Generative Engine Optimization (GEO): What Actually Works

GEO is the work of getting your business named and cited in answers from ChatGPT, Gemini, Claude and Google's AI features. Here is what the research behind the term really showed, how GEO differs from AEO and SEO, and a checklist you can work through this month.

AmIPicked · 13 min read · updated

Generative engine optimization (GEO) is the practice of making your business more likely to be mentioned, recommended or cited when AI systems such as ChatGPT, Gemini, Claude, Perplexity or Google AI Overviews answer a question. The core of it is not new: your site has to be crawlable and rank in normal search, and other trusted sites have to describe you clearly. What is new is how you measure success, because the goal is a mention inside an answer rather than a blue link on page one.

This guide covers where the term came from, what the research actually found (it is often misquoted), how GEO compares with AEO and SEO, which tools are worth using, and a checklist a small business can work through without an agency.

What is generative engine optimization?

A generative engine is any system that searches for information and then writes an answer instead of showing a list of links. ChatGPT with search, Google AI Mode and AI Overviews, Perplexity, Gemini and Microsoft Copilot all work this way. They run one or more searches in the background, read a handful of pages, and write a summary that may name businesses and link to sources.

GEO is everything you do to make your business one of those named businesses or linked sources. For a dentist in Austin, that means showing up when someone asks ChatGPT "which dentist in Austin is good for nervous patients?" For a bike shop in Leeds, it means being on the list when someone asks Gemini "where can I get a road bike serviced in Leeds this week?"

Wikipedia's article on the topic notes that GEO is also called answer engine optimization (AEO) or AI optimization (AIO), and that the terms are used inconsistently. That inconsistency matters less than people think, because the practical work overlaps almost completely.

Where does the term GEO come from?

The term comes from a research paper, "GEO: Generative Engine Optimization", by Pranjal Aggarwal (IIT Delhi), Vishvak Murahari, Karthik Narasimhan and Ameet Deshpande (Princeton University), with Tanmay Rajpurohit and Ashwin Kalyan. It was first posted on arXiv on 16 November 2023 and later accepted at the KDD 2024 conference.

The authors built a benchmark called GEO-bench: 10,000 queries drawn from nine sources, including MS MARCO, Natural Questions, ELI5 and queries from Perplexity's Discover page. For each query they took the top five Google results, rewrote one of the pages using a given method, and measured how much of that page ended up in the AI-generated answer.

What did the GEO paper actually find?

The paper tested nine rewriting methods. On its main metric, which counts how many words from a source appear in the answer and weights earlier positions more heavily, the results on the test set were:

MethodScore (baseline 19.5)
Quotation addition27.8
Statistics addition25.9
Fluency optimization25.1
Cite sources24.9
Technical terms23.1
Easy-to-understand22.2
Authoritative tone21.8
Unique words20.7
Keyword stuffing17.8

Four findings are worth remembering:

  1. Adding quotations, statistics and citations to sources helped most. The paper reports a best-case improvement of about 40 percent in visibility on its benchmark.
  2. Keyword stuffing made things worse. It scored below the unmodified page, and in the tests on Perplexity it was about 10 percent worse than baseline.
  3. Lower-ranked pages gained the most. A page in fifth place in the search results gained about 115 percent from citing sources, while the page in first place actually lost visibility. GEO helped smaller sites more than it helped the leaders.
  4. What works depends on the topic. Citing sources worked best for factual and legal questions, quotations for history and people, statistics for law, government and opinion questions.

What the GEO paper did not show

The paper is often quoted as "GEO increases visibility by 40%", which overstates it. The tests were run on a research setup built on GPT-3.5 and on a 200-query sample on Perplexity, where the pages were uploaded as files because Perplexity does not accept source URLs. The paper did not test ChatGPT search, Google AI Overviews or Gemini as they work today, and it did not measure whether the rewritten pages would still rank in Google. It also did not study local businesses or product recommendations.

So the honest takeaway is narrow: once a page is already among the sources an AI reads, content that contains quotable facts, numbers and references gets used more. It says nothing about how to get into that source list in the first place. That part is still mostly SEO, listings and reputation.

What is answer engine optimization (AEO)?

Answer engine optimization is the older term. It grew out of featured snippets and voice assistants, where the goal was to have your page supply the one direct answer to a question. Today people use AEO and GEO almost interchangeably.

Where companies do separate them, the definitions differ. Microsoft Advertising's January 2026 guide, "From Discovery to Influence", describes AEO (answer or agentic engine optimization) as making content and product information easy for AI assistants and agents to retrieve and present as direct answers, and GEO as making content discoverable and persuasive inside generative AI systems through clarity, trust and authority. Profound treats the two as one practice. Google's own guide simply names both terms and then says they are part of SEO.

GEO vs AEO vs SEO: what is the difference?

