**TL;DR** — Across 30 client sites through June 2026 we audited a structural choice that lives in how the answer sentence carries its quantity: whether the passage that answers a query grounds its claim in a **concrete quantity** — a percentage, a count, a proportion ("73% of pages fail the reserved-space check.") — or in a **vague quantifier** that gestures at the amount without bounding it ("Most pages fail the reserved-space check."), and whether stating the number changes how often the AI Overview lifts that sentence into the card. Across 7,690 cited-passage events we joined each cited sentence to whether its quantity was concrete or vague. The headline is that quantity-precision is a real citation lever, and it is really a verifiability lever wearing an arithmetic costume. An answer sentence grounded in a concrete quantity was cited 1.8× more often than a matched sentence making the same claim with a vague quantifier on the same query. The strongest predictor was the quantity being checkable — a sentence whose number a reader could hold and verify was lifted far more than one whose "most" or "many" the reader had to take on faith. The second was quantity-in-the-sentence — a number that sat in the lifted sentence beat one parked in an adjacent chart the composer did not lift. The third, and the warning, was false precision — a fabricated-looking decimal ("73.4% of pages") on a claim that could not plausibly be measured that finely was cited no more, and on 6% of pages the composer paired it with a competitor's cleaner round figure. One change — replacing vague quantifiers in answer sentences with the concrete quantities the page already had — lifted cited-passage rate by 18% on the affected sites over a 30-day follow-up.
Why we ran this audit
The AI Overview composer lifts a single sentence and drops it into a card as the answer to a query, and that sentence stands in front of a reader who has none of the page around it to back it up. A whole class of answer sentences leans on a vague quantifier to carry the load a number should carry: "most sites get this wrong", "many pages reflow", "a few simple fixes cover it". Read inside a flowing article these feel authoritative, because the surrounding paragraphs supply the evidence; lifted cold into a card, the quantifier is a claim about an amount with no amount attached — "most" could be 51% or 95%, and the reader has no way to tell, so the card asserts a proportion it never states. We had spent weeks on the shape of the answer sentence — its polarity, whether it named a cause, whether it defined its term — and how the sentence carries its quantity is the natural next structural variable, because a quantifier is the one place a sentence claims a measurement while withholding the measurement.
The second motivation was a drafting habit that reaches for the quantifier because the number is elsewhere. A page runs the study, puts the figure in a chart or a table, and then, in the prose, writes "most pages failed" rather than "73% of pages failed" — because the writer knows the reader can see the chart, and repeating the number in the sentence feels redundant to someone reading top-to-bottom. But the composer does not see the chart; it lifts the prose sentence, and the prose sentence gave away the number to the figure beside it. We needed to know whether folding the concrete quantity the page already had into the answer sentence — at the cost of a number that reads slightly redundant next to its own chart — bought the citation, because if it did, the fix is nearly free: move the figure from the caption into the sentence the composer would lift.
How we ran the measurement
30 client sites — 11 SaaS, 6 publisher, 8 B2B services, 5 DTC — each with a fixed 200-query basket of its real in-market queries, deliberately weighted toward queries whose best answer turns on an amount ("how common is X", "how much does Y help", "what share of Z"). Twice daily through June 2026 we captured every AI Overview card, and for cards citing a client page we identified the specific lifted sentence and classified how it carried its quantity: concrete (a percentage, count, or proportion sits in the sentence), vague (a quantifier like "most", "many", "several", "a few" with no number), or off-sentence (the number lives in an adjacent chart or caption but not in the prose). For each cited sentence we built a matched control: a comparable sentence on a similar query whose quantity-carrying differed but whose underlying claim was the same, so the comparison was concrete-vs-vague rather than good-page-vs-bad-page. The cited cohort was 7,690 events.
Two normalisation moves matter. We scored shape on the sentence as it would be lifted — alone, with no surrounding context — because that is the unit the composer extracts, and a vague quantifier that reads as well-supported inside an article full of charts reads as an unbacked assertion in the card. And we matched on sentence citability before comparing shape — we paired each cited sentence with a control our existing cited-paragraph rubric scored as equally liftable (concrete, on the query, factually complete), so the effect we attribute to the quantity is not just the numbered pages being better written overall. The 1.8× and 1.5× figures are from those matched comparisons, not raw averages.
