What Is CITED Framework?

How the CITED Framework Works

An AI answer is assembled in two moves. First the engine retrieves a set of candidate documents; then it synthesizes a reply that lifts passages from a few of them and attributes each to a source. A page can fail at either move, and it fails silently — nothing tells you that your best paragraph was never in the candidate pool, or that it was retrieved and passed over. Generative engine optimization is the practice of winning both moves; the CITED Framework is a way to break that practice into five parts you can check one at a time.

The five letters are ordered roughly the way an engine encounters them. Indexability and Topical authority decide whether a page is retrieved at all. Credibility, Entity optimization, and Distribution decide whether a retrieved passage is the one the model quotes and whether it trusts you enough to put your name on the claim. Because the letters are sequential, they behave like a chain rather than a scorecard: a perfect score on four of them counts for nothing if the fifth is broken. The framework’s job is to make that broken link visible.

CITED is diagnostic, not a metric an engine reports. No AI system exposes a “CITED score.” What it gives you is a shared vocabulary for the question every content team eventually asks — why did the machine cite them and not us? — and a way to answer it that points at a specific, fixable gap instead of a vibe.

The Five Levers

• Credibility — Is the claim trustworthy and cleanly stated? This is the letter most under your direct control: specific numbers attributed to a named origin, direct quotes, plain declarative phrasing. It maps to citation readiness and extractability — a passage that can be lifted whole without inheriting your ambiguity.
• Indexability — Can the machine actually reach and read the page? An AI crawler has to be allowed in robots.txt, the page has to render its content without a JavaScript wall, and it has to be indexable. This is the gate: if it’s shut, the other four letters never get evaluated.
• Topical authority — Does the site cover the subject with enough depth that a retrieval system treats it as a relevant candidate for the query? Thin, one-off pages rarely enter the candidate set for competitive questions; a coherent body of work on the topic does.
• Entity optimization — Does the engine recognize you as a real, resolved thing rather than an unattached string of words? Entity optimization ties your brand, authors, and products to established entities so the model can attribute a claim to a “who” it already understands.
• Distribution — Is your reputation echoed off your own domain? Mentions, references, and corroborating sources across the web raise the odds that a model, cross-checking, keeps landing on you — feeding share of voice in the answers themselves.

Example of the CITED Framework

The walkthrough below is illustrative — a way to see the five levers acting on one page — but every mechanism it leans on is real and documented.

Start with Indexability, because it is the gate. To be quoted by ChatGPT’s browsing tool, a page has to be reachable by OpenAI’s user agents; OpenAI documents GPTBot and OAI-SearchBot as the crawlers that must be allowed in robots.txt for that content to be fetched. A single disallow line here zeroes out everything downstream. Google’s own crawler overview documents the same principle for Google-Extended and AI features: no access, no candidacy.

Now Credibility. The 2023 paper “GEO: Generative Engine Optimization” measured which content changes made a generative engine more likely to feature a passage, holding the underlying facts constant. Adding cited statistics, direct quotations from authorities, and authoritative sourced language produced the largest lifts — the best methods raised visibility inside AI answers by up to roughly 40% (the paper reports +41% on one visibility metric), while keyword stuffing did not help. So a credible answer block is not a stylistic nicety; it is the single content change with the most measured evidence behind it.

The remaining three letters explain the gap that Credibility alone can’t close. Topical authority is why a lone strong page on an unfamiliar domain still struggles against a site with a deep, coherent body of work on the subject — retrieval favors the established candidate. Entity optimization is why a recognized brand gets named where an unresolved string does not: the model can attribute the claim to an entity it already understands. And Distribution is why corroboration off your own site matters — a model cross-checking a fact keeps landing on the source that other sources also point to. Run a page through all five and the diagnosis is rarely “improve everything.” It is one specific letter, and that letter is your ceiling.

Frequently Asked Questions

The Bottom Line

CITED breaks the vague goal of "show up in AI answers" into five things you can actually audit: can the machine reach your page, does it cover the topic deeply, is the claim credible and clean enough to quote, does the engine recognize you as a real entity, and is that reputation echoed across the web. Work the chain and fix the weakest link first — the letter you are worst at is the one deciding whether you get named at all.