SEOAEOGEO
Main goalRank pages in search results and earn clicksHave your content supply the direct answer to a questionBe mentioned, recommended or cited inside AI-written answers
Where it showsGoogle and Bing results pagesFeatured snippets, voice assistants, AI answer boxesChatGPT, Gemini, Claude, Perplexity, Copilot, Google AI Overviews and AI Mode
Unit of successPosition and clicksBeing the quoted answerShare of answers that name you, and how you are described
Typical query"dentist austin""how much is a dental crown in texas""which dentist in Austin is best for nervous patients and takes Delta Dental?"
Key inputsCrawlability, relevance, links, page qualityClear question-and-answer content, structured dataEverything in SEO, plus consistent mentions on review sites, directories, news and forums
Main free measurementGoogle Search Console, Bing Webmaster ToolsSearch Console, manual checksBing AI Performance report, Search Console generative AI report, manual prompt tests

The table shows the real difference. SEO is about your pages. GEO is about your business as an entity: what the web as a whole says about you, because an AI answer often blends your site with Yelp reviews, a local news article and a Reddit thread.

Is generative engine optimization different from SEO?

Less than vendors suggest. Google's guide to optimizing for generative AI features, published on 15 May 2026, says that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." The same guide says you do not need llms.txt or other AI text files, special schema markup, chunked content or writing in a special style to appear in Google's AI features. To be eligible, a page must be indexed and able to show in Google Search with a snippet.

Other engines work on the same foundations. OpenAI's crawler documentation says that sites that block OAI-SearchBot will not be shown in ChatGPT search answers, and OpenAI's help pages say ChatGPT search draws on third-party search providers, including Bing. If a page is not crawlable or not in the indexes these systems use, no amount of rewriting will help.

Where GEO does add something:

  • You measure differently. There is no rank position to track. You track how often you are mentioned across many runs of the same question, because the answers change every time (more on this in how to check AI visibility).
  • Off-site mentions carry more weight. A recommendation answer is assembled from several sources. If three review sites and a local newspaper describe a competitor and nobody describes you, the AI has nothing to say about you.
  • Accuracy of description matters. An AI may name you but describe you wrongly, for example with old opening hours or a service you stopped offering.

How to do generative engine optimization: a practical checklist

This is the order we would work in for a small business. The first block is technical and takes an afternoon. The rest is ongoing.

1. Make sure AI systems can read your site

  • Check robots.txt for OAI-SearchBot (ChatGPT search), Googlebot, Bingbot, PerplexityBot and ClaudeBot. You can block GPTBot (training) and still allow OAI-SearchBot (search). Our AI crawler checker reads your robots.txt and shows which bots are allowed, and the GPTBot and robots.txt guide explains each one.
  • Check that your firewall or CDN does not block these bots. Aggressive bot protection settings can block crawlers even when robots.txt allows them.
  • Make sure key text (services, prices, location, hours) is in the HTML, not loaded only by JavaScript.
  • Verify your site in Google Search Console and Bing Webmaster Tools and submit a sitemap. Bing matters more than it used to, because ChatGPT search uses it as one of its sources.

2. Make your pages easy to quote

  • Answer the main question of each page in the first two or three sentences.
  • Add concrete facts: prices or price ranges, service areas, opening hours, years in business, certifications. These are the "statistics" that the GEO paper found useful.
  • Cite sources when you make claims, and quote real people where it fits (a named staff member, a customer review with permission). The paper's two strongest methods were quotations and statistics.
  • Create one page per important service rather than one page listing everything. "Emergency boiler repair in Leeds" deserves its own page if it is a core service.
  • Add an FAQ with the questions customers really ask you on the phone.
  • Use structured data (LocalBusiness, Organization, Product, FAQ) because it helps search engines, but do not expect it to work as a GEO switch. Google says it is not required for its AI features.

3. Build the off-site picture

  • Claim and complete your listings: Google Business Profile, Bing Places, Apple Business Connect, Yelp, and the directories that matter in your industry (for example, Healthgrades for doctors or Checkatrade in the UK).
  • If you are an eligible US business, look at OpenAI's Business Hub, launched as a beta in November 2026, which lets you claim the profile ChatGPT shows for your business. Our ChatGPT ranking guide covers what it does and does not change.
  • Keep your name, address, phone number and description identical everywhere.
  • Ask happy customers for reviews regularly, and reply to them.
  • Look at which sites AI answers cite for your category and try to get listed or mentioned there: "best of" lists, local news, trade associations, industry blogs.

4. Measure, then repeat

  • Write 10 to 20 questions a customer would ask, run them in ChatGPT and Gemini with search on, and record whether you are mentioned, who is mentioned instead, and which sources are cited.
  • Repeat monthly with the same questions.
  • Check the free reports: Bing Webmaster Tools has an AI Performance report (public preview since February 2026) showing citations in Copilot and Bing AI answers, and Google Search Console has a generative AI performance report.

What about llms.txt?

llms.txt is a proposed file that lists your most important pages in Markdown for language models. It is cheap to create and harmless, but no major AI company has confirmed that its search product uses it, and Google says its Search ignores it. Treat it as optional. If you want one, our llms.txt generator creates it in a minute, and the llms.txt guide explains the format and its limits.

