Getting ranked in Google is a process most people understand by now. There’s a body of practice around it — keywords, backlinks, technical health, content quality — and while the details shift, the fundamentals have been stable for years.
Getting cited in an AI-generated answer is newer territory. ChatGPT is giving answers based on its training data and, increasingly, live web browsing. Perplexity is crawling the web in real time and citing sources directly. Google’s AI Overviews are synthesising answers from organic results before a single blue link appears.
For many businesses, this is becoming the layer of search that matters most — particularly for consideration-stage research, where a buyer asks “what should I look for in a [service]” and gets an answer that either includes you or doesn’t.
The good news: the fundamentals aren’t as different from SEO as you might expect. The signals that AI models use to select and trust a source overlap substantially with what makes a site rank well. But there are some specific technical steps worth taking, and a WordPress site is well-positioned to implement all of them.
Here’s what actually moves the needle.
How AI models decide what to cite
Before getting into the specific steps, it’s worth understanding what’s happening when an AI model chooses a source.
ChatGPT’s responses are partly drawn from training data — websites and documents the model processed before its knowledge cutoff — and partly from live web browsing when that feature is active. Perplexity operates primarily as a real-time web search tool, pulling from current pages and citing them explicitly. Google’s AI Overviews synthesise from organic search results in real time.
In all three cases, the model is making a judgement about which sources are trustworthy, clear, and relevant enough to include. That judgement is based on signals — some visible, some structural — that your site either has or doesn’t.
The sites that get cited consistently share a few characteristics: they’re clearly attributed to a real entity, they’re structured so the content is easy to parse, they demonstrate genuine expertise rather than generic claims, and they answer questions directly rather than burying the answer in waffle.
With that in mind, here are the six signals that matter most.
Signal 1: Structured data and schema markup
Schema markup is code you add to your pages — invisible to visitors, readable by machines — that tells crawlers exactly what a page contains. For AI citation purposes, it’s one of the highest-leverage things you can implement on a WordPress site.
The key schema types for a service business are:
- Organization (or LocalBusiness) — your business name, description, URL, location, contact details, and social profiles. Goes on the homepage and ideally in the sitewide footer.
- Person — establishes you as a real individual, with a name, description, job title, and links to your profiles. Important for solo practitioners and personal brands.
- Service — describes what you offer, who it’s for, and what it costs. Goes on individual service pages.
- FAQ — marks up question-and-answer content so it’s directly parseable. Goes on FAQ sections or dedicated FAQ pages.
- BreadcrumbList — communicates the site’s hierarchy to crawlers.
- Article — on blog posts, with author, datePublished, and dateModified.
On WordPress, the easiest implementation path is Rank Math, which handles Organization and Person schema automatically and lets you add FAQ and Article schema through the post editor. For custom JSON-LD — which gives you more control — you can add it directly in your theme’s functions.php or via a plugin.
A minimal but correct Organization schema for a WordPress developer might look like this:
{
"@context": "<https://schema.org>",
"@type": "LocalBusiness",
"name": "Warren Groom",
"description": "Freelance WordPress designer and developer based in Toronto, specialising in custom WordPress builds for agencies and regulated sectors in Ontario.",
"url": "<https://warrengroom.com>",
"address": {
"@type": "PostalAddress",
"addressLocality": "Toronto",
"addressRegion": "ON",
"addressCountry": "CA"
},
"sameAs": [
"<https://linkedin.com/in/warrengroom>"
]
}
This gives an AI model everything it needs to represent the business accurately — name, description, location, and a canonical URL it can verify.
Signal 2: Entity clarity
AI models think in entities — people, organisations, places, products, topics. When an AI is trying to decide whether to cite your site, one of the first questions it’s answering is: do I know who this is?
Entity clarity means your site leaves no ambiguity about who you are, what you do, and where you operate. In practice that means:
- Your business name is consistent everywhere it appears — homepage, about page, footer, contact page, social profiles, Google Business Profile. Not “Warren Groom”, “Warren Groom Design”, and “WG WordPress” on different pages.
- Your location is stated clearly and consistently. If you serve Toronto clients, that should be apparent on your homepage, your contact page, and in your schema.
- Your About page reads like a declaration, not a mystery. Who you are, what you do, how long you’ve been doing it, who you do it for. This is the page an AI model is most likely to draw from when constructing a description of your business.
- Your NAP (name, address, phone number) is consistent across every place it appears — your website, Google Business Profile, any directories or citation sites.
Entity clarity is the foundation everything else builds on. A site with excellent schema but inconsistent entity information is still confusing to an AI model — because the structured data and the prose are contradicting each other.
Signal 3: E-E-A-T content
Google’s E-E-A-T framework — Experience, Expertise, Authoritativeness, Trustworthiness — was developed as a content quality assessment rubric, and AI models apply similar judgements when evaluating whether to cite a source.
For a WordPress developer’s site, this translates to a few concrete things:
Author information on published content. Blog posts and articles should have a named author with a bio that establishes genuine expertise. Not just a name — a sentence or two that explains why this person is worth citing on this topic. Rank Math’s Author schema handles this automatically if your user profiles are filled in correctly.
Specific, verifiable claims. Generic statements like “I deliver high-quality WordPress sites on time” don’t help. Specific ones do: “I’ve been building custom WordPress sites for agencies in Toronto since 2008” is verifiable and attributable. “I’ve built sites for clients in twelve regulated sectors, including dental practices, mortgage brokers, and First Nations Band Councils” demonstrates actual experience.
