When a potential client asks ChatGPT, "which automation agency would you recommend for a growing business?", the answer names two or three companies. That list is being written right now, and it has nothing to do with bidding on a keyword.
This guide is the hands-on follow-up to our article on what LLMO is: what to do, in what order, and how to check whether it's working.
In this guide
- How each engine decides who to cite
- Step 1: let them crawl you
- Step 2: publish citable content
- Step 3: build presence beyond your own website
- How to measure whether you're being cited
How each engine decides who to cite
They don't all work the same way, and it's worth knowing that before you invest any effort:
| Engine | Where it pulls information from | What it prioritizes |
|---|---|---|
| Perplexity | Live search with source citations | Pages that answer the question directly and recently |
| ChatGPT (with search) | Its own index + live browsing | Recognizable sources and well-structured content |
| ChatGPT (without search) | Training knowledge | Brands with broad, repeated presence across the web |
| Google AI Overviews | Google's index | What already ranks well + clarity of the answer |
The practical takeaway: Perplexity and AI Overviews respond to focused work relatively quickly. Showing up in ChatGPT without search enabled is a years-long game built on distributed presence.
Step 1: let them crawl you
This is the most boring step, and the one most people skip. If your robots.txt doesn't explicitly allow AI crawlers, no strategy will save you.
- Allow
GPTBotandOAI-SearchBot(OpenAI),PerplexityBot,ClaudeBot, andGoogle-Extended - Publish an
llms.txtfile at your root with a summary of what your company does and links to your key pages - Make sure your content lives in the HTML, not only behind JavaScript: several crawlers don't execute JS
- Mark up your entity profile with
OrganizationJSON-LD: name, logo, address, official profiles
The mistake we see most often: companies investing in content while their CDN blocks AI crawlers by default. Months of work that no model ever reads.
Step 2: publish citable content
A model cites fragments, not full articles. Anything that makes a fragment easy to extract works in your favor:
- Answer before you set the context. A two- or three-line paragraph up top that directly answers the question in the heading.
- Define things precisely. "X is the process of..." is more citable than a narrative paragraph.
- Use tables and lists. Comparisons, numbered steps, and criteria are formats models extract well.
- Bring your own data. What only you can say — results from your own projects, patterns from your audits — is what sets you apart from aggregated content.
- Include a real FAQ at the end. Every well-answered question is an independent citation opportunity.
- Sign your work with a verifiable author. A LinkedIn profile and a bio on your site. It's a direct E-E-A-T signal.
If you had to prioritize one format, prioritize the glossary: short, precise definitions of the terms in your industry. They're the assets that get cited most for definition-type questions, and they're the cheapest to produce.
Step 3: build presence beyond your own website
Models don't only read your domain. They consult industry publications, forums, LinkedIn, comparison sites, and directories. A brand mentioned on six different sites has a much higher chance of showing up than one mentioned only on its own website.
- Articles in industry publications with links and context
- Case studies published by the client, not just by you
- Up-to-date profiles on directories and comparison sites in your category
- Sustained LinkedIn activity from your team, with real names and titles
Five mistakes that keep you out
Before you add new tactics, remove the ones working against you. These are the issues we see most often when auditing AI visibility:
- Blocking crawlers without realizing it. Many CDNs and security plugins block AI bots by default. Check your actual
robots.txt, not the one you assume you have. - Content that only exists behind JavaScript. If the text renders client-side, most crawlers see an empty page.
- Service pages with no definitions. They sell well and explain nothing. A model has nothing to cite.
- Inconsistent contact details. A different name, address, or phone number across your website, Google Business Profile, and LinkedIn weakens your entity profile.
- Content with no author. Without a verifiable byline, content carries less weight in experience and authority signals.
How to measure whether you're being cited
How to measure whether you're being cited
Keep it simple before you buy any tools:
- Write down 10 questions one of your clients might ask. Be literal: "best automation agency for small businesses."
- Run them every month in ChatGPT, Perplexity, and Gemini. Note whether you appear, in what position, and which source gets cited.
- Check GA4 for referral traffic from
chatgpt.comandperplexity.ai. It already arrives labeled. - Track which competitors show up. That's your map of content still to build.
With that spreadsheet alone, you'll have more insight than most companies in your industry.
Conclusion
Showing up in ChatGPT and Perplexity isn't magic or a technical trick. It's letting yourself be read, publishing answers worth citing, and existing beyond your own website.
What makes this urgent is that competition in this space is still low and the signals build up slowly. A year from now, what takes six months today will take twice as long.
Want to know what AI is saying about your brand right now? Request an AI visibility audit.
