First-party data

The proof. Measured, dated, verifiable.

An agency shows you client logos. We show you data. These three cases come from sites we own or operate: we tested everything on ourselves before selling any of it.

Three measured cases, at a glance A page ranking first with the featured snippet against a domain rating 67 brand site, holding one backlink and a URL rating of zero, ranking for 81 keywords on a query searched 2,500 times a month. A site cited 653 times inside ChatGPT answers. A practice site that lost about half its organic traffic and whose rebuilt pages regain positions in two to six weeks. Content beats authority #1 Position held, featured snippet inc… 1 Backlink, URL rating zero 81 Keywords ranked by that page Cited by assistants 653 Citations measured in ChatGPT answe… URL Verified one URL at a time 2 Assistants citing the site Repair, not only construction 50 % Of organic traffic lost in six mont… 8 Weeks in the rebuild plan 2-6 Weeks for a rebuilt page to move
Independent measurement tools, dashboards shown live on request

02

653 citations inside ChatGPT

653
Citations measured in ChatGPT answers
URL
Verified one URL at a time, not by domain average
2
Assistants citing the site, including Copilot

The context

Our English-language dental media for the US market. The objective: be the source AI assistants cite when someone asks a dental question.

The result

653 citations measured inside ChatGPT answers, verified URL by URL with independent tools. The site is also cited by Copilot.

The method, in plain terms

  1. Direct answer structure: every page opens with the answer, and assistants lift it as written.
  2. FAQ sections marked up with FAQPage schema on every piece of content.
  3. An anti-regression protocol: on a cited page, every edit is additive, never destructive. We learned that, measured it, and documented it after a real incident that we resolved and used as a laboratory.
  4. Continuous monitoring: every page is tracked on its own citation count, page by page, never as a domain average.

What it proves. AI visibility is not declared, it is engineered. And it is measurable. In France, 38.1 percent of the population was already using an AI assistant in September 2025 (Mediametrie). This channel is no longer optional.

03

The recovery of a Paris practice site

50%
Of organic traffic lost in six months
8
Weeks in the rebuild plan
2 to 6
Weeks for a rebuilt page to regain positions

The context

The site of an orthodontic practice we operate had lost roughly half its organic traffic between January and July 2026: publishing had stopped, freshness had decayed, competitors were active, and AI Overviews were eroding click-through rate.

The result

An eight-week plan: six existing pages rebuilt, two new pieces of content, a fixed weekly publication rhythm, internal linking rebuilt, FAQPage schema applied systematically. Production orchestrated end to end by our system, from brief to delivery.

The first effects

  1. Rebuilt pages regain positions in 2 to 6 weeks, where new content takes 8 to 12. That gap is the reason for the refresh-to-creation ratio we chose.
  2. The site is in active recovery on its main local query, worth 1,800 searches a month.

What it proves. The method works in repair, not only in construction. And a practice does not need the budget of a dental group to apply it: it needs consistency and method.

Verifiability

These figures come from independent measurement tools (Ahrefs) and are dated August 2026. In a meeting we show the dashboards live, positions and citations included. No polished screenshots: the data, on screen, on the day.

Every number published anywhere on this site traces back to one of these measurements. When we have no first-party data on a subject, we say so rather than borrow someone else's statistic.

Request a free audit of your site

The methods behind the numbers

Each of these cases is written up in full inside one of our guides.