What we found
A US Federal Trade Commission consumer alert page dated August 2026 carries the address wording "have you lost money scam watch scammers who say they can help", and two community members asked about it. A second FTC consumer alert page from the same month carries the address wording "searching online bill pay impersonators", and five community members asked about that one. Our harvest captured only those two web addresses and the counts of people asking about them — no headline text, no article body — so we cannot describe what either alert actually says beyond the wording inside the link itself. Nothing we hold names a company being impersonated, a dollar amount, a country or region, or a number of people who lost money. When you need to pay a bill, reach the company through the address printed on your own paper or emailed statement, or an app you installed yourself — never through a search result, a sponsored listing or a link someone sent you. If anyone approaches you offering to get back money you already lost, treat any request for a fee, a card number or account access as the moment to stop and go read the FTC's own consumer alert page yourself. LIVEFRAUD ALERTS is an independent consumer awareness desk and is not affiliated with the Federal Trade Commission or any company referenced here. This check pass rests on two community-shared links and the interest they attracted; we will revisit it if the underlying advisory text reaches our harvest.
- Both rows are bare URLs with community interest counts — no article text was harvested, so the substance of each advisory is inferred only from the words inside the web address.
- The publisher domain is an official consumer protection site, which supports the topic being real, but no quotation, date of incident, figure or named target is available to us.
- Community interest counts are small (two and five people) and say nothing about how widespread either scam is.
- Reviewed by 2 models, 1 from independent houses.
THE RULE
Two live FTC alert topics — approaches promising to recover money you already lost, and impersonator pages surfacing when you search for a company's bill pay — are drawing questions, with no detail in hand to size either.
What we don’t know
- What either FTC alert actually states — no article text was harvested.
- Which companies or bill types the impersonators are posing as.
- How the impersonators are reaching people: paid search listings, ads, calls, texts or otherwise.
- What payment methods are requested and how funds move.
- How many people have been affected, where they are, and how much has been lost.
- Whether the two alerts describe related activity or entirely separate schemes.
The bench — who voted
2 INDEPENDENT AI MODELS REVIEWED THIS. 1 OF 2 — ONE MODEL DISSENTED. READ WHY.
The card names a count. Here are the seats behind it, with what each one said.
ANTHROPICclaude-opus-5 · HIGH
The two rows may not be two distinct pages. Row [2]'s limitation states "The row is the same link repeated with an interest count of five," which conflicts with the finding's assertion of "a second FTC consumer alert page." Either the harvest contains one URL duplicated (in which case the claim of two separate August 2026 alerts, and the separate interest counts, are not established) or the limitation note is describing intra-row duplication. The finding should not resolve that ambiguity in favour of two pages without saying so.
DISSENT, PRESERVEDI grade the pattern high while agreeing the evidence here is thin. That is deliberate: recovery-fraud approaches to people who have already lost money, and impersonated bill-pay contact points surfacing in search, are among the highest-harm consumer patterns in circulation, because the first re-victimises a person already out of pocket and the second captures a payment the victim believes is legitimate and owed. The grade reflects the pattern, not the strength of this harvest.
DISSENT, PRESERVEDThe finding is unusually honest about its own emptiness — it says outright that it cannot describe what either alert says. That candour should be preserved in publication, not smoothed away in editing. Stripping the 'we hold only the link' language would turn a defensible low-confidence note into an implied summary of FTC advisories.
DISSENT, PRESERVEDI would not publish this as a fraud alert at all. It contains no scheme, no target, no method and no loss data. It is a pointer to two FTC pages plus generic hygiene advice. Presenting it under an alert masthead risks training readers that alerts can be content-free.
GROQopenai/gpt-oss-120b · MODERATE
The finding adds consumer advice (e.g., "never through a search result, a sponsored listing or a link someone sent you") that is not present in the evidence rows, which only document link wording and interest counts.
Reviewed by 2 independent models: 1 found it carried by the evidence, 1 did not.
▼ Protocol & challenge record
Objection (high): Content-of-article claims are built from URL slug words while the same finding says the content is unknown. The claim field asserts one page is "addressed to people who have lost money to a scam" and the other is "about impersonators appearing in online searches for bill pay" — both are readings of slug text — then the finding states "we cannot describe what either alert actually says." That is a direct internal contradiction. Slugs are usually descriptive, but they are editorial shorthand, not the article's assertions; the recovery-scam slug could, for example, front an alert about a specific named recovery operation, a report release, or a rule action. Either the summary must be stated strictly as "the link wording reads X" (as the two claim sentences correctly do) or the article text must be fetched.
Not resolved — preserved on the record.
Objection (high): The advice specifies a delivery mechanism the draft admits it does not know. "Never through a search result, a sponsored listing or a link someone sent you" names sponsored/paid listings and inbound links as the vector, while unknowns list "How the impersonators are reaching people: paid search listings, ads, calls, texts or otherwise" as open. Likewise "treat any request for a fee, a card number or account access" imputes a payment pattern to the recovery-scam alert that nothing harvested supports. If this advice is to stay, it must be flagged as generic desk safety guidance, not as guidance derived from these two alerts.
Not resolved — preserved on the record.
Objection (high): The directive_options include "Send this to any searching-online-bill-pay-impersonator you know" and "Forward this to the searching-online-bill-pay-impersonators in your life." These instruct readers to forward the alert to the perpetrators. They are slug-tokenisation artifacts, not audience descriptors, and are self-evidently unpublishable. The same artifact contaminates targeting_dropped, which lists "money", "scam", "searches", "find", "company's", "bill", "page" as if they were audience groups — evidence the targeting extractor is chopping the URL path into person-nouns. None of this should reach a public record.
Not resolved — preserved on the record.
