Two years ago, when a client asked us to «come first on Google», we all understood exactly what they meant: that first blue link in the list of results. Today, the conversation has changed completely. Every week, someone comes to Vandelay with a different concern: «Why does ChatGPT recommend my competitors when people ask about my sector, yet doesn’t even mention me?» Answering that question is precisely the domain of the optimisation for generative engines (GEO, from the English Generative Engine Optimisation): the set of techniques designed to ensure that artificial intelligence systems such as ChatGPT, Gemini, Perplexity, Copilot or Google’s AI Overviews cite, mention and recommend your brand within their responses.
It’s not just a passing fad, nor is it a substitute for traditional SEO. It’s a new layer being added on top, and it’s worth understanding how it works before your competitors get ahead of you in the place where more and more people are starting to search.
What is generative engine optimisation and how does it differ from SEO?
Traditional SEO aims for positions in a list of ten results. Optimisation for generative search engines aims for something different: to appear inside of the response that the AI drafts for the user. You’ll see two circular acronyms that are almost synonymous. AEO (Answer Engine Optimisation) usually refers to answer engines, such as Google’s AI Overviews; GEO focuses on generative engines, such as ChatGPT or Perplexity. In practice, both describe the same objective: to gain visibility when a machine synthesises the answer rather than returning links.
The technical difference matters. A large proportion of these systems operate using Retrieval-Augmented Generation (RAG): they retrieve content in real time, interpret it and construct a response by citing certain sources. Your job is no longer simply to rank a URL, but to ensure that your content is easy for a language model to retrieve, easy to understand and easy to cite. It’s a shift in mindset that we’ve already discussed when talking about the loss of SEO visibility due to generative AI.
Why optimisation for generative engines is already affecting your business
The data confirms that this is not mere speculation. According to Conductor’s analysis of millions of queries, Google’s AI Overviews appeared in around 25 % of searches in early 2026, compared with just over 13 % in March 2025; other metrics, such as those from Advanced Web Ranking, put that figure at over 60 % in the US market. The range is wide because each study measures different sets of keywords, but the trend is unmistakable: AI-generated answers are gaining ground at the expense of traditional links.
Added to this is user adoption. ChatGPT reached around 900 million weekly active users in February 2026, more than double the figure from a year earlier. And Gartner predicts that by 2028, half of all online searches will involve an AI assistant. If half of all queries go through a conversational layer, not being present there is tantamount to being invisible to a huge proportion of your market.
What the data tells us about how to get a date in the AI world
This is where evidence helps to separate the noise from what really makes a difference. The seminal academic paper, «GEO: Generative Engine Optimisation» (Aggarwal et al., presented at the 2024 KDD conference and led by researchers from Princeton), demonstrated that a brand’s visibility within generative responses can improve by up to 40 % by applying three specific tactics: citing authoritative sources, incorporating statistics and adding direct quotes. These aren’t tricks: they are signals that the model interprets as reliability.
The second insight comes from Semrush, through what it calls the «Mention-Source Divide»: fewer than one in five brands manage both to be mentioned frequently and to be cited as a source in AI-generated responses. In other words, being mentioned does not guarantee that you will be linked to, and vice versa. Working on both aspects simultaneously is what makes the difference.
There is a third factor that should not be overlooked: cross-referencing between sources. These systems tend to place greater trust in a brand when it is consistently mentioned across various independent domains — industry media, directories, reviews, press releases. The consistency of your information across the web reinforces what the model recognises as a solid and recognisable entity.
A step-by-step guide to optimising for generative engines
With all of the above in mind, this is what we’re implementing in the accounts we manage. There’s no need to reinvent your website; you do, however, need to organise the content with a machine’s reading habits in mind:
- Answer before you go off on a tangent. Begin each section with a clear and direct definition or answer to the question. Research into retrieval in language models shows that passages with a definitional structure at the start are retrieved more effectively.
- Organise the content into collapsible sections. Lists, comparisons, tables and «top X for Y» formats are easy to quote because the model can extract the data without misinterpreting it. Well-organised information beats endless prose.
- Provide facts, figures and quotations. This is the tactic backed by direct academic research. In the eyes of AI, a single paragraph containing a statistic and its source is worth far more than ten paragraphs of filler.
- Mark up your content with structured data. The schema FAQ- and How-To-style content helps search engines understand which question each section answers. It is one of the most frequently cited predictive signals for appearing in search results.
- Look after your online presence outside your website. Work on ensuring consistent mentions across media outlets, directories and reviews. Domain authority is built across many sites, not just on your own domain.
- It makes it easier for the models to read. A file llms.txt When done well, it helps guide ChatGPT, Claude or Perplexity towards content that truly represents your brand.
- Keep your content fresh. Freshness is a recurring factor: reviewing and updating key pages sends a signal that the information is still valid.
How to gauge whether your GEO strategy is working
Here’s the tricky bit: measuring GEO isn’t the same as opening Search Console and checking rankings. The metric that’s becoming the industry standard is the Share of Model (sometimes known as share of voice (in AI): how often your brand appears in the generated responses compared to your competitors for a set of relevant questions. There are already tools that track your mentions on ChatGPT, Gemini or Perplexity, but it’s worth supplementing these with what you can actually verify: referral traffic from these platforms in your analytics and the trend in your impressions in Search Console for queries with AI Overview. As for why traditional measurement falls short, we explore this in our article on the AI-powered marketing measurement.
Let’s be honest: this is a fledgling field and the metrics are still being standardised. Anyone who promises you a foolproof formula for «coming out on top in ChatGPT» is selling you a pipe dream. What we can say for certain, based on the available evidence, is that well-structured, data-backed content that is consistent across the web is far more likely to be cited. The rest comes down to constant measurement and fine-tuning.
If you’ve got this far, you’ve probably already realised that there’s an answer to your client’s question — «Why doesn’t the AI mention me?» — and that the answer starts with reviewing how your content is currently written and structured. At Vandelay, we’re auditing real accounts using this approach and fine-tuning strategies before the gap closes. If you’d like to know how your brand appears when someone asks an AI about your sector, get in touch and we’ll look at it together: sometimes the initial assessment is the most revealing part.
Sources
- Aggarwal, P. et al. «GEO: Generative Engine Optimisation», KDD 2024 — arxiv.org/abs/2311.09735
- Gartner, predictions on search and AI assistants — gartner.com
- Conductor, «2026 AEO / GEO Benchmarks Report» — conductor.com
- Semrush, analysis of mentions and citations in AI («Mention-Source Divide») — semrush.com
- eMarketer, «FAQ on GEO and AEO» (2026) — emarketer.com
The figures cited are taken from public reports from 2025–2026 and are, in several cases, estimates or projections; methodologies vary between studies, so they should be regarded as orders of magnitude rather than absolute values.
English (UK) 
Español