HowtoGetCitedbyChatGPT&Perplexity:ADeveloper'sGEOPlaybook

Why citations are the new backlinks

A backlink used to be the unit of authority on the web. It still matters for classic ranking. But in AI search, the unit of visibility is the citation: your domain, named inline, in the answer itself.

That last stat is the opportunity. An analysis of roughly 400 million AI-search citations found that comparison content is the most-cited format on the web — and when I checked who ranks for GEO topics right now, every first-page result was a tool vendor selling something. Zero neutral practitioner guides. That gap is why this post exists, and why it can rank.

How AI citations actually work

Strip away the hype and the pipeline is simple:

1. Retrieval. For your query, the system pulls candidate passages from its web index — the same index classic search uses.
2. Selection. It picks passages that directly answer the question with specific, verifiable claims.
3. Generation. It writes the answer and attaches citations to the claims.

Step 2 is where you win or lose. The model is not judging your brand. It is judging whether a 40-word passage from your page cleanly answers the query. Everything in this playbook serves that one moment.

The playbook

1. Put the answer in the first 60 words

   Every section should open with a direct answer block of 40 to 60 words — no throat-clearing, no "in today's fast-paced world." Models lift passages verbatim, and they lift the passage that already reads like an answer. Write the section so its first paragraph could stand alone as the cited text. If it can't, rewrite it until it can.

2. Be quotable: own a definition, a stat, or a framework

   Citations attach to claims. "GEO is SEO's additional layer" is quotable. "You should do GEO well" is not. Coin one crisp definition per post, publish one original number (even from your own analytics), and name your framework. Named things get cited; vague advice gets skipped.

3. Publish comparisons as real tables

   Citation analyses consistently show comparison content earning disproportionate citations — it is Perplexity's favorite format. If your post compares tools, approaches, or costs, put the comparison in a markdown table with concrete values, not in flowing prose. Tables survive extraction cleanly; paragraphs get mangled.

4. Ship schema markup

   This page carries BlogPosting, BreadcrumbList, and FAQPage structured data. Schema doesn't guarantee citations, but it makes your content's structure machine-readable, which makes it cheaper for retrieval systems to extract clean passages. If you run Next.js, this is thirty lines of code — there is no excuse to skip it.

   json

   Copy

   {
     "@context": "https://schema.org",
     "@type": "FAQPage",
     "mainEntity": [{
       "@type": "Question",
       "name": "What is generative engine optimization?",
       "acceptedAnswer": {
         "@type": "Answer",
         "text": "GEO structures content so AI search engines retrieve and cite it..."
       }
     }]
   }

5. Date everything and refresh quarterly

   AI search heavily favors fresh content for fast-moving topics — every "best AI tools" comparison older than a few months is already stale, which is a ranking window, not a problem. Put visible published and updated dates on the page, and actually update the content when the facts change. A quarterly refresh habit beats a one-time publish on any topic that moves.

6. Build entity mentions off your site

   Models cross-check who you are. A name that appears on GitHub, dev.to, LinkedIn, and in other people's citations is an entity; a name that exists on one blog is a string. Distribute every post: a canonical version here, a repost on dev.to with rel=canonical pointing back, a thread on X, a post on LinkedIn. Each mention is a vote that your entity is real.

7. One canonical page per question

   Don't split "what is GEO" across five posts. One page should be the definitive answer to one question, and everything else should link to it. Fragmented coverage means the model finds five weak passages instead of one strong one. Clusters beat confetti.

The tools worth paying for

You don't need a GEO platform to start. You need keyword research, a way to see what questions people ask, and analytics. When the content starts earning, these are the upgrades that pay for themselves:

Tool	What it does	Why it matters for GEO
Semrush	Keyword research, position tracking	Finds the low-competition questions worth answering
Frase	Content briefs, answer optimization	Structures drafts around the questions models answer
AlsoAsked	People-also-ask mapping	Gives you the FAQ section, pre-written by demand
Google Search Console	Indexing, impressions, queries	Free. Shows which queries already surface your pages

What not to do

FAQ

What is generative engine optimization (GEO)?

GEO is the practice of structuring and publishing content so AI search engines like ChatGPT, Perplexity, and Gemini retrieve and cite it in their generated answers. Where SEO optimizes for ranked blue links, GEO optimizes for being quoted as a source.

How do ChatGPT and Perplexity choose which sources to cite?

They retrieve candidate pages for a query, then cite passages that directly answer the question with specific, verifiable claims. Observed patterns show they favor direct answer blocks, comparison tables, original statistics, and well-structured long-form content over thin or purely opinionated pages.

Can a small blog get cited, or only big brands?

Small sites get cited regularly. AI search cares more about whether a passage directly answers the query than about domain authority. A focused practitioner post that answers one question precisely can out-cite a generic article from a major publisher.

Do I need schema markup to get cited by AI search?

Not strictly, but it helps. Article, FAQPage, and HowTo schema make your content's structure machine-readable, which makes it cheaper for retrieval systems to extract clean answer passages. This post ships with FAQ schema for exactly that reason.

Does llms.txt help with AI citations?

llms.txt is an emerging convention for giving LLMs a clean markdown map of your site. It is not a ranking factor and no major AI search engine has confirmed using it for citations. Treat it as a nice-to-have, not a strategy.

Is GEO replacing SEO?

No. GEO builds on SEO. AI search retrieves from the same web index that classic search uses, so crawlability, page experience, and topical authority still matter. If you haven't read it, start with my notes on perceptual performance — speed is still the cheapest ranking factor there is. GEO is the additional layer: structuring content so it survives being quoted.

Get cited on purpose

I run GEO audits for dev teams and founders: I take your site, find the pages that should be getting cited, and hand you the exact structural changes that fix it — answer blocks, schema, comparison tables, refresh cadence. If AI search is eating your traffic, that's the fastest way to take it back. Details are on my about page.

“

Write the passage the model wishes it had found.

”