The shape of the quantifier pattern
The flat headline first. A sentence grounded in a concrete quantity was cited more. An answer sentence carrying a percentage, count, or proportion was lifted 1.8× more often than a matched sentence making the same claim with a vague quantifier on the same query. The effect held through the quality match and the citability control: among sentences our rubric scored as equally liftable, the numbered ones were lifted far more than the quantifier ones. The composer behaves as though it prefers a claim a reader could check — one that states the amount it is asserting — over one that names an amount ("most", "many") and leaves its size to the reader's guess, because a card that says "73% of pages fail this" answers the amount the query asked about while a card that says "most pages fail this" restates the question as a vaguer version of itself.
The most decision-relevant cut was that this is about the quantity being checkable, not about a digit being present. We tested whether the win came from any number appearing or from the number actually pinning the claim, and the second was the whole story: a sentence with a decorative number that did not bound the claim ("There are countless ways — at least 5 — to break this") was cited no better than a bare quantifier, while a sentence whose number pinned the amount ("73% of pages break this") was lifted far more. The concrete quantity wins when it makes the claim verifiable. State the number that bounds the claim, not one that decorates it.
Driver one: give the reader a number they can check
The single strongest predictor was whether a reader could verify the amount the sentence asserted. Holding the claim constant, a sentence with a concrete quantity was lifted at 1.8× the rate of one with a vague quantifier. The composer extracts a sentence and puts it in front of a user who did not read the page; a sentence that says "73% of pages fail the reserved-space check" answers the question and hands the reader a figure they could confirm, while one that says "most pages fail the reserved-space check" answers with a word that could mean anything above half — and a card that asserts an unstated amount is, to the composer, a worse answer than one that states it. A human reading the page in order sees the chart under the sentence; the reader in front of the card sees only the sentence, and the composer rewards the sentence that carries its own number.
We ran a structural test on 28 answer sentences across 15 clients, each a strong claim on a query where the sentence used a vague quantifier for an amount the page had actually measured and shown in a figure. We rewrote each to fold the concrete quantity into the sentence, changing no underlying claim — only lifting the number that lived in the chart into the prose the composer would extract. Over the 45 days that followed, 20 of the 28 sentences began being lifted on at least one query where the vague version had been skipped. The lever was not new data; it was moving the figure the page already had into the single sentence the composer would lift, so the sentence carried the amount it was asserting.
Driver two: the number has to be in the sentence, not the chart
Holding concrete-quantity constant, the second driver was where the number sat. A figure that lived in the answer sentence beat one parked in an adjacent chart, caption, or table cell that the composer did not lift. The reading consistent with the data is that the composer lifts one self-contained sentence and does not reach into the figure beside it, so a page that put its number in a chart and its prose in a quantifier gave the composer a sentence that still read as vague — the number was on the page but not in the unit that got extracted. A sentence-adjacent chart is invisible to the lift; only the number inside the sentence rides along into the card.
We ran a structural test on 18 off-sentence pages across 11 clients, each of which had the exact figure in a chart or caption but a vague quantifier in the prose sentence. We rewrote each to state the figure in the sentence itself, keeping the chart as the fuller evidence, changing no data. Over the 60 days after the change, 13 of the 18 sentences improved their cited-passage rate. The two drivers compound: a concrete quantity beats a vague one, but only if the concrete quantity is in the sentence — the sentences that won stated their number in the prose the composer lifts, not in the figure it leaves behind.
Driver three: false precision, and the number no one could measure
The third driver was the warning. A concrete quantity helps only when it is plausibly measured, and bolting a fabricated-looking decimal onto a claim that could not be pinned that finely backfires. A sentence like "73.48% of pages have a suboptimal hero image" — a spurious two-decimal figure on a soft, judgement-laden claim — was cited no more often than the honest round version ("about three-quarters of pages have an oversized hero image"), and on 6% of audited pages the composer paired the false-precise sentence with a competitor's cleaner round figure, so the citation was shared rather than won. The reading consistent with the data is that the composer discounts a precision the claim could not support; a decimal on an un-measurable quantity reads as invented, and an invented-looking number is a trust cost, not a trust signal. Match the precision to the measurement — a round number for a rough claim, a sharp one only for a sharply measured amount.
