Why it matters.
Statistics help a page get used in an AI answer. The GEO study by Aggarwal and colleagues (KDD 2024; arXiv version of 28 June 2024) found that adding statistics, quotations and citations could raise a source’s visibility in generative engine responses by up to 40%.
But a number only helps if it survives being lifted. “3x growth” with no unit or period cannot be checked, and an engine may pair it with an older figure it found elsewhere. Units, dates and a named source make the number usable and harder to garble.
The 10-minute self-check.
Copy every number from your homepage and your three main service pages into a sheet. For each one, mark:
- Unit: users, key events, pounds, impressions?
- Period: which dates, compared with which dates?
- Source: which tool or report?
- Evidence: is there a dated screenshot on file?
Any number missing two of the four should be rewritten or removed.
Steps.
- Start a claims register: one row per number, with source, date range, evidence file and status.
- Rewrite each claim in the full pattern: from X to Y, over the period, measured in the tool.
- Pick one canonical version of each result and retire the others everywhere you can find them.
- Keep numbers in text next to the claim, not only in an image.
- Publish only rows marked approved. Pending numbers never go live.
- Review the register every quarter.
The register is the base of every proof page in our AI search visibility work.
What we saw on prolinkage.com.
How to measure it.
- Completeness: share of published numbers with unit, period and source. Aim for all of them.
- Echo check: each month, ask the engines what results your company has achieved and note whether the numbers they repeat are current.
- Conflicts: versions of the same result still live anywhere on the web.
Sources.
- Aggarwal et al., the GEO study (KDD 2024), arXiv 2311.09735, v3, 28 June 2024.
- Prolinkage claims register, 8 November 2026.