What is generative engine optimization (GEO)? How to do it
GEO is getting your brand named and cited in the answers AI assistants write. Here is what it means, how it differs from SEO and AEO, what the research supports, and five steps to do it.
The seoscore team 10 min read
On this page
- What generative engine optimization means
- GEO vs SEO vs AEO: how they fit together
- How AI assistants decide which brands to name
- What the research says actually moves GEO
- Step 1: measure where you stand today
- Step 2: make your pages easy for AI to read and quote
- Step 3: earn mentions on the sources AI cites
- Step 4: keep entity signals and facts consistent
- Does llms.txt help GEO?
- Step 5: track GEO over time
- Common GEO myths
- Check your GEO with seoscore
- Next steps
Generative engine optimization (GEO) means getting your brand named and cited in the answers that AI assistants write. Those assistants include ChatGPT, Claude, Gemini, Perplexity and Google's AI Overviews and AI Mode. GEO has two halves. The first is making your pages easy for those engines to read and quote. The second is earning enough trusted mentions across the web that they choose to name you. You measure it by asking the assistants your customers' questions and recording how often, and how early, you appear.
What generative engine optimization means
A generative engine is a search system that writes an answer instead of handing you a list of links. GEO is the practice of shaping how your brand appears inside those written answers.
The term comes from a 2023 paper by Pranjal Aggarwal and co-authors from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi. It was peer-reviewed and presented at the KDD 2024 conference (Wikipedia summary, Search Engine Land). Since then the label has spread well beyond academia. People now use "GEO", "GEO SEO" or "AI SEO" for any effort to show up when someone asks ChatGPT, Claude, Gemini or Perplexity a question, or reads a Google AI Overview.
GEO vs SEO vs AEO: how they fit together
The three terms overlap, but each answers a different question.
| SEO | AEO | GEO | |
|---|---|---|---|
| Goal | Rank pages in search results | Make a page readable, trustworthy and quotable for AI engines | Get your brand named and cited in AI answers |
| Where you show up | Links on Google and Bing results pages | Quoted passages and cited links in AI answers | Inside the answers from ChatGPT, Claude, Gemini, Perplexity, AI Overviews and AI Mode |
| What is measured | Positions, impressions, clicks | Crawler access, readable content, markup, direct answers | Mention rate, position in the answer and citations, per engine |
| Main levers | Technical health, useful content, links | robots.txt, content that loads without JavaScript, structured data, answer blocks | Third-party mentions, consistent brand facts, quotable pages |
GEO builds on the other two rather than replacing them. Most assistants look things up on the web before answering, so a page that ranks badly or can't be crawled rarely gets picked up. Google says there are no additional requirements to appear in AI Overviews or AI Mode. A page only has to be indexed and eligible to be shown in Search with a snippet (Google: AI features and your website). Its later guide, Optimizing your website for generative AI features on Google Search, makes the same point more firmly. Search Engine Journal summarised it as saying AEO and GEO are still SEO.
AEO (answer engine optimization) is the page side of GEO. For that detail, read our article on answer engine optimization. This article concentrates on the brand side.
How AI assistants decide which brands to name
An assistant can learn about your brand in two ways.
- Training data. The model absorbed text about you when it was trained. This is slow to change. What the web said about you months or years ago shapes what the model "knows".
- Live retrieval. For many questions the assistant searches the web first, reads the results, and writes its answer from them. This is often called grounding. Google says AI Overviews and AI Mode may use "query fan-out", which means running several related searches across subtopics and data sources before writing the answer (Google).
With live retrieval, the pages an engine fetches decide which brands it mentions. Those pages include search results, review sites, comparison articles, forums, directories, Wikipedia and news. If those pages name three competitors and not you, the answer usually does the same.
Third-party pages matter a great deal. A 2025 study by Chen and colleagues looked at ChatGPT, Perplexity and Gemini. It found AI search shows "a systematic and overwhelming bias towards earned media", meaning third-party, authoritative sources, over brand-owned and social content. Google's ordinary results showed a more balanced mix (arXiv 2509.08919). It is a preprint, so treat it as a strong signal rather than settled fact.
Answers also differ between engines, because each one retrieves from different places. They can differ between two runs of the same question too. The same study found results were sensitive to how a question was worded and which language it was asked in. A single screenshot proves very little. You need repeated checks.
What the research says actually moves GEO
The original GEO paper tested nine ways of rewriting web content, using a benchmark called GEO-bench. Seer Interactive puts the benchmark at around 10,000 queries (Seer). The main findings (paper on arXiv):
- Overall, GEO methods raised visibility by up to 40% on diverse queries.
- Three changes worked best: citing sources, adding quotations and adding statistics. They gave a relative improvement of 30 to 40% on the paper's word-count-based visibility measure and 15 to 30% on its impression measure. The authors note these changes take little effort.
