Newsjack Accuracy Audit — 22 July 2026

26 highest-risk articles checked against their own cited sources (all 54 were indexed; the 26 with both a dramatising verb and a number in the headline were verified in full).

Result

14 of 26 (54%) need a change. Not one number was fabricated. Every figure exists somewhere. The failure is almost entirely in how they’re labelled.

The two failure modes

1. Vintage drift — last year’s data presented as this year’s. Pet insurance ($5.2bn is 2025), vending ($40.04bn is 2025), digital exhaustion (Asana, fielded 2025), household net worth ($184T was Q4 2025 and has since been superseded by $183.0T for Q1 2026), 1099-K (the threshold changed retroactively in 2025, so 2026 is simply the second year at that level).

2. Denominator swaps — one population relabelled as another. Squaremouth’s insured travellers → “the average American vacation” (Deloitte puts the typical trip just over $4,000, not $9,032). Zillow listings offering concessions → “40% of landlords”. JCHS multifamily properties → “landlords”. Vanguard plan participants → “workers”. TICAS’s one illustrative household → “payments just jumped $400/month”. Fed directly-held equities → “stocks” (total exposure is ~36%, not 25.6%).

Both are cheap to fix and cheap to prevent.

❌ WRONG — fix first

credit-card-debt-1-25-trillion-delinquencies-2026.md Headline: “Credit Card Debt Hits $1.25 Trillion as Delinquencies Reach a 15-Year High” The cited NY Fed Q1 2026 release is titled “Delinquency Transition Rates Hold Steady”. Card balances fell $25bn (down from Q4’s $1.277T record); card early-delinquency transitions ticked down 8.7%→8.6%; serious flat 7.04%→7.10%. → Both halves of the headline are wrong. The real story in that release: student loan 90+ delinquency rose to 10.3% and ~2.6m borrowers were sent to the Dept of Education’s Default Resolution Group.

bnpl-now-affects-fico-score-2026-hidden-debt.md Headline: “Your BNPL Debt Now Hits Your FICO Score”; FAQ: “It can now”; body: a missed payment “can shave roughly 20 to 50 points.” Affirm, 25 Nov 2025: “Paying Affirm back on time or early doesn’t raise your score today… Missing Affirm payments doesn’t lower your credit score today.” Senate Banking letters, 4 May 2026: most providers don’t report; Experian collects BNPL data but doesn’t use it in its scoring model. No lender has publicly confirmed using FICO 10 BNPL. → Reframe to “built, but not in use yet.”

only-4-percent-pets-insured-2026-naphia-track-insurance-roi.md $5.2bn and 4.27% are calendar-2025 figures from NAPHIA’s 2026 report. The article also cites the 2025 report URL. → “Pet Insurance Hit $5.2 Billion in 2025…” and cite the 2026 SOI report.

student-loan-payments-spike-july-2026-budget.md “$400/month” is not an average. It’s TICAS’s single illustration: a family of four on the median $81,000 income moving from SAVE to RAP goes $36 → $440. → “SAVE ended July 1 — one advocacy estimate has a median-income family of four going from $36 to $440/month on RAP.”

⚠️ MISLEADING — reword

• 1099-K — threshold is right, but it reverted retroactively in 2025. “Jumped back in 2026” → “Is back at $20,000”.
• 85% freelancers — 85% are paid late “at least some of the time” (Remote, worldwide). The 120-day terms are ANA client-to-agency data from 2013–19, not client-to-freelancer.
• Grocery 3-year high / coffee — headline blames tariffs, but coffee was removed from reciprocal tariffs in Nov 2025; the driver is Brazilian drought (the body says this correctly). USDA forecasts food-at-home +2.8% for 2026. Body also says coffee is “nearly doubling” while the headline says +18% — they contradict each other.
• Vending $40bn — that’s convenience services (vending + micro markets + coolers + office coffee) in 2025, not vending alone.
• Rental vacancy “hit 7.3%” — Census says it is not statistically different from 7.1% a year ago. And the 40% is Zillow listings, not landlords.
• Household wealth $184T — superseded. Q1 2026 is $183.0T (Z.1, 11 Jun 2026).
• Summer vacation $9,032 — Squaremouth’s insured trips. Squaremouth itself published $9,668/+24% in May and $9,032/+17% in July.
• Digital exhaustion 84% — not in the cited Lokalise report at all; it’s Asana’s 2025 five-country knowledge-worker survey.
• Americans hold 25.6% in stocks — directly-held equities only, Q4 2025. Including funds and retirement accounts it’s ~36%.

❓ UNVERIFIABLE

• Bathroom remodel +4–6% — the only support is two contractor marketing pages. No authoritative body publishes this. Harvard JCHS actually forecasts remodelling spend growth decelerating through 2026. Rebuild on Angi/NAHB data or drop.

✅ ACCURATE — no change needed

Wedding $34,200 (The Knot 2026 study) · Teachers $895, +49%/decade (AdoptAClassroom 2025) · 401(k) $24,500 / IRA $7,500 (IRS IR-2025-111) · ATRI $2.336/mile record (2025 ops) · QBI permanent + $400 minimum (OBBBA §199A(i), first applies 2026) · Upwork 38% (Future Workforce Index 2026, n=2,400) · Tariffs on steel/lumber/cabinets (NAHB) · Kitchen remodel $26,943 (Angi 2026) · Landlord opex +10–20%, insurance doubled (JCHS 2026) · 47% scaled back travel (Omnisend/Cint, n=1,075) · Section 179 $2.56M + 100% bonus (Rev. Proc. 2025-32) · Grocery 4.5% forecast (real, but it’s one named economist, not “economists”)

Also verified for reuse: 2026 IRS standard business mileage rate = 72.5 cents/mile (IR-2025-128).

The prevention fix

Both failure modes are caught by two questions added to the newsjack prompt, asked after drafting and before publishing:

1. “What year is this data actually from, and does my headline say that year?”
2. “Who exactly was measured, and does my headline describe that same group?”

Plus one rule: a superlative (“record”, “highest”, “hit”, “3-year high”) may only appear in the headline if those words appear in the primary source. Every failure above is a superlative or a direction of travel the source didn’t claim.