What we found
Five community lookups in our harvest asked about a Federal Trade Commission consumer alert page whose web address refers to searching online for bill pay impersonators. Two lookups asked about a second FTC consumer alert page whose address refers to having lost money to a scam and to scammers who say they can help. Three lookups asked about an American Bar Association page whose address refers to a grandparent kidnapping scam. Our harvest captured only those web addresses and the number of people asking about each one; we hold none of the text of the three pages, so we cannot say what any of them actually describes. We do not know who asked, where they were, whether they had been approached by anyone, or whether any money was lost. When a bill needs paying, open the company's own statement, card or account app and use the contact details printed there — never a number or link that came back from a search. LIVEFRAUD ALERTS is independent and not affiliated with any agency, association or company mentioned here, and this alert reports only what our harvested rows contain.
- All three rows are community lookup records containing a web address and a count, with no page text harvested.
- No source narrative, victim account, loss figure or timeline was captured, so nothing describes how any approach is actually made.
- The subject matter of each page is inferred only from the wording inside its own address.
- Reviewed by 3 models, 2 from independent houses.
THE RULE
People are actively searching for help after a scam approach, and the search results themselves are a place impersonators are known to be looked for.
What we don’t know
- What each of the three pages actually says.
- How contact is made in any of these scams, and what the caller or website asks for.
- How any money is said to move, and whether anyone in our harvest lost money.
- Who the people asking are, and where they are.
The bench — who voted
3 INDEPENDENT AI MODELS REVIEWED THIS. 2 OF 3 — ONE MODEL DISSENTED. READ WHY.
The card names a count. Here are the seats behind it, with what each one said.
ANTHROPICclaude-opus-5 · MODERATE
The word 'clustered' in the claim dresses up ten lookups spread across three addresses (5/2/3) as a pattern. Nothing in the rows establishes that these lookups are related to one another, came from distinct people, or occurred in any shared window. The finding body is appropriately flat, but the headline claim leans on a word the counts do not earn.
DISSENT, PRESERVEDThe evidence here is genuinely thin — three URL strings and three small counts — and the finding deserves credit for saying so four separate times rather than papering over it. My 'holds' verdict is about that discipline, not about the strength of the underlying material.
DISSENT, PRESERVEDI grade the pattern moderate rather than high only because the harvest establishes no contact, no approach and no loss. If any of these addresses do correspond to what their wording suggests — recovery-fraud follow-on and grandparent kidnapping impersonation in particular — the real-world harm to the people who eventually meet those scams is severe, and readers should not take my moderate as a comment on those scam families themselves.
DISSENT, PRESERVEDI would record that a lookup count is not a victim count. Nothing in these rows tells us whether anyone asking had been targeted or was merely curious, and future alerts in this series should not begin treating repeat lookups as an incidence signal.
GROQopenai/gpt-oss-120b · LOW
Raised no objection; read the evidence as holds.
OPENROUTERopenrouter/free · MODERATE
The finding attributes specific content to the three pages (e.g., 'bill pay impersonators,' 'lost money to a scam,' 'grandparent kidnapping scam') based solely on their web addresses, but the evidence rows explicitly state that the alert/article text was not harvested. The evidence only confirms lookups occurred for those addresses and counts, not the pages' actual content or context.
Reviewed by 3 independent models: 2 found it carried by the evidence, 1 did not.
▼ Protocol & challenge record
Objection (high): The risk_line contradicts the finding's own limitations. "People are actively searching for help after a scam approach" is not in evidence: the three rows contain URLs and counts only. Nothing indicates any asker was approached, victimised, or seeking help — a person may look up an FTC alert out of curiosity, for work, as a researcher, or because a link was shared. The second half, "the search results themselves are a place impersonators are known to be looked for," is derived entirely from the slug 'searching-online-bill-pay-impersonators', i.e. from the very page text the draft admits it does not hold. The risk_line re-imports the inference that the limitation sentence disclaims, which is the most serious integrity failure in this draft.
