Vizoryo › Blog › LLM SEO: how to make your business citable by ChatGPT, Claude, and Perplexity

Start with what is different

Classic SEO targets the SERP — a ranked list of links the user picks from. LLM SEO targets the answer — a synthesized response the user reads as the answer. Two consequences fall out of that:

• Position is binary. You are inside the answer or you are not. There is no #4.
• Extraction matters more than ranking. The assistant builds the answer by pulling facts out of source documents. Documents that are easy to extract from get used; documents that are not, do not.

Everything in this guide flows from those two facts.

Step 1: Get the crawler welcome mat right

If LLM crawlers cannot reach your pages, nothing else matters. The crawlers you care about:

• GPTBot — OpenAI's training crawler
• OAI-SearchBot and ChatGPT-User — used by ChatGPT search and ChatGPT browsing
• ClaudeBot and anthropic-ai — Anthropic's training crawler
• PerplexityBot and Perplexity-User — Perplexity's retrieval crawlers
• Google-Extended — Gemini and AI Overviews training
• Applebot-Extended — Apple Intelligence training
• Meta-ExternalAgent, Amazonbot, CCBot, Bytespider — other LLM training crawlers

Open your robots.txt. Confirm every bot above is allowed. A default User-agent: * + Allow: / covers them, but many sites have older boilerplate that blocks unknown crawlers. WAF rules at Cloudflare, AWS, or your hosting provider sometimes block these crawlers without anyone realizing.

After updating robots.txt, tail your server logs for the same user agents. If you do not see them within a week, you are still being blocked somewhere downstream.

Step 2: Render facts on the server

Most LLM crawlers do not execute JavaScript — GPTBot, ClaudeBot, and PerplexityBot read the raw HTML your server returns; of the major crawlers, only Google's and Apple's render pages like a browser. Pages that ship facts in client-rendered React only — without server-side HTML — get partially crawled, partially indexed, and partially cited.

For any page that matters for LLM SEO — homepage, services, locations, FAQ, pricing — make sure the core text is in the HTML that comes back from your server, not injected by a script that runs after page load. Use SSR, static generation, or a hybrid approach where bots get the rendered version.

Step 3: Add structured data on every important page

Schema.org JSON-LD is the highest-leverage thing you can ship for LLM SEO. The minimum kit per page:

• Homepage and About: Organization or LocalBusiness with name, URL, logo, sameAs (social profiles), contact info, address.
• Every internal page: BreadcrumbList so the assistant understands site structure.
• FAQ pages and service pages: FAQPage with five to ten real questions and direct answers.
• Pricing page: Product or SoftwareApplication with offers and prices.
• Service pages: Service with name, description, provider, area served, and (where it applies) offers.
• Review pages: Review or aggregated AggregateRating with reviewer, rating, body — if you have real reviews.

Verify every page with Google's Rich Results Test. If it parses there, it parses for LLM crawlers too.

Step 4: Write FAQ blocks that match real questions

FAQ blocks are LLM SEO gold for a reason: they are pre-parsed, declarative, and exactly the format an assistant is looking for when answering "people also ask" style queries.

Three rules for FAQ content:

1. Use the actual phrases customers use. Not "What are our service hours?" but "When are you open?" Use the wording from your sales calls and support tickets.
2. Answer in 1-3 sentences. Long answers get truncated. Short answers get cited verbatim.
3. Lead with the fact, not the context. "We are open Sunday through Thursday, 9 to 6" — not "Great question — for many years we have served our customers..." Assistants extract the lead.

Mark up every FAQ block as FAQPage JSON-LD. Six to eight Qs per page is the sweet spot.

Step 5: Be unambiguous about what you are

LLMs build a mental category for each business before recommending it. Ambiguous businesses do not get recommended.

Pick the one sentence you want every LLM to know about your business. Make sure that sentence is on your homepage, your About page, and your social profiles, in roughly the same words. Examples that work:

• "A kosher catering company in Tel Aviv serving weddings of 80–400 people."
• "A boutique commercial law firm in Haifa specializing in tech M&A."
• "An independent veterinary clinic in Petah Tikva — small animals, walk-in hours, English- and Hebrew-speaking staff."

Examples that lose to the above:

• "Premium experiences for discerning clients."
• "Your trusted partner in modern legal solutions."
• "Caring for the family members who matter most."

Step 6: Make your name, address, and phone agree everywhere

NAP consistency — Name, Address, Phone — is the oldest local SEO discipline. It matters more for LLM SEO, not less. Assistants discount data they see expressed inconsistently across sources.

Pick the canonical version of your business name. Spell it the same way on your website, Google Business Profile, Wikipedia, Facebook, every directory listing. "Cohen & Associates" vs "Cohen and Associates" vs "Cohen & Associates LLP" — pick one, stick to it. Same for phone format and address format.

Step 7: Earn citations on sources LLMs trust

When two sources disagree, the assistant picks the more authoritative one. Authority comes from being mentioned across multiple credible sources — not from your own claims about yourself.

Where to focus:

• Wikipedia — if you are notable enough to clear the bar, this is the single biggest signal. Most SMBs cannot, but every category leader should try.
• Wikidata — easier to enter than Wikipedia, and LLMs do consume Wikidata.
• Industry-specific directories — Avvo for lawyers, Healthgrades for doctors, G2 for software, etc.
• Trade publications and news — even one or two mentions in named outlets shift assistant confidence noticeably.
• Structured business indexes — services like Vizoryo that publish your business as a structured AI-readable entity on a stable canonical URL.

Step 8: Measure

You cannot improve what you do not measure. For LLM SEO, classic Google Search Console tells you nothing about ChatGPT or Perplexity. You need three measurement layers:

1. Manual probes. Ask each assistant the queries you actually want to win, once a week. Write down whether you appear, in what position, with what attribution. Cheap, low-tech, high-signal.
2. AI search analytics tools. Profound, Otterly, Bluefish AI, and similar tools track brand mentions in LLM answers across thousands of prompts. Useful at the enterprise scale; expensive but precise.
3. Server log analysis. Filter your server logs for AI crawler user agents. You will see exactly which crawlers are hitting which URLs and how often. Free if you already store logs.

What to skip

• Stuffing keywords into your page. Useless. Assistants do not rank by keyword density.
• AI-generated filler content. Both Google and LLM assistants are increasingly good at detecting generic AI prose. It gets discounted.
• Link buying. Always a bad idea; especially bad for LLM SEO where the goal is real citations from real sources, not artificial ranking signals.
• Obsessing over llms.txt. The format is new and may matter eventually, but it is no substitute for the work in this guide.

A 30-day plan

If you have one month and a small team:

• Week 1: Audit. Robots.txt + crawler access. Run Google's Rich Results Test on your top 10 pages. List the schema gaps.
• Week 2: Add schema.org JSON-LD on every important page — homepage, services, locations, FAQ, pricing. Verify each.
• Week 3: Write or rewrite FAQ blocks for your top 5 pages. Mark up as FAQPage. Tighten the one-sentence description of what your business does and put it on the homepage and About page.
• Week 4: NAP audit across every external listing. Pick up two or three new third-party listings — industry directories, structured business indexes, Wikidata.

At the end of those four weeks you will have moved further than most of your competitors will move all year. Measure for the next 60 days. Adjust. Repeat.