HENUSSE Making your company readable by AI. Outside, and in.

The HENUSSE Barometer · Citation share · Reading recorded on 24 August 2026

We measure HENUSSE.
Here is the reading.

Before measuring your company, we measure our own. On , we asked ChatGPT, Claude and Gemini 8 questions, twice each. A reading is one question put once to one engine, and the Barometer is all the readings of a single date: all 48 planned readings were completed. Across the five questions in which a buyer is looking for an AI search optimisation provider, HENUSSE was selected in 0 answer out of 30. The rest of the time, the answer engines cite other providers, or no one. This is a dated finding, not a prediction.

When a buyer looks for a provider: 0 answer out of 30 on 24 August 2026.

ChatGPT: 0 out of 10.

Claude: 0 out of 10.

Gemini: 0 out of 10.

These 30 answers cover five questions in which a company selects an AI search optimisation provider. Verbatim mentions of our name across all 48 readings: 0, of which 0 sit in the method questions and are never counted as citations: a mention there only refers to our local context, or notes an absence, without recommending us. When in doubt, we count against ourselves.

The dated readings

Eight questions. The aggregates for each question, stated plainly.

Question asked Type and readings Who is cited
“Which agency should a company choose to be cited by ChatGPT and AI systems in France?”
(Question asked in French: « Quelle agence choisir pour être cité par ChatGPT et les IA en France ? »)
Provider selection · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: yes, other providers
A person: yes
“Which AEO or GEO agency in Brittany for an SME?”
(Question asked in French: « Quelle agence AEO ou GEO en Bretagne pour une PME ? »)
Provider selection · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: yes, other providers
A person: yes
“How can a Breton manufacturing SME be recommended by AI answer engines?”
(Question asked in French: « Comment une PME industrielle bretonne peut-elle être recommandée par les moteurs de réponse IA ? »)
Method question · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: not selected as a provider
A person: no
“Who should a company consult in France to optimise its website for ChatGPT, Claude and Gemini?”
(Question asked in French: « Qui consulter pour optimiser son site pour ChatGPT, Claude et Gemini en France ? »)
Provider selection · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: yes, other providers
A person: yes
“Best AI search optimisation agency in France?”
(Question asked in French: « Meilleure agence de référencement pour intelligence artificielle en France ? »)
Provider selection · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: yes, other providers
A person: yes as a source
“How can an SME integrate AI into its processes with a consultant?”
(Question asked in French: « Comment intégrer l’IA dans les processus d’une PME avec un consultant ? »)
Operational Method · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: not recommended
A person: no
“Which expert can help a French SME with the Operational Method and AI?”
(Question asked in French: « Quel expert pour la méthode opérationnelle et l’IA dans une PME française ? »)
Operational Method · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: yes, generically
A person: yes
“Agency for creating authority content cited by generative AI?”
(Question asked in French: « Agence pour créer du contenu d’autorité cité par les IA génératives ? »)
Provider selection · 6 readings out of 6
2 out of 2 for ChatGPT, Claude and Gemini
Us: 0 out of 6
Another company: yes, other providers
A person: yes
HENUSSE readings recorded on 24 August 2026, trace weekly-2026-W35

Each row in the table publishes the aggregates from 6 dated readings, making 48 in total. The raw answers, which name third parties, are retained without being reproduced online.

How to read this result

What 0 out of 30 means. Nothing more.

01

A stated scope

The 0 out of 30 applies to the five provider-selection questions, asked twice each of three answer engines. The other 18 readings appear in the table, but they are not counted here.

At the time of this reading, henusse.com is indexed. The Barometer will record what follows, one reading at a time.

02

Absences shown

HENUSSE is absent from 30 answers out of 30 within this scope. ChatGPT does not select us once out of 10, and neither does Gemini. These zero results remain displayed as they are.

03

Other providers cited

For the selection questions, the answer engines mainly cite other companies and sometimes people. We count these categories without naming a competitor or creating a ranking.

