What is llms.txt and why does it matter?

A tiny file where you list solid-answer pages and allow citation with a link. It helps AI assistants know what they can safely use.

How much does it cost and when’s ROI?

Typically 3,000–6,000 EUR setup and 490–1,200 EUR/month. If there’s real demand and healthy margins, many see ROI in 3–6 months.

Can I force ChatGPT or Perplexity to mention me?

No. But you can improve odds with clear answers, proof, and quotable pages. Test weekly and refine.

Does GEO work for local businesses?

Yes. Especially where “now” and “near me” matter. Clinics, repair shops, plumbers, lawyers, local travel tend to see fast impact.

GEO: how to show up in ChatGPT, Perplexity, AI Overviews

Q: Is GEO the same as SEO?

No. SEO wins classic results. GEO gets you quoted inside AI answers. Combine both for best effect.

Hook

Be honest: how often do you get a full AI answer and never click a website? It’s normal now. In 2026, a big slice of customer attention lives inside AI answers. The real question: when someone asks “who can fix my heating in Iași today?”, do you show up in that answer?

In plain English: what GEO is and why it matters

GEO stands for Generative Engine Optimization. It’s how you present your content so ChatGPT, Perplexity, and Google’s AI Overviews can quote you. Not “secret SEO sauce”. It’s common sense for the one-second summary era.

You’re no longer fighting only for the blue link. You fight to be the source behind the answer. If the AI cites or mentions you, trust goes up on the spot. The assistant your customer relies on “recommends” you.

What works? Clear pages with short answers to real buyer questions. Ballpark prices, opening hours, coverage areas, process, warranties. Real cases with numbers. A simple “facts page” that’s easy to quote. There’s also a tiny file called llms.txt. Think of it as a menu for the AI waiter: you list the pages with solid answers and say they can be quoted with a link.

Perplexity favors clean, precise sources. AI Overviews likes verifiable, short formulations. ChatGPT rewards clarity and evidence. In practice, a site with concrete statements, arranged for a hurried human, has a fair shot at being pulled into the response.

A concrete example: the Cluj clinic that “entered” AI answers

A dental clinic in Cluj with 8 doctors had two issues. 1) Many prospects asked ChatGPT “best emergency dentist in Cluj”. 2) Perplexity quoted competitors. In Google, AI Overviews showed lists without them.

What they did in six weeks with their marketing team (plus light consulting from DevoneX):

Published a page “Dental emergencies in Cluj: what we do in the first 60 minutes” with short answers, “from” prices, extended hours, average wait time to appointment.
Added 3 mini case studies with photos and numbers: “acute pain – fixed in 45 min”, “fracture – temporary in 60 min”, costs, timing, patient feedback.
Created a one-page “facts dossier”: years in business, number of doctors, equipment, licenses, addresses, parking, warranties.
Placed an llms.txt file at the site root, listing those pages and clearly allowing quoting with a link.
Rewrote 10 FAQs into 5–7 line answers, at the top of service pages.

Results after 90 days:

Perplexity cited them consistently for “dental emergency Cluj”, “dentist Sunday Cluj”.
Google’s AI Overviews pulled two paragraphs from the emergency page, with a link.
ChatGPT mentioned them in 3 out of 5 test prompts like “where can I go now?”, including address and hours.
AI-sourced leads in the CRM: +29% vs previous quarter.
New calls mentioning “the assistant recommended you”: +37/month.
New patients over 3 months: +52, average revenue/patient 168 EUR.

Total investment: 4,800 EUR setup, 690 EUR/month. Payback in ~4 months. Not magic, just tidy execution. At DevoneX, we’ve seen similar patterns with auto repair shops and small dermatology clinics.

Real 2026 costs (EUR)

No fluff, no shiny tiers. Expect something like this:

Setup: 3,000–6,000 EUR. Content audit, rewriting key pages, “facts dossier”, 2–4 case studies, llms.txt, lead tracking.
Monthly: 490–1,200 EUR. Q&A updates, 1–2 new case studies/month, monitoring citations in Perplexity/AI Overviews, minor tweaks.
Optional photo/video: 300–900 EUR per session if you lack assets.
Typical ROI: 3–6 months if your average ticket is 80–100 EUR+ and there’s active demand locally.

For context: 690 EUR/month is roughly one-third of a receptionist’s salary, but it works 24/7 and never gets tired.

When it’s worth it and when it’s NOT

It’s worth it if

You sell services with decent margin (clinic, repair, local IT, courses, local travel).
People actually search in your city (queries like “urgent”, “now”, “near me”).
You can show proof: starter prices, response times, real cases, reviews.
You’re okay with plain language and short answers. No corporate fog.

It’s not worth it yet if

99% of sales are word-of-mouth and you won’t touch the site for 3 months.
You sell ultra-cheap items where more traffic won’t move profit.
You can’t respond fast to new calls/messages. You’ll burn the leads.
No one searches your service locally (or it’s ultra-niche with ~10 searches/month).

Brutal truth: GEO isn’t a trick. It’s order and proof. If you have nothing concrete to show, it won’t help. If you do, it puts you on the AI shelf at eye level.

How to start in 5 simple steps

List your top 10 buyer questions. Put short answers (50–120 words) at the top of service pages.
Create a “Facts & figures” page: from-prices, hours, average wait time, service areas, warranties, years in business, team size.
Publish 3 case studies with photos and numbers. Problem, what you did, duration, cost, client quote.
Add an llms.txt file at your site root. Point to the “quotable” pages and clearly allow citation with a link. It’s a signpost for assistants.
Weekly tests: ask ChatGPT, Perplexity, and Google AI Overviews 10 buyer-like questions. See if you appear. Tweak page answers.

Our default at DevoneX is a 60-minute kickoff: questions, numbers, what you want to sell more of. Then we rewrite for hurried humans. If you want a blunt yes/no for your case, ping us at the end.

Bottom line: pitfalls to avoid

Don’t chase “ChatGPT Search ranking” hacks. Assistants quote clarity, proof, and usefulness. Don’t overpromise or hide prices. If AI quotes you and the page disagrees, trust collapses. Don’t ship 20 weak pages. Five excellent, data-backed pages beat them every time.

Short FAQ

Is GEO the same as SEO? No. SEO helps you win classic results. GEO helps you get quoted in AI answers. Best results come from doing both.

Do I really need llms.txt? Not mandatory, but helpful. It signals “here are solid answers” and allows quoting with a link. For some clients, it sped up citations.

How long until I see results? Usually 4–12 weeks. It depends on competition and how clear your pages are. First citations can show up in 2–3 weeks.

Can I control AI Overviews? Not 100%. But you can influence it with short, unambiguous answers high on the page and clear numbers. That improves correct inclusion.

Does this work for e-commerce? Yes, especially for “X vs Y” and “which should I choose” questions. You need honest comparison guides with data and clear recommendations.

Any risks? AI may summarize poorly. That’s why you use numbers and simple sentences. Check mentions regularly and adjust wording.

Want an honest take on whether GEO is worth it for your business? Send us a note. We’ll answer plainly. Contact