What is generative engine optimisation (GEO), and is it real?

Generative engine optimisation (GEO) means improving content so it appears more often in AI-generated answers. Yes, the work is real. Researchers have tested changes that improved visibility, and Google publishes guidance for its AI search features. That supports doing the work, but gives no fixed uplift to expect from it.

What does GEO actually involve?

GEO is about getting content into a generated answer. The original research defines its purpose as increasing a content creator’s visibility in those answers.

For Google, established SEO practices still matter because its AI search features use the core Search ranking and quality systems. Google also asks for original content that adds expert knowledge or experience beyond what is already commonly known.

We would expect a GEO proposal to name the questions your company needs to answer and what its content will add. Ask how the work will help the reader before accepting a new label as a reason to spend.

What evidence says GEO works?

The GEO paper reports visibility gains of up to 40% in its tests, with results varying by subject. The experiment used gpt3.5-turbo to generate answers, and the authors say methods may need to change as engines evolve. The result shows what worked in that experiment. It is not a forecast for your company.

For any proposed improvement, we would ask what changed, which engine was tested and what the result measured. Then test it against the questions that matter to your business.

Is GEO just SEO with a different name?

Google’s guidance gives SEO and GEO plenty of common ground. SEO practices still matter, and Google requires no special schema.org markup for generative AI search. These are the machine-readable labels added to a page.

But it is still worth examining the AI answers themselves. Google says AI Overviews and AI Mode may run several related searches to build an answer. So the words someone types can be only part of what Google searches for.

Ahrefs compared 15,000 long-tail queries across conventional search and AI assistants. Of the URLs Perplexity cited, 28.6% also appeared in Google’s top ten in that study. Ahrefs collected the data in early July and says results may have changed. It is a measured example of the difference, not a current benchmark.

What should you measure?

Open full-size screenshot in a new tab

Select the screenshot to enlarge it.

Read the labels before the trend

The example plots citations and cited pages. Its example numbers are not a benchmark for your company.

Keep the business question beside the report

Read visibility alongside enquiries and revenue, and check whether the answer actually recommends your company.

Source: Microsoft Bing Webmaster Blog, official example

Google points publishers to the Generative AI performance report in Search Console to measure how content performs in its generative features. It also suggests tracking conversions and time on site with tools such as Google Analytics.

Bing’s AI Performance Report covers Microsoft Copilot, Bing and selected partner AI experiences. Its Citation Share measure records what appears, but Bing says it does not measure traffic share or score content quality.

We would read those visibility measures alongside enquiries and revenue. Check whether the answer mentions your company, cites your content or recommends you for the job. There’s a difference between showing up and being recommended.

How do you judge a GEO proposal?

Ask where you stand now, what the supplier will change and how they will judge the result. The proposal should explain how that change could help someone choose you.

Ross Simmonds puts the priority plainly: do work that benefits people, whatever you call it. His point is about where to put the effort, not evidence of a measured uplift.