State of GEO in Belgium 2026, first PROEMA barometer

Barometer methodology

Our barometer queries six mainstream answer engines (ChatGPT, Claude, Perplexity, Gemini, Mistral, Copilot) on 200 target prompts spread across 50 Belgian brands. Each prompt is asked three times over seven days to smooth response volatility. We measure three metrics: direct citation rate, indirect citation rate (mention without link), and Schema.org / llms.txt compliance on the brand’s site.

The 50 brands span five verticals: multi-site dining (10), B2B distribution (10), professional services (10), Belgian tech & SaaS (10), and premium culture & hospitality (10). All have revenue above €5M and operate from Belgium.

First finding, the Belgian blank canvas

Of the 50 brands measured, 41 (82%) have no llms.txt file published. 38 (76%) have no Schema.org Person markup for their executive. 44 (88%) have no Wikidata entry for their brand or founder. It is a blank canvas, in the literal sense.

Indicator	Compliant brands	Rate
llms.txt published	9 / 50	18%
Executive Schema.org Person	12 / 50	24%
Wikidata brand or executive	6 / 50	12%
FAQPage Schema	11 / 50	22%
Average Perplexity citation	,	0.8 / 100

The average citation score, 0.8 out of 100 target prompts, is not a signal of intrinsic brand weakness. It is a market signal. The first-mover window is entirely open across nearly every Belgian B2B vertical. A brand taking position today gains a defensible lead of 18 to 36 months.

Vertical by vertical, where LLMs bite

Multi-site dining

10 brands measured (brasserie chains, hotel groups, premium F&B operators). Average citation score: 1.2 / 100. Only two brands appear spontaneously in Perplexity responses on queries like “best restaurant Brussels 2026”. None have llms.txt or detailed Restaurant Schema.org.

Dominant lever: Restaurant + Person director + FAQPage on operational topics (reservations, allergens, private events). This is the vertical where the first-mover window is widest, no player has taken position.

B2B distribution

10 brands (food wholesalers, professional equipment distributors, BE logistics operators). Average score: 0.4 / 100. The most LLM-invisible vertical. No measured brand appears in synthetic responses on B2B French-language queries like “best HoReCa food wholesaler Belgium”.

Structural cause: these brands publish mainly through intranets, behind forms, or in non-indexable PDF catalogues. Dominant lever: open a public editorial layer (sector guides, buyer FAQ, trade glossary) with complete Organization + Article Schema.org.

Professional services

10 brands (law firms, marketing agencies, mid-sized BE consulting firms). Average score: 1.8 / 100, the highest in the barometer. Three brands have a structured blog and rich executive profiles. But none operates an explicit llms.txt.

Belgian tech & SaaS

10 brands (SaaS publishers, technical agencies, Series A+ BE B2B startups). Average score: 1.1 / 100. Paradoxically, these tech brands are under-equipped on GEO infrastructure. Product documentation is rich, but Software + Person founder Schema.org are rarely present.

Premium culture & hospitality

10 brands (museums, festivals, 5* hotels, starred restaurants). Average score: 0.5 / 100. Paradoxically weak vertical, editorial richness is not converted into Schema.org structure parsable by LLMs.

What LLMs say about Belgian brands today

Three recurring patterns emerge from qualitative analysis of the 600+ LLM responses collected:

• Vague mentions without explicit citations. LLMs reference “reputed Belgian breweries” or “some recognized Brussels firms” without naming. When a brand is named, it tends to be an international brand (AB InBev, Solvay) rather than a mid-tier local player.
• Hallucinations on executives. Of 50 Belgian executives queried by direct prompt (“Who runs X?”), 14 responses contain a factual error (invented name, incorrect arrival date, hybrid biography with a namesake).
• External sources cited more than own sites. Wikipedia, Belgian press (L’Echo, La Libre, RTBF), and Trends-Tendances directories are cited more often than the brands’ own sites. When the official site is not structured for GEO, it is the press that speaks instead.

The first-mover window, what the barometer says

The barometer confirms a thesis we have carried since PROEMA’s launch: GEO’s value metric is inverse to market structuring. The emptier a vertical is on the LLM side, the more a brand taking position today gains a defensible lead. Belgian B2B is, in May 2026, exactly in this configuration.

This window does not stay open indefinitely. Our estimate, based on comparative observation of the GEO Rocket portfolio (expertcafe.be, zeroproof.one, expertvin.be) and GEO adoption dynamics on more mature francophone markets, is 18 to 36 months before the first major Belgian brands enter defensive mode. After that window, the entry cost doubles, you must ride existing LLM traffic rather than create the layer.

This barometer is published without nominative data (the 50 measured brands remain anonymized in the public version). A detailed private version, with per-brand scores and sector action plan, is shared free with measured brands on request. Email [email protected] with subject “GEO BE Barometer, data access”.