Cited sources where appropriate. When you make a factual claim — about a regulatory body, a platform’s behaviour, a search trend — linking to the source signals that your content is anchored in reality, not fabricated.
Depth over breadth. A post that fully answers one question is more citable than a post that superficially touches ten. AI models are looking for the most complete, accurate answer to a specific query — not the most comprehensive list of related points.
Signal 4: Conversational content structure
Traditional SEO content was often structured around keyword density — getting the target phrase into the headline, subheadings, and body a certain number of times. That approach produced a lot of content that read awkwardly and wasn’t particularly useful to anyone.
AI models are asked questions by real people in natural language. The content they cite tends to be content that answers those questions directly — which means structuring your pages around how people actually ask, not just what they search.
In practice this means:
- Headings that reflect actual questions. “How does a WordPress developer structure their fees?” is a better H2 for AI citation purposes than “Pricing and fee structures”.
- Intro paragraphs that state the answer upfront. Don’t make a reader (or an AI) wade through three paragraphs of context before getting to what the page is actually about. State it clearly in the first sentence.
- FAQ sections with real questions. The questions in your FAQ should be the questions your clients actually ask — not the questions you wish they’d ask. “Do you work with agencies who have their own PM process?” is real. “Why choose a custom WordPress build?” is marketing.
- Conclusion paragraphs that summarise. AI models often pull from summary content. A clear concluding paragraph that restates the key point is useful.
This doesn’t require rewriting your entire site. Start with your highest-value service pages and your most important blog posts — the ones where you most want to appear in AI-generated answers.
Signal 5: An llms.txt file
llms.txt is an emerging standard — proposed by Jeremy Howard at fast.ai — that lets you give AI models a curated, structured summary of your site directly. It’s a plain text file that lives at yoursite.com/llms.txt and tells AI models who you are, what you do, and which pages matter.
It’s not yet a universal standard, but adoption is growing fast among technically-oriented businesses and publishers, and several major AI tools are already using it. For a WordPress site, it takes about twenty minutes to set up and requires no plugin.
I’ve written a full guide to llms.txt — what it contains, how to write it, and exactly how to add it to a WordPress site — in What is llms.txt and does your WordPress site need one?
Signal 6: Clean technical foundation
None of the above works if your site has basic technical problems. AI crawlers — like search engine crawlers — deprioritise slow, broken, or inaccessible sites.
The technical checklist for AI citation readiness is largely the same as for good SEO:
- Page speed. Core Web Vitals passing on mobile and desktop. A slow site creates a poor crawling experience and signals low quality. In WordPress, this means a lightweight theme, a well-configured caching layer, and images served in next-gen formats (WebP, AVIF).
- Clean crawlability. No important pages blocked in
robots.txt. No crawl errors in Google Search Console. Your XML sitemap is submitted, up to date, and contains only the pages you want indexed. - No broken links or 404 errors. These signal abandonment and reduce crawl budget efficiency.
- Accessible HTML. Semantic heading structure (one H1 per page, logical H2/H3 hierarchy), meaningful alt text on images, and no content locked behind JavaScript that crawlers can’t render.
- HTTPS throughout. No mixed content warnings. All internal links using the canonical URL.
In WordPress, the heavy lifting here is done by your theme, your caching setup, and your SEO plugin. If you’re running Rank Math with a lightweight theme and a caching layer, you’re likely in reasonable shape — but it’s worth running a quick audit to confirm.
What not to do
A few things that won’t help — and may actively hurt:
Keyword stuffing. AI models aren’t counting keyword density. They’re assessing relevance and quality. A page that repeatedly forces in a phrase reads as low quality to both humans and machines.
Thin content written to a template. A 300-word page that technically covers the topic but doesn’t actually answer anything is not going to be cited. The bar for AI citation is genuine usefulness — the model is trying to give the user a good answer, and it needs content that can actually do that.
Inconsistent information across your site. If your homepage says you’re based in Toronto, your about page says you serve clients globally, and your schema says you’re in Ontario, an AI model encounters conflicting signals and may default to not citing you rather than risk giving inaccurate information.
Schema that doesn’t match the page content. Adding FAQ schema with questions that don’t actually appear on the page, or Organisation schema with a description that contradicts your homepage copy, creates inconsistency that undermines trust.
How long does it take to see results?
Honest answer: it depends on the tool and the change.
Perplexity operates primarily on real-time web crawling, so improvements to your crawlability and content quality can show effects within days or weeks. ChatGPT’s responses are partly based on training data with a knowledge cutoff, so changes you make now may not be reflected in responses for some time — though the browsing feature updates faster.
Google’s AI Overviews respond to changes in organic rankings — so improvements that lift your pages in traditional search will generally improve your AI Overview presence too.
The practical advice: implement the signals above, track your citations in the tools you care about (you can do this manually by asking the AI directly), and assess over a 60-90 day window. This isn’t a game of overnight results — it’s the same long-term compound return as good SEO.
The six signals above — structured data, entity clarity, E-E-A-T content, conversational structure, llms.txt, and a clean technical foundation — are the foundation of a WordPress site that AI tools can accurately understand and confidently cite.
None of them require starting from scratch. Most can be applied to an existing site methodically, starting with the pages where AI citation matters most to your business.
If you’d like to understand how this applies to your specific WordPress site, GEO and AI visibility is a service I offer — or you can start with the foundational reading: What is GEO? and What is llms.txt?
