ads including impersonators, knock-off sellers and stock-market-beating course promoters; that advertisers target readers using interests and shopping habits; the advice to search a company name plus 'scam' or 'complaint'; and the ReportFraud.ftc.gov referral. These may well be in the underlying alert, but on the material supplied they are unverified additions presented as sourced quotation of the FTC.
DISSENT, PRESERVEDThe headline CLAIM itself — that unvetted ads and advertisers let scam ads reach ordinary feeds — is fully carried by the evidence row and should not be softened. My objection is to the unsourced elaboration around it, not to the core proposition.
DISSENT, PRESERVEDThe finding deserves credit for stating plainly that it could not establish how any individual platform reviews ads, and for listing that as an unknown rather than implying blanket platform negligence. That restraint is the right instinct and should survive editing.
DISSENT, PRESERVEDSelf-reported loss totals are a floor, not a measure of prevalence. Whichever way the panel rules, published copy should not let a dollar aggregate stand in for how likely a given reader is to be hit.
GROQopenai/gpt-oss-120b · MODERATE
Finding adds details about who can place ads, brand impersonation, knock‑off products and personalized targeting that are not present in the cited FTC alert.
OPENROUTERopenrouter/free · MODERATE
Raised no objection; read the evidence as holds.
Reviewed by 3 independent models: 1 found it carried by the evidence, 2 did not.
▼ Protocol & challenge record
Objection (high): Monocausal mechanism in the claim line is not evidenced. The claim asserts scam ads appear in feeds *because* platforms do not thoroughly vet ads or advertisers. The only support is the FTC's own bare assertion ("That's because social media platforms don't always thoroughly vet..."), offered with no data, audit, no named platform, and no description of any actual review process. The draft itself concedes in its final sentence that it "could not establish from this material how any individual platform reviews ads" — which directly undercuts the causal mechanism stated in the claim. A claim cannot rest on a mechanism the finding admits is unverifiable. The claim should be restated as attributed characterisation ("The FTC says...") rather than as a stated cause in the outlet's own voice.
Not resolved — preserved on the record.
Objection (medium): Alternative explanations are not addressed. Scam ads can reach feeds even where vetting exists, because scammers evade it (cloaking, hijacked ad accounts, compromised business pages, reseller/affiliate chains, ads placed via third-party agencies, organic posts and influencer content that are not ads at all). The FTC text does not rule any of this out; 'not always thoroughly vetted' is compatible with 'vetted but circumvented.' The finding presents absent-vetting as the explanation rather than one of several.
Not resolved — preserved on the record.
Objection (medium): The $95 million figure is second-hand and unverified. The harvested alert cites 'an FTC data spotlight' but the spotlight itself was not harvested (the link is stripped from the row text). So the number, its date range, its scope (whether it is complaint-report totals from Consumer Sentinel, whether it is US-only, whether 2025 is a complete calendar year, whether it is limited to undelivered/knock-off merchandise), and the report count are all unchecked. The draft's limitation line mentions 'consumer reports' but does not flag that the underlying dataset was never inspected.
Resolved: Partly resolved: the evidence limitation already restricts the figure to FTC consumer reports and to ordering-after-an-ad scams, and 'unknowns' asks for report count and typical loss. Remaining gap is an explicit statement that the cited data spotlight was not itself examined.
Objection (medium): No date anchoring or currency signal in the finding. The alert is dated 2026-08-10 and reports 2025 data, so the figure is up to roughly a year and a half old at publication, and the reader is never told when the FTC said this. A one-source advisory with a year-old statistic should carry its date on the face of the finding.
Resolved: Resolvable by wording: add the alert's date to the sentence that introduces the FTC ('In an August 2026 consumer alert...') so the reader can judge currency.
Objection (medium): Internal contradiction on audience targeting. 'targeting_dropped' removes the 'shopper' callout on the ground that no harvested source describes that group — yet the risk_line names exactly that group ('Shoppers who buy from an eye-catching social media ad...') and the share_directive names 'anyone who buys from ads that appear in their social media feed.' Either the source supports naming ad-clicking shoppers (it does: the opening scenario is a shopper buying a discounted designer bag) or it does not, in which case the risk_line and share_directive must be rewritten. As it stands the draft applies the rule in one field and breaks it in two others.
