The problem wasn't visible. That's often the case with AI engines.

The Agency — a B2B consulting firm based in French-speaking Belgium, active for 12 years in supporting industrial SMEs — had everything needed to be credible online. A polished website, published case studies, regular presence on LinkedIn, and solid Google positions on its main keywords. No reason to worry.

Except that its prospects had changed their search habits.

The context: a silent evolution

Between 2024 and 2025, B2B buyer search behaviour changed faster than anyone was measuring. According to Similarweb and SparkToro data, traffic from LLMs to professional sites reached 10 to 12% of global visits in some B2B sectors in Western Europe — a figure that was near zero 18 months earlier. Over the same period, ChatGPT grew to more than 200 million weekly active users worldwide, with particularly strong penetration among decision-maker profiles (executives, managers, purchasing managers).

Concretely: when a marketing manager looks for a consulting agency for a transformation project, they no longer just run a Google search. They ask ChatGPT or Gemini a question. And what the AI tells them directly influences what comes next.

The Agency had no visibility on this channel. Nobody had flagged it.

What we found when testing

The diagnostic covered 44 prompts, spread across 6 query types, tested on ChatGPT, Gemini, Copilot and Claude. The prompts were custom-built from the Agency's actual positioning and targets.

The overall result: a score of 31 out of 100.

Out of 44 prompts tested, the Agency was cited in 6 responses. Partially present (mentioned without being recommended) in 4 others. Absent in the remaining 34.

The most problematic queries were the most commercially important — "best provider" and "direct comparison" type queries, where the prospect is in active selection mode.

"which B2B transformation consulting agency in Belgium for a 50-person SME?"

On this type of prompt, three competitors appeared systematically. The Agency appeared in no response, on any engine.

Why the AIs didn't cite it

After analysis, three main causes were identified.

The positioning was too generic on external sources. AI engines don't rely solely on the company's website. They rely on what they find elsewhere: press articles, sector directories, comparison sites, mentions in third-party publications. The Agency had an excellent website but little documented external presence. Result: the AIs couldn't find enough consistent signals to recommend it with confidence.

The specialisation wasn't legible enough. The Agency positioned itself on "SME consulting", an angle too broad for an AI to capture as a differentiated expertise. Its best-ranked competitors had sharper positioning ("digital transformation for manufacturing industry", "B2B complex sales process optimisation"). The AI associated these specific formulations with precise queries. The Agency's generic positioning corresponded to nothing.

No third-party source mentioned it in the right context. On local queries ("French-speaking Belgian consulting agency"), Gemini picked up some partial signals. But on thematic queries, no sector publication, no comparison site, no external content cited it with the right keywords. For an AI, a company that doesn't exist in third-party sources is a company whose relevance cannot be guaranteed.

What was recommended

The report delivered included 10 prioritised recommendations. The three main areas:

1. Clarify positioning on key pages. Rewrite the homepage and "About" page so that positioning is immediately legible to an AI: precise sector, client type, problems solved, geographic area. No vague marketing reformulations — factual and specific formulations.

2. Build a consistent external presence. Identify 5 to 8 Belgian B2B sector directories and publications where the Agency could obtain a listing or mention. Not classic link building — documented presence on sources that AIs consider reliable.

3. Publish indexable content on client problems. AIs recommend sources that answer the questions prospects ask. Publishing 3 to 4 detailed articles on problems specific to industrial SMEs, with formulations that match real queries, increases the probability of being cited on those topics.

What this changes concretely

The diagnostic isn't an end in itself. What it produces is a list of actions ranked by impact and effort — not a theoretical audit, but a work plan.

For the Agency, the quick wins identified were accessible in less than a month: rewrite two pages, create listings on three directories, fix metadata. Actions any team can execute without particular technical skills.

Results on AI engines aren't immediate. LLMs take a few weeks to several months to integrate new sources into their responses. But the window to act is open now, before competitors consolidate their position.

This case is an anonymised illustration based on our diagnostic method. The sector data cited (LLM traffic, ChatGPT adoption) comes from public sources available in 2025-2026.

Want to know where your business stands? Order your diagnostic