- The results held up on a live engine. In a test on Perplexity, visibility improved by up to 37%. Adding quotations did best on the word-count measure, with a 22% improvement (ar5iv version).
- Lower-ranked sites gained most. Sites further down the results page benefited more than those at the top (Search Engine Land). Seer reports that citing sources increased visibility by 115.1% for sites ranked fifth.
- Keyword stuffing did not help. In Seer's reading of the results, an authoritative tone made no significant difference either.
The paper also has limits. Most of its tests ran on a simulated engine built on a 2023 model, and today's assistants retrieve and write differently. It measured how much of a page's text made it into an answer, not whether a brand was recommended. Nobody has published solid evidence on how much any single tactic moves brand mentions in today's ChatGPT, Claude or Gemini. Treat the findings as good writing habits, not a formula.
Step 1: measure where you stand today
Start with a baseline. Without one you can't tell whether anything you change has worked.
- Write down the questions customers really ask. Use your sales calls, support tickets and site search, plus the searches you already rank for. Mix the types: "best [category] for [need]", "[you] vs [competitor]", "is [product] worth it", and problem questions that never mention a brand.
- Ask each assistant every question. Cover ChatGPT, Claude, Gemini and Perplexity. Ask each question at least twice, because answers vary between runs.
- Record four things for each answer:
- whether you were named
- how early you were named
- which competitors were named
- which sources were cited
A simple sheet is enough:
| Question | Engine | Named? | Position | Competitors named | Sources cited |
|---|---|---|---|---|---|
| Best invoicing tool for freelancers | Perplexity | No | – | A, B, C | review site, forum thread |
Doing this by hand gets slow once you have more than a dozen questions. AI visibility in seoscore does it for you. It asks ChatGPT, Claude, Gemini and Perplexity the questions your customers ask and scores how often, and how early, they name your site. It keeps every answer and the sources each engine cited. It is on every plan, including Free.
Step 2: make your pages easy for AI to read and quote
This is the AEO half of GEO. In short:
- Let AI search crawlers in. Check robots.txt for the search crawlers, which are separate from the training crawlers. OpenAI, for example, lists OAI-SearchBot for ChatGPT search and GPTBot for training (OpenAI crawlers). Blocking a training crawler and allowing a search crawler is a common setup:
User-agent: OAI-SearchBot
Allow: /
User-agent: Claude-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
- Make content readable without JavaScript. Key text should be in the HTML the server sends, not only drawn in later by a script.
- Answer the question near the top in two or three plain sentences, then give the detail.
- Add statistics, quotations and cited sources where they are genuine. This is the change the GEO paper found most useful.
- Use structured data that matches the page. Google's guidance for AI search asks that structured data match the visible content.
- Show dates. Put a published or updated date on pages where freshness matters.
- Check your snippet rules. Google says nosnippet, data-nosnippet, max-snippet and noindex limit how your content can appear in its AI features.
The free AEO and GEO checker runs these checks on any page. It covers AI crawler access, content without JavaScript, snippet rules, entity and FAQ markup, freshness, citations and direct answers. Run it on your home page, your main product pages and the pages engines cited for competitors in Step 1.
Step 3: earn mentions on the sources AI cites
Go back to the sources column from Step 1 and sort them into review sites, comparison articles, forums, directories, news and Wikipedia. Those are the pages shaping the answers, so they are where to spend effort.
- Reviews and directories. Claim your listings, fix wrong details and ask real customers to leave reviews.
- Comparison and "best of" articles. Contact the authors with accurate, current facts and an offer to answer questions. Don't pay for placement disguised as editorial.
- Forums and communities. Take part as yourself, answer questions properly and say who you are when you mention your product.
- News and industry sites. Original data, such as a survey or figures from your own product, gives journalists a reason to cite you.
Google's guide warns against pursuing inauthentic mentions (Google). Fake reviews and planted forum posts are a risk, not a strategy.
AI mentions helps here. It reports where Google AI Overviews and ChatGPT cite your site, which of your pages they use, the questions people asked and the other sites cited beside you. That list of other sites is your outreach list. AI mentions is on Starter and up, with a free trial.
Step 4: keep entity signals and facts consistent
An entity is a thing a search engine or model can identify on its own terms: your company, a product, a person. Models connect your brand to a category more reliably when every source describes it the same way.
- One description, everywhere. Write a one-sentence description of what you do and who it is for. Use it on your site, social profiles, directories and press pages.
- A real about page. Include your name, what you make, where you are, who runs it and how to contact you.
- Up-to-date product facts. Keep features, availability and pricing pages current. Old facts get repeated in answers.