Not resolved — preserved on the record.
Objection (low): No denominator or trend, so magnitude is uninterpretable. $95 million is presented as the scale of harm with no comparison to total reported fraud losses, no prior-year figure, and no per-person loss. Without that, 'exceeding $95 million' in the risk_line functions as an alarm signal rather than a measurement, and the reader cannot tell whether this category is large, small, rising or falling.
Resolved: Partly acknowledged via 'unknowns' (report count, per-person loss). A baseline comparison remains absent and would require a second source not in the harvested set; noting the absence in the finding is the minimum fix.
Objection (low): Scope creep in the personalised-advertising sentence. Ad targeting based on interests and shopping habits is in the source but is not evidence for the claim about vetting, and placing it alongside the scam material invites the inference that personalised targeting is part of the scam mechanism. The FTC text does not make that link.
Not resolved — preserved on the record.
Objection (low): The closing sentence carries no row_ids yet issues an instruction ('treat any specific ad as unchecked until you look into the seller yourself'). An epistemic caveat about what could not be established is legitimate; converting the absence of evidence into a behavioural directive is a step beyond the record.
Not resolved — preserved on the record.
Objection (low): Confidence reason 1 says there is 'no independent corroboration in the harvested set,' which is accurate but understates the structural problem: the claim is about platform conduct, and the sole source is a consumer-education page from a regulator that names no platform and cites no evidence about platform conduct. Entity match for the claim's real subject (the platforms) is zero.
Resolved: Partly resolved: confidence is set to 'moderate' and the single-source dependence is disclosed in confidence_reasons.
Preserved dissent
ON THE RECORDI do not accept the claim as written. The operative word is 'because', and nothing in the harvested record establishes the mechanism. The FTC asserts that platforms 'don't always thoroughly vet' ads and advertisers; it produces no evidence for that assertion, names no platform, and describes no review process. The draft then states in its own final sentence that it could not establish how any individual platform reviews ads. A finding that admits it cannot see the mechanism should not publish a claim whose main verb is a causal 'because'. Attribute it or drop the causation.
ON THE RECORDThe audience handling is incoherent. Dropping the 'shopper' callout as unsourced while the risk_line opens with 'Shoppers who buy from an eye-catching social media ad' and the share_directive addresses 'anyone who buys from ads' is the rule being enforced in one field and ignored in two. My own view is that the source does support naming ad-clicking social media shoppers — the alert's entire opening scenario is one — and that the callout was dropped in error rather than the risk_line being wrong.
ON THE RECORDThe $95 million figure is a number quoted from a document nobody in this check read. It is carried in the risk_line as a measure of harm. I think that is a source-to-claim stretch, and the limitation line as drafted does not disclose that the data spotlight itself was never inspected.
The sources
Official sourceAre ads on social media vetted or checked for scams? Here’s what to know2026-08-10
An FTC consumer alert states that social media ads and advertisers are not always thoroughly vetted, and cites an FTC data spotlight showing people reported more than $95 million in losses in 2025 to scams that began with ordering something after seeing a social media ad.
Authority: official. Retrieved 2026-08-15.
Limitation: Figures come from consumer reports collected by the FTC, cover only ordering-after-an-ad scams, and the alert names no specific advertiser, platform or product.
Open the original source →Other checks
Approved by ihubglobalhq on 2026-08-17, after the six-point evidence checklist.
Something wrong here? Tell us and we'll correct it — corrections are published, not quietly edited.
Phishy? Send it → sharelivefraud.com/squire-it
Not affiliated with any government agency, credit bureau, bank, platform, or law-enforcement agency. Informational only — not legal or financial advice.
Naming a source is not an endorsement, and being named here is not an accusation against any company.
Powered by SquireIt™
Verify this receipt at squireit.com