Can Anyone Quote Your Page? 70% of Marketing Content Says No

For the surrounding account decisions, compare the full benchmark and the method.

In short

29.5% of pages pass citation readiness. The check is simple: does the page contain a concrete figure, and is there attribution language near it — according to, source:, research shows, study found? Seven in ten pages fail, and they nearly always fail for the same reason.

The two halves, and which one is missing

The check needs both. The industry has one.

statistics_with_sources — which asks for at least three figures plus an attribution phrase — passes on 57.6%. So a majority of pages do carry numbers. But citation_friendliness, which asks whether a figure and its attribution sit close enough together to be lifted as a unit, passes on only 29.5%.

The gap between those two rates is the finding. Pages have numbers. The numbers have no visible provenance. A page asserting that CPMs rose 23% year over year, with no indication of where that came from, is unquotable by anything that cares about attribution — which is precisely what an assistant citing a source cares about.

Why this happens

Three causes, in rough order of frequency.

The number came from somewhere the writer cannot name. An internal deck, a conference talk, a figure that has circulated so long its origin is lost. Writers know it is soft, so they state it without a source rather than cite something they cannot stand behind. The instinct is right; the output is an unusable claim.

The source is in a link, not in the text. A hyperlinked word carries the source for a human who hovers. Text extraction frequently loses the anchor, and the sentence arrives naked.

The page has no numbers to source. This is our own failure mode. Our corpus passes citation readiness on 44.4% — well above the industry’s 29.5% — and passes statistics_with_sources on only 47.7%, below the industry’s 57.6%. The reason is a deliberate house rule against publishing benchmark figures we cannot source, which leaves a large number of posts carrying no figures at all. Refusing to invent numbers is correct; it has a measurable cost and this is where it shows.

What passing looks like

Not complicated. The source goes in the sentence, in text, next to the number:

According to Meta’s Business Help Centre, an ad set exits the learning phase after roughly 50 optimisation events in the week following its last significant edit.

That sentence can be lifted whole, and it arrives with its provenance attached. Compare:

Ad sets need about 50 events to exit learning.

Same fact. Unquotable, because a careful consumer has no way to know whether it is documented or folklore.

The rule that actually works

If you cannot name the source, do not publish the number. This is more demanding than it sounds and it is the right constraint, because the alternative — publishing plausible figures with vague provenance — degrades the thing that makes content worth citing in the first place.

The corollary is the harder work: if a page genuinely needs a figure and no public source exists, the options are to measure it yourself and publish the method, or to write the page without the number. Both are legitimate. Inventing a defensible-sounding percentage is not.

Interpretation boundary

The check looks for attribution language near a figure, which means a page can pass it with a phrase like “research shows” attached to nothing real. It measures the form of sourcing, not its validity. A page scoring well here can still be citing a source that does not support the claim — which no structural check can catch, and which remains a matter of editorial honesty rather than markup.

Can software help?

Bach.ai audits your connected Meta account, estimates the revenue impact of what it finds, and proposes specific fixes. It applies a change only after you approve it. Think of it as an automated audit layer that surfaces issues and proposed fixes for your review — not a replacement for your team’s judgment, and creative production is not its core job, though the Pro and Agency plans can generate a limited number of variants.

FAQ

What makes content easy for AI to cite?

A concrete figure and an attribution phrase sitting close enough together that a sentence can be lifted with its source intact. Only 29.5% of the pages we measured managed both.

Why do pages with statistics still fail citation checks?

Because the statistics have no visible provenance. 57.6% of pages carry figures but only 29.5% carry them with attribution nearby. The source is often in a hyperlink, which text extraction frequently loses.

Should I publish a number I cannot source?

No. Either measure it yourself and publish the method, or write the page without it. Publishing plausible figures with vague provenance degrades exactly the property that makes content worth citing.

Does refusing to publish unsourced numbers hurt your scores?

Measurably. Our corpus passes citation readiness on 44.4% against the industry’s 29.5%, but passes the statistics check on only 47.7% against their 57.6% — because many of our posts carry no figures at all. That is the cost of the rule, and we think it is worth paying.

Method and sources

“A figure and an attribution phrase close together, so a sentence can be lifted with its source intact.”

Source: the Bach.ai AI Extractability Benchmark, run 2026-09-22. Sample: 132 pages returning HTTP 200, drawn from the published sitemaps of 12 advertising and analytics tools, plus our own 392 published posts. Every page was rendered in a headless browser and scored by the same 17-check engine the product runs against customer sites, so each figure is reproducible against the live web rather than asserted. Read the per-check rates as an industry signal, not a precise ranking of any one vendor: between 3 and 12 pages were sampled per tool, spread across the whole sitemap rather than drawn from the newest posts. The method is written up in full at how we measure AI extractability.