- Organization markup with sameAs links to your official profiles:
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Example Ltd",
"url": "https://www.example.com",
"logo": "https://www.example.com/logo.png",
"description": "Invoicing software for UK freelancers.",
"sameAs": [
"https://www.linkedin.com/company/example",
"https://github.com/example"
]
}
Does llms.txt help GEO?
llms.txt is a proposed plain-text file at the root of a site that gives AI tools a short, curated map of your most useful pages. Google has said plainly that Google Search does not use it. Creating one will neither help nor harm your visibility in Google, including its AI features (Google). We have not seen any confirmation that OpenAI, Anthropic or Perplexity read it when answering questions. Some of them publish llms.txt files on their own sites, but that is not the same thing.
If you want one anyway, it takes minutes. The free llms.txt generator creates and checks the file. Just don't expect it to move your GEO numbers on its own. Steps 1 to 4 matter far more.
Step 5: track GEO over time
Re-run your question set at least monthly, because answers drift as models update and the web changes. Watch these metrics:
| Metric | What it tells you |
|---|---|
| Mention rate per engine | How often each assistant names you across your questions |
| Position in the answer | Whether you are named first or as an afterthought |
| Citation share | How often your pages are among the cited sources |
| Cited pages | Which of your pages engines rely on, and which they ignore |
| Competitor share | Who is gaining the answers you are missing |
AI visibility keeps these runs so you can compare each one with your baseline. Look at Google AI Overviews next to your ordinary rankings. Rank tracking (Starter and up, with a free trial) shows your position for each keyword. It also shows whether Google displayed an AI Overview and cited your site. In Search Console, clicks from AI Overviews and AI Mode are counted within your overall search traffic (Google).
Common GEO myths
- "GEO replaces SEO." It doesn't. Retrieval leans on search results, and Google says its AI features need no extra requirements beyond normal indexing and snippet eligibility.
- "You can pay or prompt your way into answers." You can't buy a place in an assistant's answer. Asking the assistant to recommend you won't change what it tells other people.
- "Stuff in keywords for the AI." Keyword stuffing did not improve visibility in the GEO research. Clear answers, real figures and cited sources did.
- "Optimise once and you're done." Answers change between runs and over time. GEO needs regular measurement, like rankings do.
Check your GEO with seoscore
Start with AI visibility. Add your customers' questions and get a baseline across ChatGPT, Claude, Gemini and Perplexity, with every answer and cited source. Then run the AEO and GEO checker on the pages that should be cited. AEO is also 15 of the 100 points in every seoscore audit. When you want to see where Google AI Overviews and ChatGPT already cite you, and who sits beside you, add AI mentions. Our GEO guide covers how seoscore measures and scores it.
Next steps
- List 20 or more real customer questions this week and record a baseline for each engine.
- Fix crawler access, server-rendered content and direct answers on your five most important pages.
- Sort the sources engines cite into types and pick three to work on first.
- Write your one-sentence brand description and use it everywhere, including your Organization markup.
- Put a monthly recheck in the diary and compare each run with your baseline.
Questions people ask
How long does GEO take to show results?
Changes that affect live retrieval can show up once engines re-crawl the pages they cite, which can take days or weeks. What a model learned in training changes much more slowly, often only when a new version is released. Plan in months and judge progress against your baseline, not a single answer.
Can a small site compete with big brands in AI answers?
Yes, especially on narrow questions where the big brands have little specific content. Assistants answer very particular questions, and a site that answers one clearly with real figures and sources can be cited for it. Pick the questions where you have genuine expertise rather than the broadest category terms.
Can I see AI referral traffic in Google Analytics?
Partly. Visits from assistants such as ChatGPT, Perplexity or Claude usually arrive as referrals, and you can group them with a custom channel that matches those domains in the referrer. Clicks from Google AI Overviews and AI Mode are counted inside ordinary Google organic traffic, so you can't separate them in analytics.
Should I block AI crawlers?
Decide separately for training crawlers and search crawlers. Blocking a training crawler such as GPTBot keeps your content out of future model training. Blocking a search crawler such as OAI-SearchBot can stop your pages appearing in that assistant's answers. If you want to be named in AI answers, keep the search crawlers allowed.
Sources
- Generative engine optimization en.wikipedia.org
- Generative Engine Optimization framework introduced in new research searchengineland.com
- AI Features and Your Website developers.google.com
- Google's Guide to Optimizing for Generative AI Features on Google Search developers.google.com
- Google's New AI Search Guide Calls AEO And GEO 'Still SEO' www.searchenginejournal.com
- Generative Engine Optimization: How to Dominate AI Search Mahe Chen∗ arxiv.org
- Optimizing Content for Generative Search Resulted in +40% Visibility www.seerinteractive.com
- GEO: Generative Engine Optimization arxiv.org
- [2311.09735] GEO: Generative Engine Optimization - ar5iv ar5iv.labs.arxiv.org
- platform.openai.com/docs/bots platform.openai.com
- Top ways to ensure your content performs well in Google's AI experiences on Search | Google Search Central Blog | Google for Developers developers.google.com
seoscore
Do AI assistants mention you?
Track how often ChatGPT, Claude and Gemini name your site, and what to change.
Get my free score