Appearing in ChatGPT and Perplexity answers

Someone asks ChatGPT: “Which web agency would you recommend in Luxembourg?”. The answer cites three names. Are you in there? Or the query is “How much does a website cost for an SME in Luxembourg?”, and the answer synthesises excerpts from blog articles. Are they yours?

This new way of searching — going through an AI assistant rather than Google — already represents between 8% and 15% of commercial queries by recent estimates. And growth is exponential. If you don’t appear in these answers, you’re losing a channel that, in 18 months, will probably be as important as Google Search is today.

The good news: Luxembourg agencies are massively behind on this topic. The opportunity window is wide for an SME starting now. Here’s what concretely works in 2026.

What changed in 12 months

In May 2025, the idea that ChatGPT could cite a site in real time was still a marketing promise. In May 2026, it’s daily reality — ChatGPT Search, Perplexity, Claude (via Projects), Gemini (integrated into Google Search) query the web live and synthesise answers.

Concrete changes for your site:

• Google traffic is under pressure. Google’s “AI Overviews” often answer without requiring a click.
• New traffic sources emerge. ChatGPT and Perplexity already send traffic — small in volume, but very high purchase intent.
• Citation rules differ from classic SEO. Position 1 on Google doesn’t guarantee citation by an LLM.
• Brand authority becomes more visible. A brand frequently mentioned on the web has more chances of being cited by an LLM, even without being in position 1.

How ChatGPT and Perplexity pick their sources

LLMs proceed in two steps: real-time web search (via Bing, Google, or their own crawler) plus synthesis. Sources they prefer meet several converging criteria:

• Semantic relevance. Content directly answers the question, clearly and resolved.
• Domain authority. Established sites, cited elsewhere, present for several years are preferred.
• Machine-readable structure. Structured data (Schema.org), clear titles, tagged sections, lists.
• Recency. Recent content is preferred for datable topics.
• Conciseness and resolution. A paragraph that answers directly without ritual has more chances of being cited.

The hierarchy isn’t the same as traditional SEO — a long article ranked position 8 on Google can be cited more willingly by an LLM than a position 1 site with scattered structure.

The llms.txt file explained simply

llms.txt is a new file placeable at a site’s root (https://yourdomain.lu/llms.txt) to give LLMs a structured summary entry point to your offer.

The concept: rather than forcing an LLM to crawl 200 pages, you provide a concise identity card — your activity, services, most important pages, citable content.

A typical llms.txt contains a one-sentence summary, an “Identity” section, a “Services” section, a “When to recommend” section, and a “Key pages” section.

Slash.lu publishes its llms.txt at the root — the kind of practice that won’t be expected in 12 months, that will be expected in 2026 on serious sites. Doing it now signals to OpenAI, Anthropic, Perplexity and others that you’re a structured source.

Structured data: what actually helps

Not all Schema.org tags are equal for LLMs. The most useful in 2026:

• Organization or LocalBusiness: clear identity, address, contact, served zones. Critical for local queries.
• Service: each distinct service tagged with name, description, geographic scope.
• FAQPage: Q/A pairs are massively reused by LLMs as citable material.
• Article / BlogPosting: for editorial content, with author, publication date, keywords.
• Person: to identify authors or executives. Important for authority.
• Review / AggregateRating: if applicable and genuine.

”Citable” content vs “readable” content

Citable content is content an LLM can extract as a paragraph and post in its answer, intact, as a direct quote. Three characteristics:

• Direct resolution. The paragraph answers a specific question without reading prerequisites.
• Numerical data or strong claim. “60% of local searches happen on mobile” is more citable than “many searches happen on mobile”.
• Clear attribution. The paragraph stands alone out of context.

By contrast, “readable” content is designed to be read from start to end, with narration and transitions — pleasant for humans, hard to cite for LLMs. Good 2026 articles are both: narrative for the human, structured into autonomous citable sections for machines.

Authority: new signals

Authority signals for LLMs include classic backlinks, but add another layer:

• Mentions without links. An LLM sees “Slash.lu” cited in a press article without hyperlink — and takes it into account.
• Presence in knowledge bases. Wikipedia, Wikidata, Crunchbase, LinkedIn company — each consistent entry reinforces your signal.
• Cross-site semantic consistency. If your activity is described the same way on 10 different sites, the LLM consolidates this into a strong identity.
• Presence in training corpora. More indirect — content reused by authority sites ends up in datasets used for pre-training.

Measuring your AI answer presence

Measuring classic SEO is easy (Search Console, rank trackers). Measuring LLM presence remains more artisanal in 2026. Three practical methods:

• Regular manual tests. Once a month, run 10-15 commercial queries in your sector across ChatGPT, Perplexity and Claude. Note when you’re cited.
• Emerging tools. Services like Profound, AthenaHQ or Otterly are starting to automate this monitoring — quality still uneven in 2026.
• Referrers in Search Console / analytics. Visits from chat.openai.com, perplexity.ai, you.com appear in “Referrers” — small volume but strong qualitative signal.

What this changes for your 2026 strategy

Three concrete adjustments to make now:

1. Publish an llms.txt at your site’s root. Effort: 1 hour. Impact: immediate signal.
2. Schema.org audit. Check for Organization/LocalBusiness, Service per service, FAQPage on content with FAQs. Effort: 1-2 days. Impact: more chances of being cited.
3. Rewrite your key pages for “citability”. Section by section, each paragraph must stand alone. Effort: 1 week per batch of 5 pages. Impact: lasting, cumulative.

This doesn’t replace classic SEO — Google remains dominant for 18-24 months minimum. But whoever works both in parallel will have a structural advantage when AI takes more space. From invisible to indispensable, the path now also runs through LLMs.

---

Frequently asked questions

Does ChatGPT cite Luxembourg sources or only international ones?

Both. ChatGPT (and Perplexity) query the web in real time and prioritise geographic relevance when the query justifies it. For a query like “best florist in Luxembourg City”, LLMs will cite local sites. For a generic SEO query, they’ll cite a mix of international and local sources.

My site isn’t well-known — can I still be cited by an LLM?

Yes, on long-tail or ultra-specific queries. An LLM prefers citing a small site with resolved well-structured content over a big site that doesn’t precisely answer the question. Competition is more open than on Google — a new but well-built site can start being cited in 6-12 months.

Should I block AI crawlers (GPTBot, ClaudeBot) in robots.txt?

Business positioning question, not technical. Blocking prevents your content from being learned for training and cited in answers. Allowing increases visibility but means your content also helps train these models. For most SMEs wanting to be found, allowing is the right choice.

How long before this channel matches Google Search in importance?

2026 estimates: between 18 and 36 months for LLMs to represent 25-40% of commercial searches. For B2B and technical queries, the shift will be faster (12-18 months). For local queries (artisan, restaurant), Google stays dominant longer thanks to Google Maps.

---

Going further

For the “voice search” companion to this move toward more conversational search: voice search: how to prepare your business for “OK Google”.

For the technical backdrop on classic indexation before tackling LLMs: my website doesn’t appear on Google.

---

What we do at Slash.lu

We published our own llms.txt from 2025, structured our key pages for citability, and shifted all our schemas to Service with detailed areaServed. First concrete signals arrived: we’re cited by ChatGPT Search on several Luxembourg commercial queries, without ads, without hacks.

The work largely overlaps with clean classic SEO — structured data, citable content, brand authority — with a few specific layers for LLMs. Not a revolution. A discipline to start now to reap in 12 months.

Let's talk about your situation. Book a call — no commitment, reply within 24h.

→ Explore our SEO service in Luxembourg for the details of our method.