What does not work in GEO

  • Keyword stuffing. The GEO paper measured it directly and it reduced visibility.
  • Mass-produced AI pages. Google's guide warns that producing pages at scale to target AI features can violate its scaled content abuse policy.
  • Hidden text or instructions aimed at AI. Prompt-injection tricks in white text are a reputational risk and search engines treat hidden content as spam.
  • Buying "guaranteed ChatGPT placement". Google's guide says no third-party tool has access to its internal ranking or AI systems, and OpenAI says ChatGPT search ranking cannot be guaranteed. Anyone promising a fixed position is guessing.
  • Chasing a single answer. One screenshot of ChatGPT naming you, or not naming you, proves very little. Answers change from run to run.

Which generative engine optimization tools are worth using?

Start with the free ones:

  • Bing Webmaster Tools AI Performance shows how often your pages are cited in Copilot and Bing AI summaries and for which queries. It does not cover ChatGPT, Gemini or Google.
  • Google Search Console has a generative AI performance report for Google's AI features.
  • A spreadsheet and the AI apps themselves. Slow, but free and transparent.
  • AmIPicked, our free AI visibility test, takes your website, writes five questions your customers would ask, runs them in ChatGPT and Gemini with live web search, and shows how often you are mentioned, your average position, your share of answers against competitors, how the AI describes you and which sources it cites. It also checks robots.txt, llms.txt and structured data.

Paid trackers make sense when you need dozens of prompts tracked weekly across several engines. Semrush's AI Visibility Toolkit and Ahrefs Brand Radar are good if you already use those SEO suites. Peec AI, Profound and Otterly.AI are dedicated trackers with stronger competitor and prompt-level reporting; Profound is aimed at larger brands. We compare them in the best AI visibility tools and Peec AI alternatives. Whatever you choose, ask the vendor how many times each prompt is run and how they calculate their score. Research by SparkToro and Gumshoe in early 2026 found that ChatGPT and Google's AI returned the same list of brands less than 1 percent of the time across repeated runs, so a tool that runs each prompt once is reporting noise.

A worked example: a bike shop in Leeds

Say a bike shop in Leeds runs five questions in ChatGPT and Gemini and is mentioned in one answer out of ten. Two competitors appear in most answers, and the cited sources are Google Maps listings, a "best bike shops in Leeds" article on a local blog, and Yelp.

The GEO plan follows directly from that. Confirm OAI-SearchBot and Bingbot can crawl the site. Add a service page for road bike servicing with prices and turnaround time. Fill in Bing Places and Yelp with the same details as Google Business Profile. Ask the last 30 customers for a review. Contact the local blog with something worth writing about, such as a free bike check day. Then run the same five questions next month and compare.

None of these steps is exotic. That is the main lesson of generative engine optimization: the businesses AI recommends are usually the ones that are easy to find, clearly described and talked about by other people.

If you want a baseline before you start, run the free AmIPicked test on your site. It takes a few minutes and gives you a starting number to beat.

Frequently asked questions

What is generative engine optimization?

+

Generative engine optimization (GEO) is the practice of making your business more likely to be mentioned, recommended or cited in answers written by AI systems such as ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews. The term comes from a 2023 research paper by Aggarwal and colleagues from Princeton and IIT Delhi. In practice it combines classic SEO with work on how your business is described across the web.

What is the difference between GEO and AEO?

+

There is no agreed definition that separates them. Answer engine optimization (AEO) usually means getting your content used as the direct answer to a question, while GEO usually means being included and cited in longer AI-generated responses. Many vendors and Profound treat them as the same thing, and Google calls both of them part of SEO.

Is GEO replacing SEO?

+

No. Google's May 2026 guide on generative AI features says that optimizing for generative AI search is still SEO, because its AI features are built on its core ranking systems. ChatGPT search also relies on web search results and its own crawler, so a site that cannot be crawled or ranked has little chance of being cited.

What did the original GEO paper find?

+

The paper tested nine ways of rewriting a web page and measured how much of the page made it into AI answers. Adding quotations, statistics and source citations raised visibility by roughly 30 to 40 percent on its benchmark, while keyword stuffing made results worse. The gains were largest for pages that ranked low in the underlying search results.

Do I need special tools for generative engine optimization?

+

Not to start. You can test your visibility by asking AI assistants the questions your customers ask, and Bing Webmaster Tools and Google Search Console now include free reports on AI citations and AI features. Paid trackers such as Semrush, Ahrefs Brand Radar, Peec AI, Profound and Otterly.AI help when you need many prompts tracked over time.

How long does GEO take to show results?

+

Technical fixes, such as unblocking OpenAI's search crawler, can take effect within about a day according to OpenAI's crawler documentation. Changes that depend on new reviews, listings and articles about your business usually take weeks or months. Measure monthly with the same set of questions so you can see the trend.

Read next