What GEO is
Generative engine optimisation (GEO) is the work of improving how visible and trusted your content is in answers that generative engines write. A generative engine uses a large language model to gather and summarise information from several sources.[1] Aggarwal and colleagues introduced the term in a 2023 paper, which they later presented at the 30th ACM SIGKDD conference in 2024.[1]
We split the work this way at Vyzpro. AEO covers how you build your own pages so a reader can lift an answer from them. GEO covers the wider ground: whether AI systems can reach you, what they find about you across the web and whether they trust it. The industry uses both terms in overlapping ways, so treat our split as a working definition.
What the research found
The authors built GEO-bench, a benchmark of varied queries paired with web sources, and tested ways of rewriting content to see how much a generative engine used it.[1] They report that GEO methods raised visibility by up to 40% in generative engine responses, and by up to 37% in a real-world test on Perplexity.ai.[1]
The strategies that performed best were citing sources, adding quotations from credible sources, adding statistics and improving fluency. Keyword stuffing, a classic SEO tactic, did not help. The effect of each strategy varied by subject area.[1]
Read the numbers with care. It’s one study, run in 2023 and 2024 on engines that have changed since. The authors measured visibility inside their own framework, and sales weren’t part of it. “Up to 40%” is a best case. Use the findings as direction: concrete, sourced writing beats padding.
How generative engines find your content
Before an engine can quote you, its crawlers have to reach your pages. Each company publishes separate crawlers with separate jobs.
- OpenAI. OAI-SearchBot surfaces sites in ChatGPT’s search features, and ChatGPT search answers won’t show sites that opt out of it. GPTBot crawls content that OpenAI may use to train its models. ChatGPT-User handles requests that people make inside ChatGPT, and because users trigger those visits, robots.txt rules may not apply.[2]
- Perplexity. PerplexityBot surfaces and links websites in Perplexity search results, and Perplexity says it doesn’t use that bot to crawl for AI foundation models. Perplexity-User fetches pages when a person asks a question. A user triggers that visit, so Perplexity says robots.txt rules often don’t apply.[3]
- Anthropic. Anthropic documents its ClaudeBot crawler and shows site owners how to block it in robots.txt.[4]
- Google. AI Overviews and AI Mode rely on the same Search crawling as ordinary results. Google-Extended limits how Google uses your content for training and grounding in its other systems.[5]
Showing up in AI search and feeding model training are separate decisions. A business that wants a mention in answers wants the search crawlers allowed. Whether to allow the training crawlers is your call. Each company’s page above explains what its controls do.
A practical GEO checklist
- Let the right crawlers in. Check robots.txt and any firewall, security plugin or CDN setting. Google advises allowing crawling in robots.txt and through any CDN or host,[5] and the same logic applies to other crawlers.
- Serve content as plain HTML. Crawlers that don’t run scripts see only the page source.
- Write quotable facts and name the source. The research found that cited sources, quotations and statistics helped.[1] Date your figures.
- Keep your business details consistent. Match your name, address, phone, hours, credentials and reviews across your site, your Google Business Profile, directories and review sites. Mismatches make a business harder to recognise and trust. That’s our practice, and a Vyzpro audit checks it.
- Show who stands behind the content. Google says trust is the most important part of E-E-A-T.[8] Real authors, real contact details and evidence matter to readers and to the systems that learn from your pages.
- Treat llms.txt as optional. It’s a proposal for pointing AI tools to clean versions of your pages,[6] and Google says its AI features need no AI text files.[5]
- Measure. Ask real buyer questions in the assistants, note who they name and repeat the check each month.
Why GEO matters now
Pew Research Center followed the browsing of 900 US adults in March 2025. On Google results pages with an AI summary, people clicked a traditional result link in 8% of visits, against 15% on pages with none. They ended their session on 26% of pages with a summary and 16% without.[7]
If fewer people click, the answer carries more weight. A buyer who reads your name in it has found you. GEO doesn’t replace SEO. Your pages still need to be crawlable, readable and indexed, and Google says the fundamentals apply to its AI features.[5]
How Vyzpro covers GEO
Every Vyzpro audit covers GEO. We check crawler access, whether your pages work without scripts, who appears for buyer-style searches in your market and where competitors get named in your place. We also check your business footprint across Google Business Profile, directories and review sites. If public information can’t settle a point, the report says so and marks the finding “to verify”.
No audit can promise a citation. A good one shows what stops AI systems reaching and trusting you, and ranks the fixes. See the AI search audit, look inside our report or check pricing.