Not resolved — preserved on the record.
Objection (high): Unaddressed alternative explanation for the 'cluster'. All three rows were published within roughly five minutes of each other (01:42:53, 01:42:55, 01:47:50 on 2026-08-18). That pattern is at least as consistent with a single ingestion batch, one crawl, one aggregator feed, or a single session/actor repeatedly hitting the same three links as it is with independent community concern. The draft treats "5 people", "2 people", "3 people" as face-value counts of distinct askers and builds a 'clustering' narrative on them, without recording the near-simultaneous timestamps or listing 'whether these lookups are independent' as an unknown. If the counts are artefactual, the entire item has no subject.
Not resolved — preserved on the record.
Objection (high): The advice sentence is not supported and is topic-mismatched. Telling readers to use contact details from their own statement/card/app rather than search results is a specific countermeasure to a specific threat model — search-result impersonation of billers — that exists here only as words inside one URL path. The draft says twice it cannot say what the page describes, then dispenses the page's presumed advice anyway. It also covers only one of the three URLs: nothing is offered on the recovery-scam page (2 lookups) or the grandparent-kidnapping page (3 lookups), so the advice silently privileges the highest-count slug and misrepresents the harvest's shape.
Not resolved — preserved on the record.
Objection (medium): Source-to-claim stretch in characterising the pages. The claim line calls all three "consumer-protection pages" and the ABA item an "article"; the finding calls the FTC items "consumer alert page[s]". Page type and subject matter are being read off URL path segments. '/consumer-alerts/' is a reasonable read for FTC, but the ABA URL is '/groups/senior_lawyers/resources/voice-of-experience/2026-march/' — that is a member-section opinion column for senior lawyers, not a consumer-protection notice. Calling it a consumer-protection page is an unsourced upgrade of its authority and purpose.
Not resolved — preserved on the record.
Objection (medium): Date currency is not disclosed to the reader. The finding gives no observation date or window at all. The two FTC URLs carry '/2026/08/' and were seen on 2026-08-18, so they are contemporaneous; the ABA URL carries '2026-march', making it roughly five months old at harvest. Presenting all three as one undifferentiated 'cluster' hides that one item is materially older, which matters if the reader is being invited to infer live activity.
Resolved: Curable in-line: state the harvest date (2026-08-18), note that both FTC addresses carry an August 2026 date segment while the ABA address carries a March 2026 segment, and drop any implication that the three are contemporaneous.
Objection (medium): watch_icons 'phone' and 'person' are unevidenced. No harvested row mentions a call, a caller, a number, or an identified individual. The icons assert a contact channel and a human actor that the unknowns list explicitly says are unknown ('How contact is made in any of these scams'). Icons are claims to a skimming reader and should be held to the same standard as sentences; only 'link' survives, and even that is thin.
Resolved: Curable by deletion: drop 'phone' and 'person'; retain 'link' only.
Objection (medium): The offered callout_options and directive_options include items that would breach the rule the draft itself cites. 'ATTENTION: GRANDPARENTS' and 'ATTENTION: PARENTS' name a risk group on the strength of one URL slug — exactly the §11 Rule 2 problem the targeting_dropped list is policing. 'Send this to any searching-online-bill-pay-impersonator you know' and 'Forward this to the searching-online-bill-pay-impersonators in your life' are incoherent (they instruct readers to forward the alert to impersonators) and slur-adjacent. Presenting these as live options is a defect even though none was selected.
Resolved: Curable by deletion: strike 'ATTENTION: GRANDPARENTS' and 'ATTENTION: PARENTS' and both 'searching-online-bill-pay-impersonator' directives from the option sets before any selection step sees them.
Objection (medium): The targeting_dropped machinery has plainly misfired: it treats bare tokens ('money', 'scam', 'search', "company's", 'bill', 'payment', 'number') as named audience groups. This is noise, and it undermines confidence that the same filter would catch a real targeting claim. It also suggests the advice sentence was rewritten around a token filter rather than around what the evidence supports.
