Generative Engine Optimization: Get Discovered by AI

You open ChatGPT at 1am and type the question your future clients type: "who are the best wedding photographers in Manchester?" or "which indie designers make good Notion templates?" It answers in two seconds. Three names, little summaries, neat source links. None of them are you. You have the portfolio, the reviews, the years of posting. The machine just has no idea you exist.

That sting is the whole reason this guide exists. LitFame's blunt thesis is simple: if you don't get famous, how can AI discover you? Assistants like ChatGPT, Perplexity, Gemini and Google's AI Overviews build their answers out of what the public web already says. If the web says nothing about you, the answer has nothing to work with.

Generative engine optimization, usually shortened to GEO, is the practice of making yourself findable and quotable inside those AI answers. This hub covers what GEO actually is, how AI answers choose their sources, what the original GEO research measured (and what it left out), the practical playbook for a person or small brand, and how to measure whether any of it is working. No hacks, no guaranteed placements, just the honest mechanics.

What generative engine optimization actually means

A generative engine is any system that answers a question by writing a response instead of handing you ten blue links: ChatGPT with search, Perplexity, Gemini, Microsoft Copilot, Claude with web search, and Google's AI Overviews and AI Mode. When these tools pull in web pages, they usually show citations, little links that point back to the pages they leaned on.

GEO is the work of getting your pages, profiles and name into those answers and citations. The term comes from a 2023 research paper, and marketers have since stretched it to cover almost anything related to AI visibility. You will also see AIO, AEO and LLMO floating around. They are overlapping industry labels with no official definition, and our breakdown of AIO vs GEO vs SEO untangles them if the jargon is making your eyes glaze.

For a creator, freelancer or small brand, a practical definition works better than a theoretical one. GEO means three things:

• Being retrievable. Your site and profiles can be crawled, indexed and found by the search systems these assistants use.
• Being recognisable. When an AI reads about you across several sites, it can tell those pages describe the same person or brand.
• Being quotable. Your content contains clear, specific, trustworthy statements an answer can lift and attribute to you.

Most of this overlaps heavily with good old search engine optimization. Google says so directly in its own guide to generative AI features, and that overlap is good news. The skills you build for Google carry straight over.

How AI answers pick their sources

Nobody outside these companies knows exactly how each model ranks sources, and anyone who claims to is guessing. What the companies publish is enough to understand the general shape.

Two places an answer can come from

An assistant can answer from what it learned during training, or from pages it fetches live at the moment you ask. Training knowledge is broad but frozen at a cutoff date and has no links attached. Live retrieval is fresh, and it is where citations come from. When an assistant fetches pages to support an answer, the industry calls it grounding or retrieval-augmented generation (RAG).

For someone trying to get discovered, live retrieval is the battleground. It updates within days, and it sends clicks.

What Google says about AI Overviews and AI Mode

Google's Search Central guide to optimizing for generative AI features is unusually direct. It says those features "are rooted in our core Search ranking and quality systems" and describes two techniques. The first is retrieval-augmented generation, where Google's ranking systems pull relevant pages from the Search index and the model writes an answer from them with links. The second is query fan-out, where the model fires off several related searches at once. Google's own example: a question about fixing a weedy lawn might fan out into searches about herbicides, chemical-free weeding and prevention.

So to show up in Google's AI answers, you mostly need to show up in Google. Google's AI features page lists two eligibility requirements: the page must be indexed, and it must be eligible to show a snippet in normal Search. Since mid-2026, Search Console also has a setting that lets site owners include or exclude their site from these generative AI features, rolled out to all sites worldwide on August 31, 2026, per Google's announcement. If your traffic from AI features looks oddly flat, check that the toggle is set to include.

Fan-out is also why a niche page can surface for a broad question. Ask about starting a home bakery and the fan-out might include licensing and pricing, so a small baker's thorough pricing page can earn a citation it would never win for the broad query. Our guide to getting cited in Google AI Overviews goes deeper on that.

What OpenAI, Perplexity and Anthropic say

OpenAI's help centre says ChatGPT search "ranks search results using multiple factors intended to help users find relevant, reliable information" and that "placement is not guaranteed." It also explains that ChatGPT sometimes sends rewritten queries to partner search providers, then reviews the results. To be eligible, OpenAI says you must allow its crawler, OAI-SearchBot, and make sure your host or CDN does not block OpenAI's published IP addresses.