We confirmed this on 15 sentences across 9 clients where an earlier optimisation pass had added spurious decimals to soft claims. We rewrote each back to a round figure honest about its own precision ("roughly 70%") while keeping the sharp decimals only on the sentences whose amounts were genuinely measured to that resolution. Over the following 45 days the honest round figures regained their solo citation while reading as credible, and none drew a shared-citation pairing. The actionable rule is blunt: a concrete quantity beats a quantifier, but a number precise beyond its measurement reads as fabricated — round the figure to the confidence the measurement actually earns.
What changed in our content checklist
Three changes. We added a quantity pass for answer sentences: before publishing, we read each section's lead answer sentence alone and check that any amount it asserts is a concrete quantity, not a vague quantifier — because the composer lifts the sentence whole and hands it to a user who cannot see the chart, so a "most" or "many" is a claim about an amount with the amount withheld. We added a location check to the same pass: the number lives in the sentence, not only in an adjacent figure, so when the composer lifts the sentence the amount rides along. And we added a precision guard: the number is rounded to the resolution the measurement actually supports, so a soft claim carries a round figure and a false-precise decimal never reads as invented.
We dropped one habit. For years our style had been to put the figure in a chart and let the prose gesture at it with a quantifier — on the belief that repeating the number in the sentence read as redundant to a reader who could see the chart right there. The audit removes that default for the answer sentence: the one sentence the composer would lift travels without its chart, and a quantifier lifted alone asserts an amount it never states. So vague quantifiers left our playbook for the answer sentence — we now fold the concrete figure into the lead answer sentence, accepting that it reads slightly redundant next to its own chart, because it is built to carry its amount to the reader who cannot see the chart at all.
- 01Ground the answer in a concrete quantity. A sentence with a percentage or count was cited 1.8× more than one making the same claim with a vague quantifier — the composer reads a numbered claim as checkable and a "most" as an unbacked assertion.
- 02Pin the claim, do not decorate it. A decorative number that did not bound the claim was lifted no more than a bare quantifier — the win comes from the number actually stating the amount asserted.
- 03Put the number in the sentence. A figure in an adjacent chart the composer did not lift was invisible to the extraction — only the number inside the sentence rides along into the card.
- 04Do not fake precision. A spurious decimal on an un-measurable claim was cited no more, and on 6% of pages the composer shared the citation with a competitor that rounded honestly.
Where this argument breaks
For queries whose answer is not about an amount — definitional, procedural, or yes/no questions where no quantity is at stake — there is nothing to make concrete and the lever is irrelevant, so it is for answer sentences whose claim turns on how much or how many. For claims a page has genuinely not measured, inventing a number to satisfy this is the false-precision failure, not the win — a vague quantifier honest about its own uncertainty beats a fabricated figure. For navigational and brand queries there is no answer sentence whose shape matters. For narrative and persuasive passages — case studies, opinion, story-driven content — a quantifier is a voice choice, not a citation lever, and the quantity pass is for the answer sentences on informational queries only. For some languages the effect may differ — in our parallel Chinese-language audit (文心一言, 元宝, 通义) the concrete-quantity win was present but the vague quantifier was carried more by a «大部分» / «不少» frame the composer sometimes read as a soft hedge and skipped cleanly, so the unbacked-assertion problem was milder. The 6% shared-citation figure is small and noisy; we are confident a false-precise decimal does not help and mildly confident it splits the citation, but it is the weakest finding here and we would not restructure a page on it alone. Our window was 60 days and the cohort was 30 sites; the multipliers are point estimates that will move by vertical and query type. Outside those carve-outs the lesson holds: in 2026 the AI Overview lifts an answer sentence grounded in a concrete quantity — the number stated in the sentence, rounded to what the measurement earns — more readily than one that leans on a vague quantifier, the unit is the individual answer sentence rather than the page, and the cheapest citation win on an amount-shaped query is to move the figure out of the chart and into the sentence that makes the claim.