Beneloo Observatory
the observatory of Belgian SME visibility in ChatGPT, Claude, Gemini and Perplexity: 3 campaigns, 1,668 companies, 96 to 98% invisible.
The product
Measuring, with numbers, who exists for the AIs.
- Role
- Design & data
- Stack
- Python · Postgres · immutable campaign pipeline
- Status
- Live
The observatory of Belgian SME visibility in ChatGPT, Claude, Gemini and Perplexity: 3 campaigns, 1,668 companies, 96 to 98% invisible.
The Observatory queries four AI models with frozen prompts over exhaustive samples of companies: 488 accountants in Namur, 382 plumbers in Namur, 798 accountants in Walloon Brabant. Every mention is validated by hand (correct, network brand, hallucination). Result: 96.9%, 98.4% and 96.6% invisibility; ChatGPT and Claude cite no local firm at all, Gemini hallucinates up to 52% of its answers.
01The problem
Everyone talks about AI visibility, nobody measures it.
When someone asks ChatGPT or Perplexity for an accountant or a plumber nearby, who gets cited? Talk is plentiful, numbers are missing. And without measurement you cannot tell an SME whether it has a problem, nor whether it is getting fixed.
It needed a reproducible protocol: same questions, same models, exhaustive samples, human validation of every name cited.
02What I built
A frozen campaign pipeline and name-by-name validation.
The Observatory queries four models (ChatGPT, Claude, Gemini, Perplexity) with frozen prompts over entire populations: 488 accountants in Namur, 382 plumbers in Namur, 798 accountants in Walloon Brabant. Every entity cited is matched against the official register and classified: correct, network brand, hallucination.
Result: 96.9%, 98.4% and 96.6% invisibility. ChatGPT and Claude cite no local firm at all; Gemini hallucinates up to 52% of its answers. Total API cost of the three campaigns: under 5 euros.
03Where it stands
Three campaigns published internally, method documented.
Aggregates are anonymised and publishable; the validation protocol, limits and biases are written down. An editorial wall strictly separates the public observatory from Beneloo consulting.
Next steps: new sectors, a published two-sector contrast, and the Beneloo Score applying this measurement to one company at a time.