Perplexity's documentation says PerplexityBot "is designed to surface and link websites in search results on Perplexity" and is not used to crawl content for AI foundation models. Anthropic's help centre lists a separate Claude-SearchBot that indexes content to improve Claude's search results. Each company runs its own crawler for search, and each one respects robots.txt. The full list of bots, what each one does and how to configure them lives in our setup guide to llms.txt, schema and AI crawlers.

The pattern: every major assistant answers fresh questions by searching the web, reading a handful of pages and summarising them. Your job is to be one of the pages they read, and easy to quote once they get there. Our guide on how to get recommended by ChatGPT applies this to OpenAI's assistant specifically.

What the GEO paper found, and what it didn't

The term generative engine optimization comes from a paper titled "GEO: Generative Engine Optimization" by Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan and Ameet Deshpande, a team that included Princeton researchers. It was posted to arXiv in November 2023 (arXiv 2311.09735) and accepted at KDD 2024, a major data mining conference. It gets quoted constantly, usually badly.

The setup

The researchers built GEO-bench, 10,000 queries from nine sources. For each query they fed the top five Google results to GPT-3.5-turbo and had it write a cited answer, mimicking tools like Perplexity at the time. Then they rewrote one source page with different techniques and measured how much more of the answer credited it. Key tests were repeated on Perplexity itself.

Visibility was scored two ways: "position-adjusted word count" (how many words of the answer credit a source, with earlier citations weighted higher) and "subjective impression" (a model's rating of how relevant and influential the cited material felt).

The results

They tested nine rewriting methods. The headline findings:

• Quotation Addition, Statistics Addition and Cite Sources were the strongest, improving position-adjusted word count by roughly 30 to 40 percent relative to the unedited page and subjective impression by 15 to 30 percent. In plain English: quotes from credible sources, concrete numbers and named sources.
• Fluency and readability edits (Fluency Optimization and Easy-to-Understand) also gave a 15 to 30 percent boost.
• Keyword stuffing, the classic old-school SEO trick, gave "little to no improvement." On Perplexity it performed about 10 percent worse than doing nothing.
• Lower-ranked pages gained the most. When the Cite Sources method was applied, the page ranked fifth in Google saw a 115.1 percent visibility increase, while the top-ranked page's visibility dropped by 30.3 percent on average.
• It depends on the topic. Statistics helped most for law, government and opinion questions. Quotations helped most for people, society and history questions.

The caveats nobody mentions

Every page in the experiment was already one of the top five Google results. The study measured how much credit a page got once it was already in the room. How to get into the room was outside its scope. If no search system retrieves your page, rewriting it with quotes and stats changes nothing.

The experiments also used GPT-3.5-turbo and a 2023 version of Perplexity. Today's assistants use different models and different retrieval pipelines, and none of them publish their ranking logic. Treat the paper as solid evidence for a direction (specific, sourced, readable content gets quoted more; stuffing keywords does not) and skip the temptation to quote "40 percent" as a promise.

The encouraging part for small creators: last-place pages benefited most. Once you are retrieved, clear and well-supported writing can punch above its weight.

Why mentions beat a big follower count

Here is the honest part, and it might cost us a sale. No AI company documents follower counts as a ranking signal for its answers. ChatGPT does not sort creators by Instagram follower number, and Google's AI features run on the same ranking systems as Search.

What gets you into AI answers is presence in the sources those systems read. That means:

• Mentions on other people's websites: interviews, roundups, podcast show notes, local news, community forums.
• Links from those mentions back to your site or profiles, which help search engines discover and trust your pages.
• Consistent profiles that describe you the same way everywhere, so a system reading five pages can tell they are all about one person.
• Quality content that answers real questions better than the alternatives.

Social reach still matters a lot, indirectly. A visible account is how journalists, podcast hosts and list-writers find names, and a post that does well gets embedded and discussed on pages crawlers read. Our piece on why you need to be famous makes the full case. Fame is the engine that produces mentions. The mentions are what AI reads.

That is also the right way to think about any boost. If a new account with a near-empty count makes people bounce before they read your work, LitFame's growth services can help a profile look established enough for real visitors to stick around. Be clear-eyed about what that does. It does not put you in a ChatGPT answer, platform rules prohibit artificially inflated engagement, and bought engagement can be removed later. Treat it as a small nudge on top of the real work below.