Objection (medium): The interest counts are probably not two and five distinct people asking. Row b31a4552 contains the same URL repeated five times in its body and reports "5 people have asked"; row 230527d9 contains the URL twice and reports "2 people have asked." The count tracks the number of duplicated link instances in the row exactly in both cases. The most economical explanation is that the counter is derived from repeated link occurrences (crossposts, re-shares, a scrape duplicate), not from a headcount of unique humans. Reporting "two community members asked" and "five community members asked" as a fact about people is a stretch on this evidence.
Not resolved — preserved on the record.
Objection (medium): "Community members" and "our community feed" are not what the source says. The rows read "2 people have asked about consumer.ftc.gov" — asked about the domain, not necessarily the specific page, and with no statement that the askers are members of our community or that they asked us. The desk is upgrading an ambiguous engagement signal into a described audience action.
Not resolved — preserved on the record.
Objection (medium): The risk_line calls these "two live FTC alert topics." Nothing harvested establishes that either page is live (no fetch, no status code, no title — only a string), nor that the underlying activity is current. Both rows carry pub 2026-08-18; if this check is being run materially later, "live" also fails on currency. Drop "live" or substitute "two FTC alert URLs dated August 2026."
Not resolved — preserved on the record.
Objection (medium): "Dated August 2026" is inferred from the /2026/08/ path segment and from the harvest row's pub timestamp, which is the date the community row was captured, not a publication date asserted by the FTC. The inference is reasonable for consumer.ftc.gov's known URL scheme, but the finding presents it as the page's date without saying where the date comes from. Say it: "the link path places it in August 2026."
Resolved: Partially resolved on the face of the draft: the two claim sentences already use the careful construction "carries the address wording" and confine themselves to what the string says, and the claim field concedes "we hold only the addresses, not the article text." Remaining fix is one clause naming the URL path as the source of the August 2026 date.
Objection (medium): Newsworthiness threshold. Stripped of slug-reading and generic advice, the entire holding is: two consumer.ftc.gov URLs exist in our feed with small engagement counters. There is no incident, no scale, no target, no figure, no quotation. Publishing a "check pass" on this manufactures an article out of two links and risks readers inferring a fresh, sized threat — precisely what confidence:low and the unknowns list say we cannot support. Fetching the two pages is trivial and would resolve nearly every unknown; shipping without doing so is the weaker choice.
Not resolved — preserved on the record.
Objection (low): Internal inconsistency on geography. The finding says "Nothing we hold names ... a country or region" while also describing the publisher as "A US Federal Trade Commission" agency. The publisher's jurisdiction is US and its consumer alerts address US consumers; callout_options "ATTENTION: EVERYONE" is therefore both unsupported and looser than the one scope hint we do have. Prefer a scope note that the alerting body is US federal, while keeping that no incident location is known.
Not resolved — preserved on the record.
Objection (low): The evidence limitation for row b31a4552 reads "The row is the same link repeated with an interest count of five." Read against row 230527d9 this suggests the two rows share one URL, which is false — they are two distinct slugs. The repetition is intra-row. Reword to avoid asserting the two evidence rows are the same page.
Resolved: Editorial rewording of the evidence limitation field only; no claim depends on it. Suggested: "This row is a distinct URL from the other evidence row; the link is duplicated within the row and carries an interest count of five."
Objection (low): watch_icons "card" and "bank" assert payment-instrument and bank involvement. Neither appears in anything harvested; both are inferred from the advice sentence, which is itself unsourced (OBJ-2). "link" is defensible from the slug; the other two are decoration that implies knowledge.
Resolved: Drop "card" and "bank" from watch_icons, retaining "link", unless the advice sentences are relabelled as generic desk guidance — in which case the icons must still not be presented as derived from these alerts.
Preserved dissent
ON THE RECORDI do not think this should ship. Once you subtract the slug-reading and the generic advice, the desk is holding two URLs and two engagement counters, and it has not even confirmed the pages load. Every question a reader would actually have is in the unknowns list. The two pages are one HTTP request away; publishing a check pass instead of making that request is a choice to write about a link rather than about a scam.
ON THE RECORDI believe the counts are artifacts, not people. Row 230527d9 repeats its URL twice and reports 2; row b31a4552 repeats its URL five times and reports 5. Two-for-two exact agreement is not how independent human question counts behave. Until the pipeline explains the counter, "two community members asked" and "five community members asked" are, in my view, false-precision claims about people, and they are the only quantitative content the piece has.
ON THE RECORDThe directive_options telling readers to forward the alert to "searching-online-bill-pay-impersonators in your life" are not a near-miss to be filtered downstream — they are proof that the targeting stage is treating URL path tokens as human categories, and the same failure produced the seven-item targeting_dropped list of words like "page", "find" and "company's". I would hold the item on that ground alone and fix the extractor before it generates something worse on a story with a named company in the slug.
ON THE RECORDThe claim field's descriptions of what each alert is "about" cannot coexist with the finding's statement that we cannot describe what either alert says. I would rather the piece say plainly "we have two FTC link addresses and nothing else" and stop, than hedge a content summary it is not entitled to make.
The sources
Official source5 people have asked about consumer.ftc.gov2026-08-18
A second FTC consumer alert page from the same month carries the address wording "searching online bill pay impersonators", and five community members asked about that one.
Authority: community. Retrieved 2026-08-18.
Limitation: The row is the same link repeated with an interest count of five; no body text, no impersonated company, no payment method and no loss figures were captured.
Open the original source →Official source2 people have asked about consumer.ftc.gov2026-08-18
A US Federal Trade Commission consumer alert page dated August 2026 carries the address wording "have you lost money scam watch scammers who say they can help", and two community members asked about it.
Authority: community. Retrieved 2026-08-18.
Limitation: Only the link and an interest count of two were harvested; the alert's text, examples, figures and any named targets 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