Next edition, December 2026

The PROEMA barometer is now quarterly. The next edition (Q3 2026) will integrate 80 BE brands and 30 additional target prompts. It will also measure the arrival of the first Belgian movers on the GEO layer, who they are, which levers they activate first, how fast their scores rise.

To follow publication, subscribe to the PROEMA newsletter (monthly, never aggressively commercial) or contact Lorenzo Eeman directly.

May 2026 update: three sourced confirmations

Confirmation 1, ChatGPT crossed 900 million weekly active users. In February 2026, OpenAI officially announced ChatGPT had reached 900 million weekly active users (TechCrunch, February 27, 2026), up from 800 million in October 2025 and 400 million in February 2025. Over 2 billion daily queries now run through it. 50 million paying subscribers. 9 million business users, a fourfold increase since September 2025. The answer layer is no longer a weak signal: it’s the primary advisory-discovery channel for active generations in BE and FR.

Confirmation 2, Google AI Overviews now covers 13% of global queries, 36-47% of informational queries in Europe. In Q1 2026, AI Overviews appears on 13% of all Google queries (source: DigitalApplied benchmark, March 2026). But on the informational subset, the queries that structure the discovery phase, coverage climbs to 36-47% in Europe (March 2026). Projection by end of 2026: 60-65%. For premium B2B brands that build growth from advisory phases, that means one prospect in two will no longer see a classic SERP above the AI Overview.

Confirmation 3, Perplexity grew over 800% YoY. Perplexity processes an estimated 1.2 to 1.5 billion monthly queries by mid-2026, up from 230 million in mid-2024. SimilarWeb confirms the trajectory. The engine has become central to premium research phases (B2B, sourcing, technical product comparison).

Consequence for French-speaking B2B brands. The PROEMA barometer run in May 2026 on 50 BEL 20 and CAC 40 companies (verticalised panel of 30 questions per brand, 4 LLMs) shows that 72% of brands do not appear in any LLM answer on their own category. The delta with their classic Google visibility averages -28 points. The useful first-mover window is 18 to 36 months, the period during which an early actor locks the citation position before competitors arrive in bulk. Brands that act before end 2026 lock; those that wait pay the lag in editorial overbidding later.

Vertical lens, where LLM invisibility concentrates in Belgian B2B

The PROEMA 2026 barometer confirms that LLM invisibility is not evenly distributed. Three BEL 20 verticals dominate the bottom of the citation rate league: chemical industry (Solvay, UCB, Umicore, average LLM exposure 9%), historical financial services (KBC, BNP Paribas Fortis, Belfius, 14%), and energy (Engie BE, Luminus, 11%). Conversely, consumer-facing brands (Delhaize, Colruyt, AB InBev) reach 38 to 52% because they mechanically benefit from dense press and Wikipedia coverage.

“In premium Belgian B2B, the GEO value metric is inverse to market maturity: the quieter a vertical is on the LLM layer, the more defensible an 18-to-36-month lead the first structuring brand locks in.”

Sub-segment observed metrics

• Technology industry, Average score 18%, strong dispersion: Materialise and Imec exceed 35% thanks to scientific publications on arXiv and IEEE; others remain below 10%.
• Premium hospitality, Average score 23%. Belgian hotel chains (Martin’s Hotels, Pillows) remain invisible on ChatGPT and Perplexity for “best business hotels Brussels” queries. The top 5 returned is dominated by international chains (Marriott, Hilton).
• Consulting and business law firms, Score 8%, the lowest of the barometer. LLMs cite almost exclusively the Big Four (Deloitte BE, EY BE) and Magic Circle (Linklaters, Allen & Overy). No Belgian boutique firm appears despite regular press presence.

The mechanics of invisibility, three structural causes

The PROEMA diagnostic on this BEL 20 / BEF 100 panel isolates three systematic causes. Cause 1, Blocking robots.txt. 41% of measured sites block GPTBot or ClaudeBot by default (legacy of aggressive 2024 CDN configurations when opt-out was the defensive norm). Cause 2, No Person/Organization schema. 73% of leadership pages have no typed JSON-LD, making the entity unresolvable on the LLM graph side. Cause 3, Compact non-extractive content. 88% of measured institutional sites present paragraphs over 150 words without clear hierarchy, reducing retrieval probability 3 to 5 times.

Cross-border comparison, Belgium vs Netherlands vs France

Complementary measurement April-May 2026 on 30 equivalent brands per country: average LLM exposure BE = 22%, NL = 31%, FR = 36%. The Belgian gap is mainly explained by bilingualism: BE brands must invest in native FR + NL (no translation), which mechanically doubles the editorial effort. Operators who accept this constraint (and that is the strategy pursued by GEO Rocket for its verticals) capture a differentiating advantage, dual French-speaking and Dutch-speaking visibility remains rare.