By JayOctober 2026
Reading the AI Visibility Numbers
If you run a salon or a local contracting business, it is tempting to read 6.81 days as the wait before AI starts naming your new page. That is not what the number measures. Several headline numbers in LinkedIn's Unlocking AI Search Visibility guide measure something narrower than they sound, which is why your own baseline matters more than anyone's statistic.
The guide's advice is sound, and we build on it in AI Visibility Takes Weeks. Start With a Baseline. It also makes a related point itself. On page 33 it quotes Cassie Dell:
Don't fall for universal 'best practices'. What works for one brand might not work for yours.
What the 6.81 days covers
The figure comes from Profound's agent logs (as reported in the guide): about 900 new pages cited by ChatGPT or Claude agents over a 60-day window, March to May 2026. The guide labels 6.81 days the median and prints the 75th and 90th percentiles beside it (three in four pages within 18.68 days, nine in ten within 37.10), which is what makes a careful reading possible. By our measurement of the chart, the plotted median sits below the halfway line (about 45%, not 50%), so treat 6.81 days as approximate.
Four things limit what the number supports:
- Cited pages only. Pages that were never cited are not in the data. Vincent Orleck's newsletter Clocking AI Citations puts it this way: "that missing denominator is the whole ballgame."
- A 60-day window. A page first cited on day 70 is outside the data.
- Two products in one curve. ChatGPT and Claude are blended into one line.
- An unstated definition. The guide defines a citation on page 30 as an answer that credits your domain or profile with a link, but the Profound chart does not say it used that definition. It may count an agent fetching a page, not an answer showing your page to a reader.
What the number does support is modest: for pages that were cited, the guide's median wait was under a week, and one in ten waited longer than 37 days.
The number describes pages that made it. It cannot say how likely yours is to.
The studies differ too
The studies the guide relies on give different numbers about LinkedIn, because each one measured something narrower. Two examples, from the guide and from Meltwater's study:
| What was measured | The guide | Meltwater |
|---|---|---|
| Where LinkedIn ranks | #1 for professional queries (Profound) | #2 overall, behind YouTube |
| Posts versus articles | Articles about 60%, posts about 40% (LinkedIn internal data) | Text posts 72%, articles 12%, video 11% |
Figures as labelled in the guide and in Meltwater's study.
The guide's #1 is Profound's figure for professional queries; Meltwater's #2 is its overall rank across the B2B prompts it tested. Meltwater also counts video as a third type, which the guide's split does not. None of these numbers describes your market.
Every company whose figures appear here sells something related: Profound, Meltwater, Semrush and LinkedIn, and so does addAI.dev. That is a reason to ask what each figure was built to show, not to dismiss it.
Two numbers that read bigger than they are
The first is similarity. The guide says AI answers "tend to quote or paraphrase you closely." The Semrush score behind that is 0.57 to 0.60 on a 0 to 1 scale, where 1 means nearly identical phrasing. Paraphrase is a fair reading of a score about shared meaning. The word "closely" is the stretch: 0.57 to 0.60 is mid-scale, modestly above the 0.53 to 0.54 Semrush reports for Reddit from its earlier studies.
The second is the 26% growth in LinkedIn's citation share. Meltwater's wording is that it "grew 26% across tracked models over the 4 week research project." On a share of about half a percent, that can only be relative growth (26% of about half a percent is a small absolute change), so it shows direction and not size.
Before repeating a number, ask for its base and its scale.
Run your own numbers
The fix is your own baseline: a fixed set of buyer questions in your own words, asked the same way across several AI products and repeated. AI Visibility Takes Weeks. Start With a Baseline. shows how to set one up.
In our pilot, one Toronto salon was named in 9 of 9 answers from each of two products to questions that did not mention it by name. A third product named it in 1 of 9 and a fourth in 3 of 9. Those questions were written around the salon's own services, and the result is a one-time snapshot.
Neither one check nor one published statistic can stand in for your own baseline.
The free scanner is a quick first look at whether AI crawlers (the programs that read websites for AI products) can fetch and read your site. addAI.dev asks for the denominator and the date behind every number, and plans each client's work around their own baseline.