Step 1: make your entity consistent everywhere

An "entity" is a thing a system can identify, like a person or a brand. If you are "Jess Park, illustrator" on your site, "jessdraws" on Instagram, "J. Park Studio" on Etsy and an old student bio on LinkedIn, a machine has to guess whether those are one person. Guessing means lower confidence, and lower confidence means you get left out.

The consistency checklist

1. Pick one name format and use it everywhere: your real name, your brand name, or "Name, from Brand." Same spelling, same capitalisation.
2. Write one bio sentence that says who you are, what you do and for whom. Example: "Jess Park is a Manchester illustrator who draws custom pet portraits and wedding stationery." Reuse it, lightly adapted, on every profile.
3. Use the same photo or logo everywhere.
4. Link everything to one home base, ideally your own domain, and link that home base back out to every profile.
5. Clean up the strays. Old portfolios, abandoned accounts and outdated job titles muddy the picture. Update or delete them.

Our guide to personal SEO and owning your name covers making the first page of Google for your name say what you want, which is exactly what AI pulls from when someone asks about you directly.

Step 2: get mentioned on the sites AI reads

Your own site is one voice. Answers to "who is good at X" questions tend to synthesise several sources, so you need other people describing you.

Where mentions come from for small creators

• Roundups and lists. "Best indie game devs to follow," "top small-batch candle makers in Leeds." Pitch writers who already publish these. Offer something useful, such as a quote, a free sample or a data point.
• Podcasts and newsletters. Guest spots usually come with show notes that name and link you. Niche shows are easier to land and the show notes are crawlable text.
• Community forums. Google's guide notes that its AI features "can show what's being said about products and services across the web, including in blogs, videos, and forum discussions." Helpful answers in your niche communities build a public trail under your name.
• Directories and marketplaces relevant to your field, from design platforms to local business listings. For local businesses, Google says a Google Business Profile helps you show up in AI responses and other Search results.

Skip the shortcuts. Google's guide says "seeking inauthentic mentions" is less helpful than it looks, because its AI features depend on both its ranking systems and its spam systems. Spammy guest posts and fake review sites get filtered. Real mentions compound.

Your own social profiles also count as searchable pages. Platform search and web search both read profile names, bios and captions, and our social SEO guide covers how to make those profiles findable across platforms.

Step 3: publish content worth citing

When an assistant retrieves your page, it is looking for sentences it can use. The GEO paper's best methods all point the same way: be specific, support your claims and write clearly.

What citable content looks like

• A clear answer near the top. If your page is about how much a custom logo costs, state your price range in the first paragraph, then explain it.
• Specific numbers you can stand behind. Your own data is gold: "Across the 140 portraits I sold last year, the most requested size was A4." Only use numbers you actually have.
• Named sources. When you state a fact you did not discover, say where it comes from. This is the Cite Sources method, and it builds trust with readers too.
• First-hand experience. Google's guide pushes "non-commodity content" and gives the example of a first-hand review versus a summary of what is already online. Your process, your mistakes, your results: nobody else can write those.
• Readable structure. Headings, short paragraphs, lists where they help. Google says there is no need to chop content into tiny chunks for AI, and no ideal page length. Write for the person reading.

Which pages to create first

For a person or small brand, a handful of pages does most of the lifting: an about page that states plainly who you are and what you have done, a services or work page with real examples, and two or three in-depth pages answering the questions clients ask you most. Our SEO guide for creators walks through picking those topics and structuring a creator website.

Skip the content farm instinct. Google's guide warns that creating separate pages for every query variation, primarily to manipulate rankings or AI responses, violates its scaled content abuse policy.

Step 4: structured data and crawler access

This technical layer cannot make weak content get cited. Getting it wrong can quietly lock you out, though.

Let the right crawlers in

Each AI company separates its bots by job. OpenAI's documentation lists OAI-SearchBot for ChatGPT search and GPTBot for model training, and says that sites which block OAI-SearchBot "will not be shown in ChatGPT search answers." Perplexity uses PerplexityBot for search. Anthropic lists Claude-SearchBot for search and ClaudeBot for training. Google's AI features use the regular Googlebot, and Google's crawler documentation says the separate Google-Extended token controls Gemini training and grounding in Gemini apps without affecting inclusion or ranking in Google Search.

So you can block training bots while allowing search bots. If discovery is your goal, allow the search bots, and check that a security plugin or CDN bot setting is not blocking them by accident.

