How to Get Your WordPress Site Cited by ChatGPT and Perplexity

To get your WordPress site cited by ChatGPT and Perplexity, be the clearest, most quotable and most machine-readable source on one specific question, then make sure an assistant can actually fetch and parse the page that answers it. Those are the two halves of the job: supply a passage worth quoting, and remove every technical reason a retrieval system might skip you. No tactic guarantees a citation, because selection logic is unpublished and probabilistic, but each step below removes a reason to be passed over.

This guide walks through the four questions worth auditing a page against, how to measure whether any of it is working, and a ten-step checklist you can run this week. If you are new to the discipline, start with our plain-English explainer on what answer engine optimisation is and come back. RankWave AI and Hydrogen SEO are both built by Sheesh Labs, so read the Hydrogen SEO examples below as a builder's illustration rather than a neutral recommendation.

How do ChatGPT and Perplexity choose what to cite?

Nobody outside these companies can describe the selection logic exactly, and it changes without notice. What is observable is the shape of the pipeline, which runs in roughly three stages.

1. Retrieval. The system assembles candidate pages, usually from a conventional search index plus its own crawling. If your page is not crawlable, or is invisible in ordinary search, it rarely enters the pool at all.
2. Extraction. The system pulls passages from those candidates. Short, self-contained passages that make sense without the paragraph before them survive this stage. Meandering prose that depends on context does not.
3. Synthesis and citation. The model composes an answer and attributes some of it to sources, typically two or three. Pages that stated something plainly, with identifiable authorship and a credible date, are the easiest to attribute.

The three surfaces differ in flavour rather than in fundamentals. Google's AI Overviews sit on top of Google's index, so conventional ranking work is the entry ticket. Perplexity is citation-forward by design and retrieves in real time, which makes freshness and crawl access matter immediately. ChatGPT blends model knowledge with browsing, so being well represented in ordinary search indexes and being trivially extractable are your levers.

Are your pages fetchable?

This is the unglamorous half, and it is where most sites quietly lose. A page that cannot be retrieved never gets to compete on quality.

Server-render the parts that matter. Titles, meta descriptions, headings, body copy and JSON-LD should all be present in the HTML the server returns, not assembled by JavaScript after the browser boots. Many fetchers do not execute scripts, and the ones that do may not wait. Check with curl or by viewing source rather than the rendered DOM in developer tools: if your answer is missing from the raw HTML, it may be missing from the machine's view too.

Audit your crawl rules deliberately. Look at robots.txt and decide, explicitly, which AI user agents you allow. Plenty of sites block AI crawlers by copying a snippet from a forum and then wonder why they are never cited. The decision is legitimate either way, but it should be a decision. Check per-page robots meta too: a stray noindex on your best explainer does the same damage.

Keep sitemaps honest. An XML sitemap is your list of what matters, and lastmod is your freshness claim. If every URL reports the same date, or the date changes whenever an unrelated widget is touched, the signal is noise. Aim for complete coverage with accurate per-URL timestamps. In the benchmark Sheesh Labs ran across seven WordPress SEO plugins, the sitemap outputs varied widely in exactly this respect, with several plugins supplying lastmod for only part of their URLs.

Publish an llms.txt file, last. It is a proposed convention for handing AI systems a curated plain-text index of your most important pages, described in your own words. Adoption is uneven and unproven, so do not expect much, but the cost is close to zero. Hydrogen SEO's llms.txt editor documentation covers publishing one from wp-admin, and the longer llms.txt guide explains what belongs in it and what does not.

Are your answers quotable?

Quotability is a writing problem, not a plugin problem. The unit of competition is no longer the page; it is the passage a machine can lift intact.

Open with the answer. Two to four sentences at the top of the page that fully answer the question in the title, and still make sense pasted into a chat window on their own. Everything after that is evidence and nuance. This is featured-snippet discipline applied to a new audience, and it works for the same reason.

Shape headings as questions. "How long should a meta description be?" declares precisely what the section answers. "Further considerations" tells a machine nothing. Then write the section to answer its heading completely and stop, so that the heading-plus-section unit is extractable on its own.

