Skip to content
Luciqo

Step‑by‑step: Create GEO content that ChatGPT, Gemini and Perplexity will quote

Gustavo De Amorim7 min read

Short answer: write one page per buyer question, make every fact easy to lift, and mark it up so assistants can recognise authorship, dates and claims. Tie each section to a clear intent, add sources you’re happy to be quoted next to, and measure which assistants actually cite you.

Below is the repeatable template we use at Luciqo to turn “helpful content” into generative‑engine‑optimised answers.

What do assistants actually cite today?

  • ChatGPT’s search and “deep research” modes add source links when they use the web; OpenAI documents that “responses that use web search may include citations” and that deep research delivers results “with citations or source links.” (OpenAI Help; OpenAI Help: Deep research; OpenAI Academy).
  • Claude’s web search “includes citations so you can easily verify sources,” both in the product and API docs. (Claude Help; Claude docs).
  • Perplexity says it “searches the internet in real time… with sources and citations included.” (Perplexity).
  • Independent research shows that which pages get cited depends heavily on the assistant’s own retrieval and query rewrites, not just on-page quality. In one October 2026 study of 158k answers and 468k citations, the retrieval step explained far more variation than page features. (arXiv; see also arXiv).

Implication: to be cited, you must align to how assistants search, surface facts cleanly, and prove recency/authority in machine‑readable ways.

What does a “GEO page” look like?

Use one page per buyer question or task. Here’s the anatomy I recommend, in order, with micro‑copy hints that make quoting easy.

  1. H1: The exact buyer question
  • Example: “What is ISO 27001? A UK SaaS guide for CTOs in 2026”.
  • Micro‑copy: Add a one‑sentence plain‑English answer under the H1. Keep it quote‑ready.
  1. TL;DR box (4–6 bullets)
  • Include: definition; who it’s for; 2–3 actionable steps; 1 stat or standard; last updated date.
  • Micro‑copy: Write each bullet so it can stand alone in an AI snippet.
  1. H2: Key facts and definitions
  • Use short H3s: “What it is”, “Why it matters”, “Who needs it”.
  • Micro‑copy: One fact per sentence. Prefer verbs over adjectives.
  1. H2: Step‑by‑step or decision flow
  • For tasks, use numbered steps. For comparisons, use a compact table (rows = options; cols = criteria).
  • Micro‑copy: Start each step with an imperative verb; finish with a measurable outcome.
  1. H2: Evidence and sources
  • Link primary sources you’re happy to appear beside in citations (standards bodies, regulators, docs). Assistants expose these links to users. (OpenAI Help; Claude Help; Perplexity).
  • Micro‑copy: Prefix each link with what it proves: “Official spec (clause 6.1)”.
  1. H2: Trade‑offs, risks, exceptions
  • Short bullets: “Good when…”, “Not ideal if…”, “Watch out for…”.
  • Micro‑copy: This section wins trust and gets quoted because it reduces bias.
  1. H2: Examples (2–3)
  • “If you’re a fintech… do X first. If you’re a manufacturer… do Y.”
  • Micro‑copy: Keep examples concrete and UK‑relevant where it helps.
  1. H2: What to do next
  • 3–5 bullets: actions, owner, time estimate.
  • Add a link to your contact or product page only once; assistants downplay overt sales.

Which structured data should I add?

Assistants and search engines use schema to understand authorship, dates and page type. Follow Google’s Article schema guidance: mark up Article/BlogPosting with author, datePublished, and dateModified; link the author to a profile page if internal. (Google Search Central). For specific intents, add:

  • FAQPage: for 3–6 genuine FAQs tied to the page’s topic.
  • HowTo: for procedural pages with steps and required tools.
  • Product/Offer/Review: for product comparisons, make eligibility for rich results clear (pricing, availability, ratings when applicable).
  • Organization and Person: to assert who you are and who wrote it.

Practical tip: keep the visible byline, published date, and last updated date close to the H1; assistants prefer recent, attributable sources. OpenAI notes that search results can show source details and dates in product UIs. (OpenAI Help: Enterprise/Edu search).

What signals help my page get chosen as a citation?

  • Recency and clarity: Put “Updated: 20 November 2026” near the top and explain what changed.
  • Atomic claims: One claim per sentence with a nearby source link. Perplexity and Claude surface citation panels that let users verify claims quickly, so tidy linkage helps. (Perplexity; Claude docs).
  • Primary sources: Prefer regulator/standard docs over blogs in your Evidence section; assistants frequently cite them. (See Google’s own advice on linking to sources in AI experiences covered by Search industry reporting. Search Engine Land).
  • Query alignment: Research shows assistants rewrite queries and rely on their own retrieval layers; shape headings and anchors to mirror real questions and synonyms. (arXiv).