Not resolved — preserved on the record.
Objection (low): Unit slippage between 'people' and 'lookups'. The rows say '3/2/5 people have asked'; the draft renders these as 'lookups' in some places and 'askers' in the evidence limitations, and writes the ungrammatical 'Two lookups asked'. Either unit may be right, but the draft should say which the field actually records and flag that it cannot distinguish distinct persons from repeat lookups.
Resolved: Curable in-line: quote the field as it stands ('N people have asked about'), use one unit consistently, and add an unknown: whether the count reflects distinct people or repeat lookups.
Objection (low): 'our harvest holds only the addresses, not the pages' is slightly inaccurate: the rows also hold the domain, a count, and a publication timestamp. Minor, but the claim line is the place where precision about what is held matters most.
Resolved: Curable in-line: rewrite the claim as 'holds only the addresses, a count and a timestamp — none of the page text'.
Objection (low): Reputational adjacency risk: an item headed as a fraud alert names the FTC and the American Bar Association while conveying no content from either. The non-affiliation disclaimer helps, but a skim reader can come away associating the named bodies with the scam rather than with the warning. Consider foregrounding that these are the bodies' own warning pages.
Not resolved — preserved on the record.
Preserved dissent
ON THE RECORDI do not think this should ship as an alert. Strip the risk_line, the advice and the unsupported icons — which is what the evidence requires — and what remains is: ten link lookups hit three warning pages on one night, and we have none of the pages. That is a pipeline observation, not a fraud warning to the public. Publishing it in alert format lends scam-alert weight to a record that contains no scam.
ON THE RECORDThe risk_line is the worst sentence in the draft. It states as fact that people were approached by scammers and that impersonators lurk in search results, three sentences before the finding says we do not know whether anyone was approached and cannot say what any page describes. If one part of the item has to be believed and another disbelieved, the item is not honest, and the confidence label 'low' does not cure that.
ON THE RECORDThe five-minute spread across all three publication timestamps is the single most important fact in this harvest and it appears nowhere in the draft. I read it as most likely a batch artefact. Until someone rules that out, the word 'clustered' should not be used, because it invites the reader to see human concern where there may only be a crawl.
ON THE RECORDOffering 'ATTENTION: GRANDPARENTS' as a callout in the same object that drops 'money' and 'bill' as unevidenced audience terms shows the safeguard is filtering on the wrong thing. A slug-derived 'grandparent' is precisely the unsourced claim about people the rule exists to stop; 'bill' was never a claim about people at all.
ON THE RECORD'Send this to any searching-online-bill-pay-impersonator you know' should never have been generated. It is nonsense, and had it been selected it would have accused the reader's contacts of being fraudsters. I want that on the record as a generation defect, not a stylistic quibble.
The sources
Reported5 people have asked about consumer.ftc.gov2026-08-18
Five community lookups in our harvest asked about a Federal Trade Commission consumer alert page whose web address refers to searching online for bill pay impersonators.
Authority: community. Retrieved 2026-08-19.
Limitation: Row contains the repeated address and a lookup count only; none of the alert's text was harvested.
Open the original source →Reported3 people have asked about americanbar.org2026-08-18
Three lookups asked about an American Bar Association page whose address refers to a grandparent kidnapping scam.
Authority: community. Retrieved 2026-08-19.
Limitation: Only the address and a count of three askers were captured; nothing from the article body was harvested.
Open the original source →Reported2 people have asked about consumer.ftc.gov2026-08-18
Two lookups asked about a second FTC consumer alert page whose address refers to having lost money to a scam and to scammers who say they can help.
Authority: community. Retrieved 2026-08-19.
Limitation: Only the address and a count of two askers were captured; the alert's contents are not in our material.
Open the original source →Other checks
Approved by ihubglobalhq on 2026-08-19, 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