Quick answer
To make your business appear in ChatGPT, Claude, Gemini and Perplexity answers, do five things: make your content exist in the HTML before JavaScript runs (AI crawlers do not execute it at all), add Schema structured data, publish an llms.txt file at your site root, open every page with a direct answer in the first 100 words, and explicitly allow AI crawlers in robots.txt. These practices are known as GEO.
Something fundamental changed in how customers reach you: a growing share no longer searches Google but asks ChatGPT directly, "who is the best app development company in Riyadh?" If your business is not part of what the model knows, you are absent from that conversation — regardless of your Google ranking.
Fact one: AI crawlers do not execute JavaScript
This is the most important technical fact in this article. As of mid-2026, not a single major AI crawler executes JavaScript — not GPTBot, ClaudeBot, PerplexityBot, or OAI-SearchBot. They all request the page, read the HTML they receive, and move on.
The difference from Google is fundamental: Googlebot runs a real browser and waits for content to load; AI crawlers do not. The practical result:
This covers very common cases: a product grid fetched via API, a pricing table built in JavaScript, a collapsible FAQ section, and customer reviews loaded from a database. All invisible.
How to check your site in one minute
| Step | Command or action |
|---|---|
| 1. Fetch the raw HTML | curl -s https://yoursite.com |
| 2. Look for your content | Can you see your product names, prices and copy? |
| 3. Without a terminal | In your browser: view page source (Ctrl+U) and search for text you know |
If your content is not in the raw source, AI engines cannot see it. The fix is server-side rendering (SSR) or static generation (SSG).
Pillar 2: Structured data (Schema)
Schema is the layer that lets a model verify information instead of inferring it. The most important types for a services business:
- Organization: who you are, contact details, areas of expertise (knowsAbout), regions served.
- Service: a page per service with description, price range and service areas.
- FAQPage: documented question/answer pairs — among the most directly quoted content in answers.
- Article / BlogPosting: with publish and modified dates and word count.
- BreadcrumbList: showing where the page sits in your site structure.
Pillar 3: The llms.txt file
A simple Markdown file at your site root (/llms.txt) that introduces models to your identity and key pages. Its accepted format:
- A single H1: your company name.
- A blockquote with a concise description — the most important line in the file, and what models use to understand your identity.
- H2 sections with link lists in the form:
- [Title](URL): Description.
The practical benefit: instead of crawling dozens of pages and possibly missing the most important, the model reads one organised file. Published studies report a measurable improvement in citation accuracy and fewer incorrect statements about the company.
Pillar 4: Write to be quoted
- Answer first: put the definitive answer in the first 100 words. Models quote passages that answer directly.
- Specific numbers: "SAR 15,000 to 40,000" is quotable; "competitive pricing" means nothing to a model.
- Clear definitions: define terms the way an encyclopedia does — this is the style models prefer.
- Tables: data organised in tables is easier for a model to extract and compare.
- Real FAQs: phrased the way people actually type the question.
- Update dates: models favour recent content with a documented date.
Pillar 5: Allow the crawlers explicitly
| Crawler | Operator |
|---|---|
| GPTBot · OAI-SearchBot · ChatGPT-User | OpenAI (ChatGPT) |
| ClaudeBot · Claude-SearchBot | Anthropic (Claude) |
| Google-Extended | Google (Gemini and AI Overviews) |
| PerplexityBot | Perplexity |
| Applebot-Extended | Apple Intelligence |
| Meta-ExternalAgent | Meta AI |
An important note: blocking these crawlers protects your content from training, but it also means your business will not be mentioned in answers. The decision depends on your content — marketing content is usually something you want spread.
Pillar 6: Presence beyond your own site
- A complete Google Business Profile with genuine reviews.
- Listings in trusted business directories and professional platforms in your field.
- Mentions of your business in external articles and sources — a third-party mention is stronger than your own claim.
- Consistency of your name and description everywhere — contradictions confuse the model and lower its confidence.
How to measure the result
- Ask the models directly about your field and city, and record whether you were mentioned and who was mentioned instead.
- Monitor server logs for GPTBot, ClaudeBot and PerplexityBot visits — their presence means you are being read.
- Track referrals in your analytics from domains such as chatgpt.com and perplexity.ai.
These standards are applied to this very site — you can review our own llms.txt file. To apply them to your site, see our web development service.
FAQ
What is the difference between SEO and GEO?
SEO targets ranking on Google's results page, while GEO (generative engine optimization) targets being mentioned inside the answer a model writes. The technical foundation is shared — speed, structure and structured data — but GEO adds a decisive condition: the content must exist in HTML without JavaScript, and be written in a quotable form.
Do AI crawlers really not execute JavaScript?
No, they do not. As of mid-2026 none of the major crawlers (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot) execute JavaScript. They read the initial HTML only, which is why single-page applications without server-side rendering arrive completely empty.
Is llms.txt mandatory or an official standard?
It is not an official standard and not every company adopts it, but it is very cheap to implement (one file), carries no downside, and measurably improves how accurately models describe your business. It costs an hour of work, and that alone justifies it.
How long until the effect shows?
Faster than traditional SEO in some cases, because answer engines that search live (such as Perplexity and ChatGPT's search feature) read your site on demand. Knowledge baked into the model itself updates with its training cycles, which is far slower and outside your control.
Should I block AI crawlers to protect my content?
It depends on your content type. Marketing and introductory content you want spread, because its spread brings customers. Paid content or sensitive intellectual property may deserve blocking. The decision is commercial rather than technical, and robots.txt lets you separate paths instead of blocking everything.