If you need a refresher on how assistants weigh source credibility, I break it down here: Which sources do ChatGPT, Gemini and Perplexity prefer — and how to get cited by them.

The repeatable GEO Writer template (copy/paste)

  • Page title (H1): buyer question + audience + “2026 update”.
  • One‑sentence answer.
  • TL;DR box: 4–6 bullets with a visible last‑updated date.
  • H2: Key facts (short H3s; one fact per sentence).
  • H2: Steps or comparison table.
  • H2: Evidence and sources (3–6 primary links; what each proves).
  • H2: Trade‑offs and risks (3–5 bullets).
  • H2: Examples by segment (2–3).
  • H2: What to do next (3–5 actions with owners/time).
  • Optional: FAQ (3–6) at the bottom; also mark up with FAQPage.
  • Footer: author bio, organisation schema, contact link.

How do I write quote‑ready micro‑copy?

  • Lead with the conclusion. “Do X because Y,” then details.
  • Use strong verbs (decide, compare, choose) and measurable nouns (cost, days, %).
  • Keep sentences under 20 words; assistants lift clean phrases.
  • Name entities exactly once per paragraph (product, version, UK spelling).
  • Add anchor‑friendly sub‑heads (“How do I…”, “What’s the cost…”, “When should I…”).

For deeper technical tactics on crawlability and rendering for AI engines, see: Measuring and improving ‘AI read‑ability’: technical SEO for generative engines and LLMs.txt and AI Visibility.

  • Internally, link from broader explainers to your GEO pages to concentrate relevance.
  • Externally, cite the exact clause, spec or law. Assistants show citation side‑panels where users inspect the supporting passage. (Perplexity).
  • Don’t over‑stuff links. A short Evidence list of authoritative sources beats 20 marginal ones; audits have shown assistants sometimes surface irrelevant links when overloaded. (EBU report).

How do I measure whether assistants actually cite me?

  • Track where you’re cited: run weekly prompt tests for your priority questions across assistants and store the outputs and links. We published a workflow: Prompt testing playbook.
  • Monitor your “recommendation share” and who gets recommended instead: How to track AI recommendation share.
  • Tie it to revenue: connect analytics and CRM to see which assistant‑referred users convert; here’s a practical guide for GA, Matomo and HubSpot: From GA to AI revenue.
  • Or use Luciqo to see which assistants recommend you, which competitors appear instead, and which sources they cite.

Quick checklist (ship in under 48 hours)

  • One page per buyer question; H1 mirrors the query.
  • TL;DR with last updated date.
  • Steps or comparison table; one fact per sentence.
  • Evidence section with 3–6 primary sources.
  • Article/FAQ/HowTo schema; visible author and dates.
  • Internal links in; one soft CTA out.
  • Run prompt tests; log citations; iterate weekly.

What next?

Pick your top five buyer questions for Q4, draft the GEO pages using the template above, add schema, and run a weekly prompt‑testing cadence. In two weeks you’ll know which claims and sources earn citations. If you want an audit or to connect assistant visibility to revenue, we at Luciqo can help you measure it end‑to‑end.

Frequently asked questions

What’s the single biggest change that increases citations?

Put a short, precise TL;DR with a visible last‑updated date and 3–6 primary sources near the top. Assistants expose dates and sources to users, and clear, recent claims are easier to lift. (See OpenAI’s notes on search with citations and Claude/Perplexity documentation on citing sources).

Should I chase AI Overviews specifically?

Treat Google’s AI surfaces as one channel; structure pages for clear questions and primary sources. Industry reporting suggests there’s no fixed formula for earning AI Overview links, but conversational, comprehensive answers and strong sources help. (Search Engine Land).

Does adding schema guarantee citations?

No. Schema helps assistants understand authorship and recency, but research shows retrieval and query rewrites drive much of what gets cited. Use schema plus great retrieval‑aligned structure. (Google; arXiv).

Three to six high‑authority, primary sources per page is plenty. Over‑linking can backfire; audits have shown assistants sometimes surface irrelevant links when overloaded. (EBU report).

When search or deep research is used, ChatGPT shows citations and often a ‘Sources’ or side panel. OpenAI’s help centre explains that web answers may include citations you can click to open. (OpenAI Help; OpenAI Help: Deep research).

Can I influence which brands assistants compare me with?

Yes—indirectly. Name the relevant alternatives in your copy and table headings to steer helpful comparisons, then measure which competitors assistants recommend instead using prompt testing or a platform like Luciqo.

See what AI says about your brand

Start with a free snapshot: 10 buyer questions, four AI assistants, results in minutes. No card needed.