Artificial intelligence: visibility, use, compliance
Search no longer always ends in a click. Ranking well is no longer enough, you have to become a source the engines cite. And the AI you already use now carries legal obligations.
Citability audit: are you cited when someone asks an assistant
Structuring content to be extracted: direct answer, dated figures, identifiable entity
Schema.org markup, llms.txt and AI crawler access
Using AI in production, with a human safeguard and a measure of what it returns
AI Act compliance: transparency, inventory of systems, traceability
A growing share of searches no longer ends in a visit. According to SparkToro (Similarweb data, June 2026), around 68% of US Google searches between January and April 2026 ended without a single click, up from 58.5% measured in 2024. A Pew Research Center study covering 68,000 queries also measures a relative drop in clicks of 46.7% when a generated summary appears at the top.
This is not the end of search marketing, it is a shift. The same study shows the opposite: clicking a link inside the generated summary remains rare, 1% of visits to pages with an AI summary, against 8% for a traditional result link on the same page. Position is no longer the currency, citation is.
Meanwhile, the AI you already use, often without realising it, has entered a regulatory framework. Since 2 August 2026, the transparency obligations of Article 50 of the AI Act apply.
Three workstreams, three horizons
| Workstream | The question it answers | First visible effect |
|---|---|---|
| Being cited | When a customer asks an assistant, who answers in your place? | 2 to 4 months |
| Producing with AI | How do you save time without publishing what everyone else publishes? | Immediate, measurable within a quarter |
| Staying compliant | Which AI systems do you actually use, and what must you declare? | A few weeks of inventory |
Our approach
- Citability audit. We put the questions your customers actually ask to several assistants, and record who gets cited. This is often the moment you discover it is your competitors, or worse, nobody in your sector.
- Structuring content. A direct answer at the top, a heading hierarchy that follows the questions, dated figures, an identifiable publisher. An engine does not cite what it cannot cut into pieces.
- Technical foundations. Consistent Schema.org markup, an llms.txt file, explicit AI crawler access, and a declared entity with what it can do and what backs that up.
- Framed use in production. Where AI saves time without damaging the brand, with mandatory human review and a measure of what it actually returns.
- Compliance. An inventory of the AI systems genuinely in use, usually more numerous than expected, and the transparency obligations that follow.
The mistake to avoid Producing more AI content to compensate for falling traffic. If everyone publishes generic material, differentiation comes back to what cannot be generated: your data, your real cases, your field experience. Volume makes worse the problem it claims to solve.
What you get
- A citability report, assistant by assistant, on your key questions.
- A content restructuring plan, prioritised by citation potential.
- The technical foundations laid and verified on your pages, not merely recommended.
- An inventory of your AI uses and the obligations that follow from them.
Who is it for?
For companies watching their informational traffic fall without understanding why, for those producing content that no longer stands apart from anyone else’s, and for those using AI tools in their shop or their customer relations without knowing what that implies.
Frequently asked questions
Is search engine optimisation dead?
No, its reward has changed. The same fundamentals still count, but the objective is no longer only to be first: it is to be the source the engine reuses in its answer.
How do I know whether ChatGPT or Perplexity cites my site?
By putting your customers’ questions to several assistants yourself and recording the sources shown. It is manual, but it is the only reliable measurement today, and it should be repeated every quarter.
Should we block AI crawlers?
It is a trade-off, not an obvious call. Blocking protects the content but rules out being cited. For a company selling its expertise, being cited with attribution is generally better than invisibility.
Does the AI Act really concern an SME?
Yes, as soon as it deploys an AI system that interacts with the public, a chatbot for example, or that generates content. The transparency obligations of Article 50 have applied since 2 August 2026. The rest of the timetable runs over several years.