Mark up who you are

Schema.org structured data lets you label your page in a machine-readable way: this page is about a Person named this, with these profiles. The key property for GEO-minded creators is sameAs, which schema.org defines as the URL of a page that "unambiguously indicates the item's identity." Listing your social profiles in sameAs is a direct way to say "all of these are me."

Keep expectations grounded. Google's own guide says structured data "isn't required for generative AI search" and there is "no special schema.org markup" for it, while recommending you keep using it for rich results. Structured data helps systems disambiguate you. It does not earn citations on its own.

About llms.txt

llms.txt is a proposed file, published at llmstxt.org by Jeremy Howard, that gives AI agents a curated markdown map of a site. Google's guide says Google Search does not use it. For a creator chasing visibility it is a low priority, and the setup guide linked above covers when it is worth an afternoon.

How to measure whether GEO is working

AI visibility is harder to measure than rankings, because answers vary between users, sessions and repeated runs of the same question. Combine a few imperfect signals.

Ask the assistants, on a schedule

Write down 10 to 20 questions a client, fan or employer might ask. Mix questions about you by name ("who is Jess Park the illustrator?"), category questions ("best pet portrait artists in Manchester") and problem questions ("how much does a custom pet portrait cost?"). Once a month, run them in ChatGPT, Perplexity, Gemini, Copilot and Google's AI Mode. Use a logged-out or fresh session where you can, since memory and personalisation skew results.

Record whether you are mentioned, whether you are linked, and which sources each answer cites. The cited sources are the real prize: they show which sites these systems trust in your niche, which makes them your next pitch list.

Check your referral traffic

When someone clicks a citation, the visit usually shows up in your analytics as a referral from the assistant's domain, such as chatgpt.com or perplexity.ai. OpenAI introduced a utm_source=chatgpt.com tag on ChatGPT links to make this easier, but Search Engine Land reported in October 2026 that the tag appears to have been dropped from some links, which can make ChatGPT visits show up as direct traffic. If your ChatGPT referrals suddenly fall while direct traffic rises on the same pages, suspect tracking before panicking.

Use the official reports

Google's help page notes the generative AI report may not appear until your site has enough impressions in those features. That is normal for a small site. Progress usually shows up first in direct questions about you by name, then slowly in category answers as mentions build.

Frequently Asked Questions

What is generative engine optimization in simple terms?

Generative engine optimization (GEO) is the practice of getting your website, profiles and name cited inside AI-written answers from tools like ChatGPT, Perplexity, Gemini and Google AI Overviews. It means being crawlable and indexed, describing yourself consistently, earning mentions on other sites and publishing specific content an AI can quote.

Is GEO different from SEO?

They overlap heavily. Google says AI Overviews and AI Mode are rooted in its core Search ranking systems, so normal SEO still applies. GEO adds focus on being quotable inside a written answer and on how you appear across many sources.

Can I pay to get my name into ChatGPT answers?

There is no documented way to buy placement in ChatGPT's organic answers, and OpenAI's help centre says placement is not guaranteed. Anyone selling guaranteed AI citations is overpromising. Allow OAI-SearchBot, build search visibility and earn genuine mentions instead.

Do follower counts help you show up in AI answers?

No AI company documents follower counts as a ranking signal. A big audience helps indirectly, because visible creators get interviewed, listed and linked, and those pages are what AI systems read. Followers make mentions easier to earn.

How long does GEO take to work?

Technical fixes can work fast: OpenAI says its systems take about 24 hours to adjust after a robots.txt change for OAI-SearchBot. Earning the mentions that put you into category answers takes months, which is why starting early matters.

Does the GEO paper prove adding statistics gets you cited?

It shows a direction, with limits. In the KDD 2024 paper by Aggarwal and colleagues, adding statistics, quotations and source citations raised a page's share of the answer by roughly 30 to 40 percent on one metric. Every tested page was already in Google's top five, and the tests used GPT-3.5-turbo, so read the figures as a direction to follow.

Should I block AI crawlers to protect my content?

That depends on your goal. Training bots and search bots are separate. OpenAI's documentation says its settings are independent, so you can block GPTBot while allowing OAI-SearchBot, and Google says the Google-Extended token does not affect inclusion in Google Search. Blocking search bots like OAI-SearchBot, PerplexityBot or Claude-SearchBot removes you from those assistants' search answers, which defeats the point if you want to be discovered.