In March 2025, I spent several days testing the AI visibility of 20 Belgian B2B agencies in ChatGPT, Perplexity and Gemini. Not to rank them, not to judge them — but to answer a simple question: when a Belgian prospect asks an AI which agency to choose for their project, who comes up?
The result surprised me, even though I was half-expecting it.
14 agencies out of 20 are completely absent from responses generated on queries directly related to their activity. Agencies sometimes well-positioned on Google, with polished sites, solid references and real expertise — invisible the moment a prospect turns to ChatGPT or Perplexity.
Here's the methodology, the complete results, and — above all — what distinguishes the 6 that make it through.
Methodology
Agency selection
I selected 20 B2B agencies active in French-speaking Belgium, covering several specialities:
- SEO and digital marketing agencies
- Automation and AI consultants
- Webflow studios and no-code development
- Digital transformation consulting firms
- HubSpot, CRM and RevOps experts
Agencies were chosen for their Google visibility (they appear in results for their main keywords) and their documented online activity. Names are anonymised in this article — the objective isn't to compare brands but to identify patterns.
Test protocol
For each agency, I tested 30 prompts covering 6 query types:
- Generic discovery: "Which [speciality] agencies would you recommend in Belgium?"
- Best provider: "What's the best agency for [specific need] in Brussels?"
- Direct comparison: "Compare [Agency X] and [Agency Y]"
- Alternatives: "What alternatives to [Agency X] for [need]?"
- By problem: "I need to automate my commercial processes, who can help in Belgium?"
- Extended local: "[speciality] agencies Brussels / Wallonia / French-speaking Belgium"
Each prompt was tested on ChatGPT (GPT-4), Perplexity and Gemini, in neutral browsing (private window, no conversation history). Critical tests were repeated at 24-hour intervals to account for response variability.
Scoring grid
Each result was scored on 4 dimensions:
- Presence: is the agency cited? (0/1)
- Position: first citation, second, third or more?
- Accuracy: does the AI correctly describe the agency's activity?
- Sentiment: is the recommendation positive, neutral or ambiguous?
Overall results
The reality is stark
Of the 20 agencies tested:
- 6 agencies have measurable presence in at least one AI engine
- 4 agencies appear consistently across multiple query types
- 2 agencies clearly dominate responses in their speciality
- 14 agencies are completely absent — zero mentions across all tested prompts
This isn't a marginal problem. 70% of tested agencies simply don't exist for AI engines, even on queries directly related to their core business.
Results by engine
The three engines don't behave the same way:
Perplexity is the most "equitable": it cites more different players, relies heavily on recent web sources, and gives a chance to agencies with recent external content. 5 of the 6 visible agencies appear on Perplexity.
Gemini favours players with a strong Google Business presence, recent reviews and consistency between the website and social profiles. It performs particularly well on geolocated local queries.
ChatGPT is the most conservative: it tends to cite the same players in a loop, often those most referenced on sources it "ingested" during training. Recent agencies or those with little media presence have very little chance of appearing there without specific effort.
What distinguishes the 6 visible agencies
Analysing the 6 agencies that stand out, 5 clear patterns emerge.
1. Solid and diverse external sources
This is the number one factor, by far. Visible agencies are cited on at least 4 credible third-party sources: sector directories (Clutch, GoodFirms, Sortlist, Digital Agency Network), press articles or B2B media, testimonials on independent platforms, mentions in studies or comparisons.
The most visible agency in the panel is referenced on 8 external sources identified in AI responses. The least visible: zero identifiable external sources in the responses.
AIs can only recommend what they "know". And they know what third-party sources tell them.
2. Clear and repeated positioning
Agencies that AIs describe correctly have one thing in common: their speciality is formulated the same way everywhere — on the site, on LinkedIn, in articles, in directories. The AI can make a coherent synthesis.
Badly described (or uncited) agencies often have vague, generalist or contradictory positioning between sources. The AI can't resolve this ambiguity — it prefers not to mention the agency rather than describe it incorrectly.
3. Visible and recent client reviews
All visible agencies have recent Google reviews (less than 6 months old) and a rating above 4.5. Some have detailed testimonials on their website or LinkedIn, giving AIs content to justify a recommendation.
This point is often underestimated by B2B agencies who consider reviews "a B2C thing". In reality, AI engines use them as a legitimacy signal.
4. Indexable educational content
3 of the 6 visible agencies regularly publish articles, guides or analyses. This content is indexed, cited by other sites, and feeds the sources that AIs use for their responses.
This isn't content to "do SEO" in the classic sense — it's content that creates external traces that AIs can mobilise to justify a recommendation.
5. Consistency across all digital presences
The last factor: consistency. Visible agencies have the same name, the same description, the same speciality on their site, LinkedIn profile, Google Business listing, directories and third-party platforms. This consistency allows AIs to build a reliable representation of the agency.
Invisible agencies often have inconsistencies: a slightly different name across platforms, a description that changes, a speciality absent from some profiles.
The most frequent errors
Beyond pure absences, I observed several types of errors in AI responses concerning the tested agencies.
Positioning error: the AI describes the agency as generalist when it has a strong niche. This happens when the specialisation is mentioned at the bottom of the page or in a submenu, but not in the main elements that AIs scan.
Confusion with a competitor: in 2 cases, the AI attributed characteristics of one agency to another with a similar name. Nominal differentiation matters more than you'd think.
Outdated information: Perplexity cited 2022 information for an agency that had since changed its positioning. AIs don't know your offer has evolved if you haven't documented it on recent sources.
Absence on local queries: some agencies appear on generic queries but disappear when you add "in Brussels" or "in Belgium". This suggests a lack of geographic signals (Google Business, local mentions, clearly stated address).
What this means concretely
If you're a Belgian B2B agency and you've never tested your AI visibility, there's a good chance you're among the 70%.
This isn't a catastrophe today — the majority of B2B purchasing decisions don't yet pass exclusively through AI engines. But behaviour is evolving rapidly, particularly among digital profiles and young executives.
The window to position early is open. Agencies building their AI presence now will have a structural lead in 18 to 24 months.
Priority actions, in order of impact:
- Check your external sources: are you listed on Clutch, Sortlist, or directories in your speciality?
- Clarify your positioning: formulate your speciality in one clear sentence and repeat it everywhere
- Activate recent reviews: ask your current clients to leave a detailed Google review
- Create dedicated service pages with an integrated FAQ answering questions that prospects ask AIs
- Publish at least 2 quantified client cases on your site and LinkedIn
Testing your own visibility
If you want to know where you stand, the first step is simple: open ChatGPT and Perplexity and ask the questions your prospects ask. "Which [your speciality] agency to choose in Belgium?" — do you appear?
To go further, our DIY guide gives a structured method with 50+ prompts and a scoring grid to analyse your results correctly.
If you prefer a complete analysis with competitive benchmark, the AI diagnostic covers exactly what I did in this study — applied to your specific company, delivered within 48 hours.
June 2026 update: we re-ran these same tests a year later. See what changed →
Full methodology available on request. Agency names are voluntarily anonymised. Results represent a snapshot from May 2025 — AI responses evolve over time.