Answer engines retrieve before they write.
Optimize the retrieval.
Generative engine optimization (GEO) gets you quoted by ChatGPT, Perplexity and Google AI Overviews. Five steps run first, and most advice fixes only the last.
View SEO expertise
Five steps to an AI citation
Every answer engine runs some version of these five steps. Each one can drop your page, and each responds to a different fix.
Fetch
Engines read only what your server sends. Put the answer in the HTML and name each AI crawler in robots.txt.
Chunk
Engines cut your page into short passages. Each must name its own subject, not open with “he” or “this approach”.
Embed
Each passage becomes a vector, a fingerprint of meaning. Names, figures and dates sharpen it. Hedges and synonyms blur it.
Rank
The engine rewrites each question four or five ways and scores passages against those. Answer every version, not one.
Cite
Retrieval gets you read. Credibility gets you named, and that takes an identifiable author, working sources and consistent markup.
Four fixes for those steps
These are files and paragraphs, not tactics. Each one targets a step above, and you can check it in the page source.
llms.txt and llms-full.txt
Two files listing your pages for AI agents. No engine has confirmed reading them, so treat them as cheap insurance.
One unambiguous entity graph
Write your author's details into every page instead of pointing at them. Two near-identical bios make two rival people.
Self-contained passages, placed high
Define the subject on the first screen, in third person, by name. Slightly formal prose reads clearly to a machine.
Outbound citations that resolve
A dead citation reads as fabrication. Auditing this site found 29 posts with unsourced numbers and one figure that traced back to nothing.
The SEO and answer engine optimization (AEO) reference sets out what llms.txt does and does not do, with Google's guidance linked.
What I measured on this site
Every number below comes from pooyagolchian.com, via Search Console and a script that queries live AI engines. I test on my own site first, so nothing here is a projection.
The engines cite more often than Google ranks
Across 30 questions on 22 August 2026, Perplexity named this site 14 times and ChatGPT 9. Google organic showed it 7 times.
Measured movement on flat impressions
Across the August 2026 work, sitewide click-through rate moved from 0.32% to 0.86% and average position from 11.9 to 8.7. Impressions stayed flat.
A brand entity thin enough to count
On 6 September 2026, Google returned 16 results for “pooya golchian”, with this site first and no sitelinks or knowledge panel. GEO here aims to grow that number.
The instrument, not the anecdote
A script asks real ChatGPT, Gemini and Copilot the same questions before and after, and records whether the site appears.
For the full method, read how to measure AI citations. For what the ChatGPT numbers mean, read how to get cited by ChatGPT. Where this work parts ways with SEO is in GEO vs SEO.
What GEO cannot do for you
This field attracts confident promises. Here is what the pitch leaves out.
You cannot buy a citation
No bid, placement or publisher deal puts you in an organic AI answer. Guaranteed AI mentions are ads or fiction.
An engine that never searches never finds you
An assistant can answer from memory without searching. In August 2026, ChatGPT searched on 8 of 26 bare-name queries for this site, and 48 of 58 with a qualifier.
Markup cannot invent authority
Schema describes what exists. It cannot create a conference talk, a package people install, or a third party who names you.
The results are unstable by construction
Answer engines change without notice and answer one question differently an hour apart. Track a fixed question set over time.
Read the SEO and AEO reference, or run the free AI visibility checker on a site.
Generative engine optimization, answered
Short answers to what people ask first.
Generative engine optimization, or GEO, makes a page retrievable and quotable by AI answer engines such as ChatGPT, Perplexity and Google AI Overviews. It works on the five stages those systems run, which are fetch, chunk, embed, rank and cite. In practice that means server-rendered HTML, self-contained third-person passages placed high in the document, one unambiguous entity graph, and outbound citations that still resolve.
Traditional SEO competes for a click from ten blue links, while GEO competes to be the passage an engine lifts into a synthesized answer. SEO's unit of value is the page. GEO's is a few hundred tokens, and its payoff is a named citation. The technical foundations overlap almost completely. What changes is that a passage must make sense alone, because the retriever will separate it from the page that gave it context.
No engine has publicly confirmed that it reads llms.txt, so nobody honest will promise you a ranking effect from shipping one. Pooya Golchian treats llms.txt and llms-full.txt as a cheap agent-readable index rather than a ranking lever. The measurable benefit is internal, because generating the file forces every important URL into one list and makes the gaps in a site's own corpus visible.
Pooya Golchian measures AI citations with a script that asks real ChatGPT, Gemini and Copilot the same questions through searchapi.io. The script records whether each answer names or links the site. Running that before and after a change produces a citation rate you can compare. In August 2026, ChatGPT and Perplexity cited pooyagolchian.com more often than Google ranked it. That pointed the work at off-site entity density instead of on-page retrieval.