What has changed in how people find you
Three years ago, a business owner who needed a supplier opened Google, typed "custom software Miami" and chose from ten links. Today, more and more, they open ChatGPT or Perplexity and ask: "Who can automate invoicing for a distributor in Miami? Give me three options with prices."
And the AI answers with three names. If yours isn't there, the conversation continues without you.
This isn't a prediction. I see it in my clients' analytics: a growing share of visits and inquiries comes from AI assistants. And they're better-quality inquiries, because they arrive already informed.
The good news: most of your competitors have done nothing about it. The window is open.
How an AI decides whom to recommend
AI assistants don't keep a list of good companies. When you ask them, they search the web (Perplexity and ChatGPT with browsing always do; Gemini and Google AI Overviews too), read the pages they find, and build an answer from what they understand.
To recommend you they need three things:
- To find you. Your website must be crawlable and show up in the searches they run underneath.
- To understand you. Your website must make clear what you do, for whom, where and at what price. No ambiguity.
- To trust you. Signals that you're real, experienced and that others have chosen you.
Each of these three is worked on concretely. Let's go one by one.
1. Getting found: the technical part
Allow AI crawlers in. Many websites block GPTBot, ClaudeBot or PerplexityBot in their robots.txt without knowing it (some WordPress themes and security plugins do it by default). Check your robots.txt. If you block, you don't exist.
Real links in the menu. If your navigation works with JavaScript and buttons instead of normal links, crawlers never reach your service pages. This is a mistake I fix constantly on modern template-built sites.
Speed and clean HTML. AI crawlers have little patience. A site that takes 6 seconds to load or needs JavaScript to display content gets read badly or not at all.
An up-to-date, multilingual sitemap with correct hreflang tags if you work in several markets. If you sell in Spain and Miami, your Spanish and English sites must be properly signposted.
2. Getting understood: structured data
This is the biggest opportunity and where almost nobody does anything.
Structured data (schema.org, in JSON-LD format) is a small block of code on each page that says, in a language machines understand without ambiguity: "This is a service, it's called X, it's offered by Y, in these areas, at this price, and it answers these questions".
What your website should have:
- Person or Organization with name, description, location, areas served, languages, links to real profiles.
- Service on every service page, with description, area served and, if you can, indicative prices.
- FAQPage with the real questions and answers for each service. AIs love this because it's exactly the format they answer in.
- BreadcrumbList so they understand the structure.
- Article with author, date and topic on every blog post.
And a development from the last two years: the llms.txt file. It's a plain-text document at the root of your site (yoursite.com/llms.txt) that summarizes, in clear language, who you are, what you offer, where and how to contact you. It's designed specifically for AI models to read. It takes an hour to make and sets you apart from 95 percent of your competitors.
Quick test: ask ChatGPT or Perplexity "what services does [your company] offer and which areas does it work in?". If the answer is vague, incomplete or made up, your website isn't explaining what you do. That's your starting point.
3. Getting trusted: trust signals
AIs, like people, recommend whoever seems trustworthy. What they look for:
A real person behind it. Name, photo, track record, linked LinkedIn profile. Anonymous "we are a team of experts" sites inspire less trust than "I'm Carlos, I've been doing this for 20 years".
Specificity. "We develop innovative digital solutions" says nothing. "I automate invoicing for distributors with over 100 invoices a month, for €3,000 to €6,000" says everything. AIs cite the specific.
Indicative prices. I know publishing them is scary. But when someone asks the AI "how much does a custom CRM cost?", the AI cites whoever gives a number. Whoever doesn't, doesn't appear.
Content with real experience. Articles that tell concrete cases, with numbers, with mistakes made, with opinions. AIs distinguish between generic content (of which there's plenty) and content that adds something. They only cite the latter.
Consistency. Your name, address, phone and description must be identical on your website, on LinkedIn, on Google Business and in any directory. Contradictions lower trust.
4. Content AIs cite (and content they ignore)
A simple rule: AIs cite content that answers concrete questions directly.
Works:
- Titles that are real questions: "How much does it cost to automate a process?", "Off-the-shelf or custom CRM?".
- Direct answers in the first paragraph, no detours.
- Lists, tables, numbered steps.
- Figures, price ranges, timelines.
- Honest comparisons ("Shopify if…, custom if…").
- Cases with measurable results.
Doesn't work:
- Filler text to "rank for keywords".
- Generic content that could be on any competitor's website.
- Articles without author, date or opinion.
- Vague promises ("guaranteed results", "industry leaders").
Everything I write on this blog follows these rules for a reason: it's what makes a business owner who asks an AI end up reading me.
5. What doesn't work (and wastes your time)
Stuffing the site with the word "AI". AIs aren't impressed by their own name. They look for clarity, not buzzwords.
Mass-produced AI content. A hundred generic articles written by ChatGPT with no review. Google penalizes them and AIs ignore them because they add nothing they don't already have.
Blocking crawlers "to protect the content". If you worry about AI using your content, I understand. But the cost is not existing for whoever asks. For a company that sells services, it doesn't pay.
Expecting results in a week. Crawlers take weeks to re-read your site. Well-made changes show in two or three months. It's an investment, not a trick.
How I apply it on the websites I build
Every website I deliver ships with complete structured data, llms.txt, FAQs on every service page, crawlable links, a multilingual sitemap and an author page with a real person behind it. It's not an extra; it's the minimum to exist in 2026. Details are in the websites that sell service.
And if you already have a website and want to know how AIs read it today, book a call. In 15 minutes I'll show you what ChatGPT and Perplexity see when they ask about your company, and what would need to change for them to recommend you.
In short
Getting AI to recommend you isn't magic or luck. It's having a website that can be crawled, that explains without ambiguity what you do and for whom, that gives prices and concrete data, and that shows a real, experienced person behind it.
It's exactly the same thing that convinces a human customer. It just has to be said, now, in a format machines understand too.
