Reliability of results surveys: audit of October 19, 2026
Audit of the 21 pages from the controlled series of October 18, 2026 (same query "musée du Louvre site officiel", hl=fr, gl=fr, fixed US exit, Camoufox), supplemented on October 19 by resolving /goto links and by the pages obtained with Chrome launched normally.
Receiving a results page does not certify a survey
After separating main results, sitelinks and modules, the Louvre homepage appears first on 3 pages, seventh on 17, and is not identified among the main results of one page. The five sitelinks present on each of the three pages where the homepage is first are not counted as independent results: they do not explain the rank 7 observed elsewhere.
The nine "NID retained / NID deleted" pairs give six identical main orders and three disjoint sets, based on the displayed titles and domains. The presentation varies more: seven pairs differ in the position of the AI Overview or of "People also ask". A link presented as the homepage remains visible in the AI Overview even when the homepage is not identified in the main list. An organic rank therefore does not sum up all visibility on the page.
The /goto links
The 180 main links go through /goto redirects. On October 19, between 11:06 and 11:09 UTC, 17 links were resolved with a plain HTTP request; four were followed to an HTTP 200 response from the target site.
| Page / rank | Displayed title | Destination reached | HTTP |
|---|---|---|---|
| 06 / 1 | Site officiel du musée du Louvre | https://www.louvre.fr/ | 200 |
| 07 / 7 | Musée du Louvre Official Website | https://www.louvre.fr/en/ | 200 |
| 02 / 7 | Musée du Louvre Official Website | https://www.louvre.fr/en/ | 200 |
| 13 / 1 | Ecole du Louvre: École du Louvre - Accueil | https://www.ecoledulouvre.fr/fr | 200 |
In pair 06/07, the first result leads to the French homepage and the seventh to the English homepage: these are two distinct URLs, and the ranking does not describe the same page moving. The 15 targets of this pair are identified and have no URL in common. The other eight pairs remain incompletely verified, and a resolution carried out on October 19 does not recreate the exact state of the links at the time of the captures.
The Validate call comes after the page
In the branches that retain NID, the six pages where the homepage is seventh record a Validate call, unlike the three pages where it is first, within the observed window. These six calls start about 8 to 10 seconds after the results page has been fully received. On all 21 pages, the main order received matches the final order. The later call therefore cannot have produced the ranking already received. Some pages obtained after challenge also show the homepage low without Validate.
An ordinary Chrome receives the same type of list
On October 19, Chrome without an automation declaration ran the same search from the VPS's native exit, on a fresh profile (12 pages, last two series of the browser checks). The Louvre homepage is never the first visible result there:
| First visible result | Pages |
|---|---|
| Le musée du Louvre en bref (Louvre press area) | 6 |
| Premiers pas au Louvre (Louvre resource) | 4 |
| Vie du musée (Louvre news) | 1 |
| École du Louvre | 1 |
The AI Overview links to louvre.fr every time. This type of list is therefore not reserved for browsers that declare themselves automated. The exit and the day differ from the October 18 series, and the screenshots only show the top of the page.
What remains open
These observations suggest an association between session state, navigation path and result variants, without isolating the cause. The corpus comprises a single query, three source profiles and a US exit despite the French parameters. It offers no contemporaneous human reference. Neither a deliberate degradation of results nor its absence is established.
For a usable survey, the access obtained, the nature of the blocks, the visibility of modules, the actual destination of links, the session state and the time of the checks must be documented separately, then compared with a human survey made at the same time from the same exit.