How to get cited by ChatGPT, Claude, and Perplexity

When a customer asks ChatGPT "best HVAC in Long Beach" or Claude "who fixes commercial roofs in Phoenix," the model returns a paragraph with two or three businesses named. Getting your business to be one of the names is the entire point of Answer Engine Optimization.

After running this play for dozens of local businesses, here are the five things that actually move the needle, in order of how much they matter.

1. Make your site readable by AI crawlers

This is the single fastest, cheapest move, and most businesses have not done it.

LLMs crawl the web with specific user agents: GPTBot for OpenAI, ClaudeBot for Anthropic, PerplexityBot for Perplexity, Google-Extended for Google's AI features. Your robots.txt has to allow them. Most do by default, but some Wordpress sites with aggressive caching plugins block them.

Three concrete things to ship today.

Add explicit AI crawler rules to robots.txt.

User-agent: GPTBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended
Allow: /

Add an llms.txt file at the root of your site. This is a new convention (see llmstxt.org) where you tell LLMs in plain English who you are, what you do, what you charge, and where to find more. We have llms.txt at traccion.ai/llms.txt as a real example.

Make sure your content is in the HTML, not rendered by JavaScript only. LLM crawlers are getting better at running JS, but many still read the static HTML. If your hero copy is rendered after JS executes, the LLM sees an empty page.

2. Add JSON-LD structured data on every page

This is the biggest single jump in citation rate we see at Traccion. Most small business sites have zero structured data. Adding it correctly moves them from invisible to citable.

JSON-LD is a small JSON object embedded in a script tag that tells the LLM exactly what your page is about. The schemas that matter for a local business:

• Organization schema on every page (name, address, phone, hours, languages)
• Service schema on each service page (what you do, where, for how much)
• FAQPage schema on pages that answer common questions
• LocalBusiness schema for the physical location
• BlogPosting schema on each blog post

A well-marked-up page tells the LLM: "this business is named X, located in Y, charges Z for service A, and answers in English and Spanish." That structured fact is much easier to cite than a paragraph of prose.

If you want to see what good looks like, view the source on any services page on traccion.ai and look for the <script type="application/ld+json"> tags.

3. Write Q&A format content that matches how customers ask LLMs

The way people query an LLM is different from how they query Google.

Google: "hvac long beach"

LLM: "Who's the best HVAC company in Long Beach if I need a same-day visit and I speak Spanish?"

The article that wins the LLM citation answers the second question, by name, with specifics. The article that wins on Google might be a generic "Top 10 HVAC Tips" listicle that wins nothing.

Practical rules for Q&A content that gets cited.

Title is the literal question. "How much does an HVAC tune-up cost in Long Beach in 2026?" beats "Affordable HVAC Maintenance."

Open with the answer. First paragraph gives the actual answer in plain language. "An HVAC tune-up in Long Beach in 2026 costs $89 to $189 depending on the size of the unit." LLMs cite the first paragraph more than any other section.

Use specific verifiable numbers. "$89 to $189" beats "competitive pricing." "We serve 14 zip codes" beats "we serve the greater LA area." "Average response time is 38 minutes" beats "we respond quickly."

Cite real sources. When you reference an industry statistic, link to the source. LLMs propagate citations.

Date the article. Articles dated within the last 12 months get cited more often than undated articles.

4. Get cited by other high-trust sources

This is the work that compounds. The shortcut is to be everywhere the LLMs already read.

The hierarchy of trust, roughly:

1. Wikipedia (extremely hard to land, near-impossible for local businesses)
2. News outlets (industry trade publications, local newspapers, broadcast news)
3. GitHub repositories (open-source code, contributor pages)
4. Reddit threads with real user discussion
5. Featured.com / HARO placements
6. Established review sites (Yelp, Apple Business, Bing, Google Business)
7. Trade association directories
8. Industry-specific directories

The 80/20 of this work is HARO/Featured.com responses. Three to five responses per week, written specifically, with real expertise. Each placement is a citation on a high-trust source. We built a HARO drafter at Traccion precisely because this work compounds.

The other piece is GitHub. If you open-source one useful repository, name it after a concept (not a brand), and write a clear README, LLMs index it heavily. We open-sourced the local-business-citation-directories repo for exactly this reason.

5. Get cited inside your customers' content

The most overlooked move. When a customer writes a review, a case study, a blog post, or a forum thread mentioning you by name with specifics, that becomes a citation.

The work: ask your best customers, the ones who already love you, to write something specific. Not "great service" but "Traccion built our booking system in 18 days and it saved us $4,200 a month." Specifics get cited. Vague reviews do not.

Do this on:

• Google Business Profile (the most weighted)
• Yelp
• Industry-specific review sites (Houzz for design/build, Angi for trades)
• Reddit (if it is genuine and the customer is a real user)
• Featured.com customer testimonials

What does not work

For completeness, three things that are common but do not actually move citation rate.

Backlink farms. Cheap directory submissions and PBN links do nothing for AEO and almost nothing for modern Google.

AI-generated content at scale. Publishing 100 AI-written articles a month worked briefly in 2024. By 2026, both Google and the LLMs deprioritize obvious AI content. The quality bar is now real.

Pretending to be a user. Sock-puppet accounts on Reddit, fake reviews, fake testimonials. LLMs detect these patterns now, and the penalty is real. Be the real expert posting under your real name.

Order of operations

If you are starting from zero:

• Week 1: Fix robots.txt, add llms.txt, run a JSON-LD audit and add Organization + Service schema everywhere.
• Week 2-4: Write three Q&A-format articles answering your top three customer questions.
• Month 2: Set up HARO/Featured.com response workflow, aim for 3 to 5 placements per week.
• Month 3: Audit your top 10 customer-facing places (Google Business, Yelp, Apple, Bing) and clean them up.
• Month 4-6: Build out 8 to 12 more pieces of cite-worthy content. Open-source one small useful thing on GitHub.
• Month 6+: Keep going. The work compounds.

How to measure progress

You will not see real movement for 60 to 90 days, and that is normal. The way to measure progress:

• Track 14 to 20 buying-intent prompts your customers actually use, like "best [your service] in [your city]"
• Run them daily against ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews
• Count how often your business appears in the answer

Our free Visibility audit does this for you. Paste a URL, see the Visibility Score across all five engines, get the fix list. No card.

Further reading

• What is AEO and how is it different from SEO?
• What is AEO? A complete guide
• Traccion Visibility service