GEO: what generative engine optimization is, and what works
GEO stands for generative engine optimization. It means improving how often, and how prominently, your content shows up inside answers written by AI systems that search the web. You'll also see it called answer engine optimization (AEO) or AI SEO. The labels overlap almost completely.
This post explains where the term came from, what the research behind it found, where Google disagrees with the hype, and what a small business can do about it this week.
Where the term came from
"Generative Engine Optimization" is the title of a research paper by Pranjal Aggarwal and colleagues from Princeton University, IIT Delhi and independent researchers, first posted in November 2023 and published at the KDD 2024 conference. The authors define a generative engine as a system that retrieves sources from the web and uses a language model to write a single answer grounded in those sources, with inline citations. GEO, in their framing, is how content creators can improve their visibility in those answers without knowing how the engine works inside.
Nobody outside OpenAI, Perplexity or Google can see how their systems choose sources. GEO is black-box work: change something, measure, repeat.
How a generative engine picks what to cite
The paper describes a common pipeline. The engine rewrites the user's question into simpler search queries, fetches the top results, and has a model write an answer that cites them. OpenAI's web search documentation describes the same shape for its API: the model searches, then returns text with url_citation annotations for the pages it used. It also says the full list of sources consulted is usually longer than the list of citations. Being read is not the same as being cited.
Because the answer is prose, visibility isn't a rank. The paper measures it as the share of an answer's words tied to a citation, weighted toward the start of the answer, since readers see that part first.
What the research found
The authors tested nine ways of rewriting a source page against a benchmark of 10,000 queries they call GEO-bench. Their main findings:
- Adding citations to credible sources, quotations and statistics worked best. The paper reports that these methods can raise visibility by up to 40% on its metrics.
- Keyword stuffing, a classic SEO habit, offered little to no improvement. In a test on Perplexity, it did 10% worse than leaving the page alone.
- Lower-ranked pages gained the most. Adding citations raised visibility by 115.1% for pages ranked fifth in search, while pages ranked first lost 30.3% on average.
- Results varied by topic. What helped for history questions was not what helped for law or opinion questions.
Read these numbers with care. The tests used a lab engine built on gpt-3.5-turbo and the top five Google results, plus a smaller test on Perplexity. Engines have changed a lot since 2023. The direction is useful: content with specific, sourced facts does better than vague or keyword-padded copy. The percentages are not a promise for your site.
What Google says about GEO
Google published a guide in May 2026 on optimizing for its generative AI features. Its position is that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." For Google Search, it says you can ignore tactics like "chunking" content, creating unnecessary AI text files such as llms.txt, and pursuing inauthentic mentions. It also says no special schema.org markup is required.
Google is one engine, though. OpenAI and Perplexity run their own crawlers with their own robots.txt rules, and some agents and developer tools read llms.txt. Our reading is that the parts everyone agrees on are the parts worth doing first: be crawlable, state your facts plainly, and earn real coverage on other sites.
GEO in practice: our own 0/100
We ran our own report on our previous domain, showsupin.com, on October 15, 2026, the day after we registered it. It asked 25 questions, 3 times on each engine, through the OpenAI API with web search and the Perplexity API, for 150 answers in total. The site scored 0 out of 100: no answer mentioned us and none cited our domain.
Splitting the questions shows two different problems.
- Category questions (14 questions, 84 answers), such as "Most affordable AI search visibility (AEO/GEO) tool": at least one of the three competitors we tracked was named in 49 of the 84 answers. We were named in none. That is a coverage problem. For these questions the most cited sources were writesonic.com (20), scrunch.com (18), superframeworks.com and thebrandhopper.com (16 each) and frase.io (14). We weren't on any of them.
- Questions using our name (11 questions, 66 answers), such as "How much does ShowsUpIn cost?": 28 of the 66 answers brought up a different product, ShowUpInAI: 16 answered as if we were it, and 12 asked whether we meant it. That is a naming problem, and it's why we renamed to Answer Checkup.
A GEO plan for a new business has to deal with both.
A GEO checklist for a small business
- Open the door to search crawlers. Allow OAI-SearchBot and PerplexityBot in robots.txt and in your firewall rules. Both companies document these bots and say changes can take up to about a day.
- Write facts, not adjectives. Prices, service areas, hours, numbers, dates. The paper's best methods added statistics and sources. Do the same on your own pages, and link to where each number comes from.
- Mark it up. Add JSON-LD for Organization, Product with Offer, Article and FAQPage, and only for content that's visible on the page.
- Add llms.txt. A short markdown summary at
/llms.txtwith links to your key pages. It's optional, cheap and aimed at agents. Don't expect it to change Google results. - Answer real questions on real pages. One clear answer per question, in the words buyers use.
- Earn third-party coverage. Find the review sites and comparison articles that answer engines cite for your category, and give their authors accurate information. Never fake reviews or mentions.
- Keep one name. Same spelling on your site, listings and profiles. If your name is close to someone else's, the engines may merge you.
- Measure on a schedule. The same questions, asked several times, every month. Our post on checking whether ChatGPT recommends your business shows how.
FAQ
- Is GEO the same as SEO?
- They share almost all their foundations. Google calls GEO "still SEO." The practical difference is what you measure: mentions and citations inside AI answers instead of positions in a list of links.
- Is GEO real or just a buzzword?
- The term comes from peer-reviewed research presented at KDD 2024. Plenty of advice sold under the label goes further than that research supports, so ask for evidence.
- Does keyword stuffing work for GEO?
- No. In the GEO paper it gave little to no gain, and on Perplexity it did 10% worse than the unedited page.
- What is GEO-bench?
- A set of 10,000 queries from nine sources, released with the GEO paper so researchers could test how page changes affect visibility in generated answers.
- How do I measure GEO results?
- Track a fixed list of buyer questions over time. For each answer, record whether you are named, whether your site is cited, and which other sites are cited.
Check your own site
Run the free scan on the Answer Checkup homepage to see whether AI crawlers can read your site and whether your structured data and llms.txt are in place.
Answer Checkup is not affiliated with OpenAI, Perplexity or Google.
Contact: hello@answercheckup.com
Sources
All pages fetched and checked on October 15, 2026.
- Aggarwal et al., "GEO: Generative Engine Optimization," KDD 2024: https://arxiv.org/abs/2311.09735
- Our own report on showsupin.com, scanned October 15, 2026 at 12:47 AM ET (25 questions, 2 engines, 150 answers)
- Google Search Central, optimizing for generative AI features: https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Google Search Central, AI features and your website: https://developers.google.com/search/docs/appearance/ai-features
- OpenAI, web search tool guide: https://developers.openai.com/api/docs/guides/tools-web-search
- OpenAI, overview of OpenAI crawlers: https://developers.openai.com/api/docs/bots
- Perplexity, Perplexity crawlers: https://docs.perplexity.ai/docs/resources/perplexity-crawlers
- The llms.txt proposal: https://llmstxt.org/