Add a genuine FAQ block. Four to six real questions with self-contained answers of roughly forty to eighty words gives you several independently quotable passages per page, each already in question-and-answer form. Do not pad it with questions nobody asks; thin FAQ blocks help nothing and look like filler to readers.

Be specific enough to be worth quoting. Assistants attribute claims, and a claim with a number, a version, a date or a named method is far more attributable than a generality. "Faster" is unquotable. "Added 12.9 ms of PHP time on a 40-post install" is a fact someone can cite you for. Publish your own measurements where you have them, and say how you took them.

Make the entity unambiguous. Use one consistent organisation name everywhere, keep a real about page, put an identifiable author on things, and cover one domain deeply rather than everything thinly. Retrieval systems are more willing to cite a source they can confidently identify.

Is your page machine-readable?

Schema markup does not make an assistant like you. It removes ambiguity about what your page is, who wrote it and when it changed, and ambiguity is what stops a system attributing a claim to you.

Three types carry most of the weight for question-driven content. Article or BlogPosting establishes headline, author, publisher and dates. FAQPage turns your question-and-answer pairs into explicit structure that mirrors what an answer engine is already trying to extract. HowTo does the same for genuine step sequences. Add Organization markup once, site-wide, with a consistent name, URL and logo, and link your author entities to it.

Two rules keep this from backfiring. First, only mark up what is visibly on the page, or you are describing a page that does not exist. Second, validate the output rather than trusting the plugin, because malformed JSON-LD is worse than none. Our guide to which schema types earn rich results on WordPress covers the choices in more detail, and Hydrogen SEO's write-up on getting cited instead of your competitors makes the same case from the plugin side.

Consistency across the machine-readable surfaces matters as much as the markup itself. The name in your Organization schema, the name in your title tags, the name on your about page and the name in your llms.txt should be identical strings. Sites that call themselves three slightly different things across those four places are asking a classifier to guess.

How do you check whether you are being cited?

Rank tracking cannot answer this question. Nothing in Search Console tells you whether ChatGPT mentioned you. You need a different measurement loop.

The manual version is free and works. Write down the twenty questions your customers actually ask, in their words. Put each one to ChatGPT, Perplexity and Google, once a month, and record one of three outcomes: cited with a link, mentioned without a link, or absent. Note which competitor was cited instead. After three months you have a trend, which is the only thing worth acting on, because individual answers vary run to run.

The tedious part is the repetition, and that is what monitoring tools exist for. The hydrogenseo.com platform includes AI-visibility monitoring that tracks whether ChatGPT, Perplexity and Google's AI Overviews cite a site for your chosen queries and charts the result over time; its answer engine optimisation documentation explains how to read the output, and Hydrogen SEO's post on why ChatGPT never mentions your site covers the reasoning behind the metrics. Sheesh Labs builds that platform as well as this site, so the useful takeaway is the method rather than the vendor: baseline first, change one thing at a time, and review monthly. Whatever you use, insist on tracking over time rather than one-off screenshots.

What does a ten-step citation checklist look like?

Run this against your ten most valuable pages. Most sites find their problems in the first four rows.

On WordPress, steps two, three, eight and the llms.txt work are plugin territory. The free Hydrogen SEO plugin we cover on this site handles server-rendered metadata, JSON-LD across the common types, sitemaps with images and per-URL lastmod, a robots.txt editor and an llms.txt editor without an account, though it is Beta software at version 0.0.20 and worth treating as such. Several other plugins cover the same ground; the point is that these four steps should not cost you developer time in 2026. Our overview of what AI SEO actually does on WordPress separates the parts that work from the parts that are marketing, and Sheesh Labs has a companion piece on answer engine optimisation for small teams if you are doing this without a dedicated SEO person.

Steps four to seven, the ones that decide whether you are worth quoting, cannot be bought. They are editing. That is the good news: the half of this that competitors cannot outspend you on is the half that matters most.