What does a line like "3× AI mentions in 60 days" actually prove? Nothing yet. Answer four questions about the denominator first. Until then that number has no unit, no floor, and no way to be wrong. (The line is a format example, not a quote from any vendor.) This part of Verify Before You Buy turns the pattern into four questions a buying desk should ask before comparing AI-visibility vendors.
The ground such claims stand on keeps moving. On Semrush's keyword panel, AI Overviews fired for 6.49% of queries in January 2025, 24.61% in July, and 15.69% in November. Published research reports 10–34% swing in output from sampling alone. That is why the check is a per-metric denominator, with n and a Wilson 95% confidence interval printed next to the rate.
About this series. Verify Before You Buy reads AI-visibility marketing claims the way a buying desk should. One reporting pattern per part, taken apart and turned into questions you can ask in writing before you sign. The subject is the reporting format, not any one vendor. Every sales line quoted in this series is a labeled format example. Every measured figure we cite carries a source and a date you can check.
What does a lift claim leave out?
A lift is a ratio of two rates, and every rate is a fraction. Before the "3×" means anything, four pieces of the fraction have to exist in writing:
- The denominator. A share of what? All tracked prompts? Only the prompts where an AI answer appeared? Only valid runs, after failures come out? Each choice gives a different number from the same data. A report that never names its denominator can switch between them and never be wrong.
- The sample. How many prompts, and how many runs of each. A "lift" taken over a handful of prompts read once is a different thing from one taken over a pinned panel with repeated runs. The headline percentage can look identical either way.
- The query-set version. Were the questions frozen between the "before" and the "after"? If the prompt list changed in the middle, the two measurements are of different things. Their ratio is not a lift. It is a coincidence.
- The window and the baseline protocol. Same engines, same locale, same repetition count, stated dates on both sides. Without this, "before versus after" quietly becomes "one setup versus another."
None of these questions accuses anyone of anything. They are the questions any rate has to answer before it goes into a budget — which is exactly why asking them in writing is such an efficient filter.
Can the denominator itself move? It does
There is a structural reason to insist on the denominator in writing: on AI surfaces, the denominator is not stable ground. It is a dial the platform turns.
The clearest published example: on Semrush's 10M+ keyword panel, AI Overviews were triggered for 6.49% of queries in January 2025, 24.61% in July, and 15.69% in November (Semrush AI Overviews study, vendor-published panel figures, data January through November 2025; accessed July 17, 2026). The surface roughly quadrupled and then shed a third of its footprint within one year, with no action required from any brand or agency.
The same denominator, three readings in one year
Share of panel queries that triggered an AI Overview (Semrush, 10M+ keywords, 2025)
Now put the format example next to that curve. A mention rate measured against "answers that appeared" can triple because your content earned it. It can also triple because the platform turned the dial and the denominator shrank under the fraction. From the headline number alone, the two cases look the same. A lift measured against a moving denominator is not something you can procure against, unless the report states which denominator it used and holds it fixed across the comparison.
And even with a frozen denominator, a single-pass reading moves on its own. Published measurement research reports 10–34% output variance from sampling alone (arXiv 2601.21339, 2026-01, preprint; accessed 2026-07-16). Rate movement smaller than that band is not evidence of anything. We collected the published record on this in a separate evidence review, which this series uses as its measurement-reliability anchor.
Which five asks fit in one procurement email?
1. Ask for the denominator definition in writing. "Share of what, exactly: all tracked prompts, appeared answers, or valid runs?" One sentence to answer; the answer becomes part of the contract's vocabulary.
2. Ask for n, k, and the interval. How many prompts (n), how many runs each (k), and the confidence interval next to the rate. A stated interval turns a marketing number into a falsifiable one.
3. Ask whether the query set is version-pinned. The before and the after must run the same frozen question list, and the version identifier should appear in the report.
4. Ask for the window and the baseline protocol. Engines, locale, repetition count, and dates, identical on both sides of the comparison, in writing.
5. Ask whether the number survives a re-run. Under the same protocol, within what band is the number expected to reproduce? A vendor who publishes intervals can answer this directly; the published evidence above tells you why a single-pass number cannot.
Where do we stand on the same asks?
We sell both measurement and execution, which is why every measurement ships with a receipt. The five asks above are the format our own reports are built to pass. Every paid CiteAngle report states the denominator for each metric. It prints n and the Wilson 95% confidence interval next to every rate that has one. It pins the query-set version, keeps measured observations apart from unmeasured ones, and seals each run with a hash you can check later. The outside figures we cite in public, including every figure on this page, are logged in our public claims registry with source and access date. The full protocol sits on our methodology page. If you want to see the format on your own brand before any contract, start with the free AI visibility check.
Here is what the repeat count buys, taken from our own measurement function and recomputed on 2026-07-17. At one run per prompt, a prompt that cited a brand every time and a prompt that never cited it come back with 95% intervals that still overlap: 20.7% to 100% against 0% to 79.3%. At seven runs they part company, 64.6% against 35.4%. The interval width falls from 79.3 points at one run to 35.4 at seven and 8.8 at forty, so the third and fourth asks above are not paperwork. They decide whether a before-and-after gap is a result or a coin toss.
Verify Before You Buy: reading AI-visibility claims
The article you are reading is Part 1. The series takes one reporting pattern per part and converts it into questions a buyer can ask in writing. It names practices, never vendors, and every quoted sales line labeled as a format example. Its measurement-reliability anchor is our published evidence review, A Single Run Is a Point Estimate, Not a Number. Next: Part 2, coverage counts versus measured counts (forthcoming). All research: CiteAngle Research, series index.
What an unfalsifiable number costs. Without a denominator a lift claim cannot be wrong, so the risk moves onto the buyer who repeats it. Every rate we publish carries n and an interval and is sealed with its conditions, which is the reason these four questions are ones we answer in public, where you can hold us to them.
Why bring the questions to us
These four questions are only useful if somebody answers them the same way twice. Ask a vendor in a meeting and you will get four confident sentences; ask again next quarter, after the panel has been quietly adjusted, and you will get four more. Nothing in that exchange leaves a record you can hold up.
So we answer them in public before you ask. The panel is arithmetic anyone can check — 50 buyer questions across 15 engines, 7 runs each, landing on 17 surfaces. Every rate carries its own denominator and a Wilson 95% interval. Readings the grid planned but could not take are printed as a count and kept in the denominator, since one that quietly shrinks manufactures improvement out of nothing. And the conditions behind each figure are sealed at measurement time, so the answer you get today is the answer you can still check next year. Put the same four questions to us in writing and the reply will match this page.
Take these denominator questions into your next vendor call
Ask them of every report on your desk, ours included. Our answers are public: the protocol in the methodology, every figure on the claims registry, the delivered format on samples.
Want these four denominators answered for your own brand? Contact us and we will scope it against your category.