Key takeaways
- AI assistants cite a handful of companies—being crawlable isn't the same as being citable.
- AI referral traffic is smaller but converts better; optimise to be named, not just clicked.
- Coherence, real reviews and credible third-party mentions are the new discoverability currency.
- Write core pages for machine extraction: state plainly who you are, for whom, and where.
- More visibility means more risk—harden security and make deliberate infrastructure choices.
Ask ChatGPT, Perplexity or Google's AI Overviews to recommend a supplier, a consultancy or a local service provider, and something uncomfortable happens: a small, repeating set of names surfaces, while thousands of perfectly competent entreprises never appear at all. The machine isn't being unfair. It simply doesn't understand who you are, what you do, or whether anyone else vouches for you. For B2B companies that spent a decade optimising for Google, this is a jarring new reality—and it's already measurable.
The traffic is shrinking, but the visitors are better
Since AI Overviews rolled out across European markets, organic click-through has softened. Fewer people scroll a list of ten blue links; many get their answer inside the AI summary and never visit a site. That sounds like bad news, and for raw traffic volume it is. But the visitors who do arrive via an AI recommendation behave differently: they're further down the decision funnel, they've already been pre-qualified by the assistant, and they convert at noticeably higher rates.
The strategic shift is clear. The goal is no longer to win the most clicks—it's to be the company the AI names when a buyer asks a high-intent question. That's a narrower, more valuable prize, and it rewards a very different kind of groundwork than classic SEO.
Why the AI doesn't know who you are
A detailed look at hundreds of small and mid-sized company websites revealed a telling pattern: the AIs could crawl the sites just fine, but they couldn't confidently answer the most basic question—who is this business and what, precisely, does it do? Pages were full of slogans and thin on facts. No clear description of services, location, specialisms, or proof. When a language model can't extract an unambiguous identity, it defaults to the safer option: it cites someone else.
This is the core insight for B2B. Being crawlable is not the same as being citable. The assistant needs structured, consistent, machine-readable signals about your identity before it will put its reputation on the line by recommending you.
The three things AI models look for
- Coherence: your company name, address, services and positioning must match across your site, your Google Business Profile, directories, and LinkedIn. Contradictions create doubt, and doubt means exclusion.
- Real third-party validation: genuine customer reviews, press mentions, industry listings and citations on sites the AI already trusts. Self-description isn't enough—the model wants corroboration.
- Explicit identity: say plainly what you do, for whom, and where. Spell out your niche. Vague "we deliver solutions" copy is invisible to a model trying to match a specific query.
Press and reputation quietly became ranking factors
There's a reason public relations is regaining influence. Everyone now publishes—blogs, posts, newsletters—but attention and trust are scarce. AI models weight third-party, editorially credible sources heavily when deciding who to cite. A mention in a respected trade outlet or a recognised directory now does double duty: it reaches humans and it teaches the machine that you're a legitimate authority. For entreprises competing to be named, earned coverage is no longer a vanity exercise—it's part of your discoverability infrastructure.
Don't forget the risk and infrastructure layer
Becoming more visible and more digitised raises your exposure. Companies that handle delegated or sensitive functions are increasingly targeted by phishing and impersonation—attackers exploit the very trust signals you work to build. As you push identity and credentials online, pair that visibility with hardened security and clear verification practices.
The same maturity applies to where your systems live. The uncritical "cloud first" era is giving way to deliberate choices—private cloud, hybrid setups, and data-sovereignty considerations are back on the table. For B2B firms building AI-driven processes, controlling your data and your infrastructure is part of controlling your reputation.
A practical starting sequence
- Rewrite your core pages so a model can state, in one sentence, who you are and what you deliver.
- Audit and align every public listing so your identity is perfectly consistent.
- Systematise real reviews and pursue a small number of credible third-party mentions.
- Add structured data and clear, factual service descriptions—write for extraction, not just for persuasion.
- Harden security and make deliberate infrastructure choices as your digital footprint grows.
The companies that will be cited by AI in the next cycle aren't necessarily the biggest—they're the ones that are legible, corroborated and consistent. That's a discipline, not a budget. Start treating your identity as data the machine can trust, and you move from invisible to recommendable.
Sources
- AI Overviews en France : pourquoi le SEO ne suffit plus — Journal du Net
- 474 sites de TPE passés au crible : les IA peuvent entrer, mais elles ne savent pas à qui elles parlent — Journal du Net
- GEO local : pourquoi être visible sur Google ne suffit plus aux PME — Journal du Net
- " Tout le monde communique, mais qui écoute encore ? " : pourquoi les relations presse reprennent le pouvoir — Journal du Net
- Quand l'État vous confie une clé, vous devenez la cible — Journal du Net
- Cloud privé : le retour du choix dans les stratégies cloud — Journal du Net