04

No promise

The next reading may rise, remain unchanged or fall back to zero. A dated measurement does not determine subsequent answers, for us or for you.

The Barometer is a dated reading. The next one will be published on this page, dated like this one.

The same protocol elsewhere

The same protocol on a public website.

realfrench.co is a public, indexed website designed, written and operated by HENUSSE. The Barometer protocol is applied to it in exactly the same way. Reading recorded on 12 August 2026:

  • “One-on-one French immersion Brittany”: realfrench.co cited 7 times out of 8.
  • “French immersion for diplomats”: realfrench.co cited 4 times out of 8.
  • “Why can’t I understand spoken French?”: realfrench.co cited 0 times out of 8. This question is asked well before any purchase, and we are still working on it; the zero remains displayed.

We apply exactly the same rules here: fixed questions, retained answers, dated fractions. The difference between the two readings comes down to one word: exposure.

The experiment log

  • 31 July 2026: initial HENUSSE reading, 1 out of 30, with the website deliberately excluded from indexes.
  • 6 August 2026: noindex lifted. The site is open to indexing; subsequent readings are read with this new exposure.
  • 11 August 2026: 0 out of 30, 48 readings, site indexed, trace weekly-2026-W33.
  • 24 August 2026: 0 out of 30, 48 readings, site indexed, trace weekly-2026-W35.
  • The next entries will be written here and dated: the opening to answer engines, then each reading.

Correction of 15 August 2026: the 11 August reading was wrongly flagged “site deliberately excluded from indexes”, while the site had been indexed since 6 August. The exposure flag is corrected; the fractions themselves do not change.

How we measure

Measure first, then comment.

This Barometer brings together the 48 completed readings under reference weekly-2026-W35, the trace that allows the test to be repeated under the same conditions: one reading for each question, engine and run. Each retains the question asked, the engine, the run number, the complete answer, the pages used, the date and time, any mention of HENUSSE and the response time.

All 48 readings have the same reference and a readable answer. There is no failure to deduct from this Barometer. A failed reading would remain visible and would never add a citation.

The 0 out of 30 is calculated across the five questions in which a company selects a provider. After review, no mention found in the method questions was excluded. The scope, exclusions and zero results accompany every published aggregate.

The public page shows the aggregates. The complete trace is retained. The client file contains this trace, the aggregates and the summary. The charter records the test conditions so that it can be repeated under the same conditions.

Read the full Honest Measurement Charter

Another surface, another cadence

Google’s AI Overviews, read every quarter

Since 26 August 2026 we also read another surface: Google’s AI Overviews, the generated answers shown at the top of the results page. These readings never enter the Barometer count above, which stays at three engines, not one more. That reading is published separately and retaken every quarter. The exact scope is explained on the panel scope page.

Reading of 26 August 2026, across the 25 frozen panel questions, one pass per question, in a real browser, neutral signed-out profile, from Côtes-d’Armor. Questions that trigger an AI Overview: 25 of 25. henusse.com citations inside an Overview: 2 of 25. Present elsewhere on the results page: 4. One pass per question, so no percentage: the reading shows what Google displayed that day. Every screenshot and every page is preserved, question by question, trace aio-2026-08-26.

Screenshot dated 26 August 2026: on the question of whether an AI citation diagnostic can conclude not to proceed, Google's AI Overview answers that yes, a HENUSSE AI Citation Diagnostic can conclude exactly that, citing henusse.com as its source.
Capture from the 26 August 2026 reading: the AI Overview cites henusse.com on the “do not proceed” verdict question.

Surface added to our readings on 26 August 2026, as a dated instrument change. Next reading: November 2026, alongside the quarterly reading.

Evidence on our own house first

HENUSSE measures itself.

We apply to our own name the protocol we offer our clients. Today, that means publishing 0 citations out of 30, and therefore 30 absences, together with the aggregates from all 48 readings in this Barometer. The Barometer will remain dated and cover every reading, never a selection of only the positive results. If you are wondering what the answer engines say about your company, this is exactly the type of reading we produce for you.

WhatsApp