A lot of business owners have tried it by now. You open ChatGPT, ask for the best supplier or service in your line of work, and read a list of your competitors. Sometimes you're on it. Usually you aren't, and it isn't obvious why.
The good news is that there's less mystery here than the industry selling "AI optimization" would like you to think. The tools that cite sources, like ChatGPT search, Perplexity and Google's AI Overviews, are built on top of web search. They find pages, read them, and quote the ones that answer the question best. Most of what gets you cited is what gets you found anyway, done more carefully.
How do ChatGPT and Google's AI answers pick sources?
They search first, then write. When you ask a question that needs current information, the tool runs one or more web searches, pulls the most relevant pages, and builds its answer from what those pages say. The pages it leaned on show up as citations.
Each tool has its own way of finding pages:
- ChatGPT search uses OpenAI's crawler. OpenAI's crawler documentation says OAI-SearchBot "is used to surface websites in search results in ChatGPT's search features."
- Perplexity uses PerplexityBot, which its bot documentation describes as the crawler that surfaces and links websites in its results.
- Google's AI Overviews and AI Mode draw on Google's normal search index. Google's page on AI features and your website says there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
So the first question is never "how do I trick the AI". It's whether your pages are in the index these tools search, and whether they're the clearest answer once they're there.
Do you need special markup or an llms.txt file?
No. Google is direct about this: you don't need new machine-readable files, AI text files or special markup to appear in its AI features.
You'll see agencies selling "AI schema" or llms.txt setups as the answer. An llms.txt file, a plain list of your key pages, is cheap to add, and we have one on this site. But nobody has shown it gets you cited, and Google says you don't need one. Treat it as housekeeping, not a strategy.
Structured data is still worth having, for the reasons it always was. It tells search engines what your business is, what a page is about, and how pages relate to each other. We cover what's worth adding in our schema checklist for manufacturers.
Can AI crawlers actually read your site?
Check this before anything else, because it's the one problem that makes everything else irrelevant. Three things commonly get in the way.
Your robots.txt blocks them. Some sites, or the plugins and security tools they run, block AI crawlers by default. Open yoursite.com/robots.txt and look for the names in the table below.
| Crawler | Company | What blocking it does |
|---|---|---|
| OAI-SearchBot | OpenAI | Keeps you out of ChatGPT search results |
| GPTBot | OpenAI | Keeps your content out of model training only |
| PerplexityBot | Perplexity | Keeps you out of Perplexity's results |
| Googlebot | Keeps you out of Google Search, AI features included | |
| Google-Extended | Controls use for Gemini training, not Search |
The difference between GPTBot and OAI-SearchBot catches people out. You can block training and still be cited in ChatGPT search, as long as OAI-SearchBot is allowed.
Your content only appears after JavaScript runs. Google renders JavaScript, so it copes. Many other crawlers don't, or do it unreliably. If you open your page with JavaScript turned off and the main text isn't there, some AI tools are seeing an empty page.
The important facts are inside PDFs or images. A price list in a PDF or a spec table saved as a picture is much harder for any tool to read and quote than the same information as text on a page. We wrote about this for spec sheets and product pages.
What makes a page worth quoting?
An AI answer is a summary built from sentences it can lift. Pages that make that easy get quoted more often. In practice, that means:
- Answer first. Put a direct, two or three sentence answer right under the question, then explain. That's why every post on this blog starts with a short answer box.
- Use the question as the heading. People ask these tools full questions. A heading like "How long does a website take to build?" matches the question far better than "Our timeline".
- Be specific. "Builds run $2,500 to $12,000" can be quoted. "Our pricing is competitive" can't.
- Back numbers with sources. Link to the study, the official documentation, or your own measured data.
There's research behind that last point. A 2024 paper from Princeton and others, GEO: Generative Engine Optimization, tested how rewriting pages changed how often generative search engines cited them. The best methods raised visibility by up to 40%, and the ones that worked best were adding citations, quotations and statistics. Keyword stuffing, the old SEO habit, made little difference.
Why do consistent facts about your business matter?
AI tools cross-check. If your homepage says you were founded in 2015, your LinkedIn page says 2017, and a directory lists a different phone number, the tool has three versions of you and no reason to trust any of them.
The same goes for your own site. We found a version of this on ours: one page listed a website build at one price range and another page at a different one. A person might not notice. A tool comparing sources will.
Pick one version of your name, description, address, phone number and prices, and use it everywhere:
- Your website, including the footer and contact page
- Your Google Business Profile and Bing Places listing
- LinkedIn and any social profiles
- Industry directories and review sites like Clutch
- Trade association member listings
Those outside profiles matter for another reason too. AI tools learn who you are partly from what other sites say about you. A business that only exists on its own website is harder to recognise than one that shows up, consistently, in the places buyers already look.
How long does it take to get cited?
Faster than ranking for a competitive term, if the page is specific. Tools that search the live web can cite a page soon after it's indexed. What slows most sites down is getting indexed at all.
Do these once:
- Verify your site in Google Search Console and submit your sitemap.
- Do the same in Bing Webmaster Tools. Bing's index also powers Microsoft Copilot, and it's the step most sites skip.
- Set up IndexNow, which tells Bing and other engines the moment a page changes instead of waiting for a crawl.
After that, the realistic path is narrow before broad. A new site will not be the source for "best web design agency" any time soon. It can be the clearest answer to a specific question in its niche within weeks.
How do you check if you're being cited?
There's no Search Console for AI answers yet, so track it by hand. It takes about an hour a month.
- Write down 10 to 15 questions your buyers actually ask. Use their words, not yours.
- Once a month, ask each one in ChatGPT with search on, Perplexity, and Google.
- Log whether you're cited, which page was used, and who was cited instead.
The "instead" column is the useful one. Open the pages that beat you and look at what they have that yours doesn't. Usually it's a direct answer, a number, or a table.
A checklist you can run today
- Does your robots.txt allow OAI-SearchBot, PerplexityBot and Googlebot?
- Does your main content show up with JavaScript turned off?
- Is your site verified in both Google Search Console and Bing Webmaster Tools?
- Do your key pages open with a direct answer to the question they're about?
- Are your prices, locations and services stated as plain text, not in a PDF?
- Do your name, phone number and description match on your site, Google, Bing and LinkedIn?
- Have you written down the questions you want to be cited for?
If you'd rather have someone check all of this for you, our website health check covers it, and our SEO service is where we fix it.