Contact me for a free no-obligation Google search visibility report for your website via this website*. *Terms and conditions apply.
Published (1.9.25 - updated 7.26)
Generative Engine Optimisation (GEO) is a new approach to optimising online content that focuses on how AI-powered search engines, (also known as generative engines) understand and use information. It is an emerging field, and I can assist with outlining integrated digital marketing strategies to boost visibility . My work is focused on Google Gemini, whose technology is frequently integrated as AI Overviews (AIOs) in 'normal' search results on GOOG as well as Chatgpt. See https://www.seolondonsurrey.co.uk/blog/quality-webspam-issues-in-some-google-ai-overviews . I also focus my work on Bing Copilot (based on Chat GPT) which is integrated into some Bing search results and appears from some of my b2c clients to be popular as it is built into Windows operating systems.
My GEO services include:
1. Technical Analysis: LLM “Chatbots” (a.k.a answer engines) like ChatGPT or Perplexity do not render web pages like Google's Googlebot or Microsoft's Bingbot yet (at time of writing) because it's costly at scale. To be able to turn up in the responses generated from their web crawls / browsing tools, key information needs to be available in the initial HTML when the bot visits the site.
2. Analysis of query fan-outs (the process by which an AI search system takes a prompt) and expands it into related sub-queries) at scale using Microsoft Clarity (which is built on the same LLM model as Chatgpt) prompt data from Bing Copilot data as well as prompt-type searches (identifiable by their length) surfacing in Google Search Console data, clustering them into core topics, and identifying the key trends among them is one of the best places to start with AEO/GEO, in my opinion .
In combination with the above I use the Topic Insights tool in Clarity to better understand how your website contributes to AI-generated answers, how your website competes, and where opportunities exist. This will be done using "comparative entity prompt" research for 15 highly relevant prompts from my research to reverse engineer Chatgpt to include a list of Chatgpt's current high-confidence information about a comparative entity topic e.g. "best online tax return australia."
3. Ensuring Google Knowledge Graph (and sources of data) used by Google (Knowledge Graph – a part of Google Gemini) are optimised as fully as possible.
4. SEO / GEO / Content Strategy and ongoing manual link building to enhance the Harmonic Centrality (HC) of your website. Mozilla Foundation's 2024 report titled “Training Data for the Price of a Sandwich” confirmed that 64% of 47 LLMs analysed used at least one filtered version of Common Crawl (CC) data. For (Chat) GPT-3, 80%+ of training tokens came from filtered CC data . "CC uses Harmonic Centrality (HC) to determine crawl priority. Sites with higher HC scores are crawled more frequently. More frequent crawling means more appearances in monthly archives. More appearances in archives means greater representation in AI training data." See https://commoncrawl.org/blog/how-seos-are-using-common-crawls-web-graph-data-for-ai-ranking-signals.
Recent research from the SEO Agency named Salt shows that a third of responses to travel prompts in ChatGPT are based on the training data of that model only.
Unlike traditional SEO, which primarily targets keyword rankings in search engine results, GEO aims to make your business / product more strongly associated in conceptual / semantic space in the LLM models for relevant prompts e.g. "best smartphones". This is the core task (according to my research) which involves an integrated digital marketing strategy and is possibly something which is a long-term strategy e.g. according to my research, for the prompt "best smartphones" Apple iPhone & Samsung Galaxy may be considered "Global Prototypical Smartphone Exemplars" of smartphones in CGPT's semantic space. This is because according to my research these products are consistently recognised across most cultural and linguistic contexts, and show up as central exemplars in Chatgpt's knowledge base:
Apple's iPhone → universally salient, culturally iconic = often “the” smartphone prototype.
Samsung's Galaxy S Series → widely available, dominant Android flagship with global recognition.
Thus trying to replace the Iphone / Samsung Galaxy's strong associations as "Global Prototypical Smartphone Exemplars" might be considered a very difficult / long-term task.
GEO can also involve making content more accessible and useful to AI models that generate direct answers and summaries for user queries (like a Chatbot).
My 20 years experience of SEO informs my research of LLMs. https://lnkd.in/eMTMiqiB argues that bias is a structural and inevitable feature of current large language models (LLMs). Because LLMs are trained to approximate the distribution of human-generated text, they can & will reproduce the patterns (including harmful biases) embedded in those texts. There is good evidence for the broad alignment of LLM bias with human bias. But LLMs don’t simply reflect biases present in human language use. Thus it may be possible to engineer bias in the medium / long term..
Note r.e. my general approach: analysing Bing Copilot "query fan-outs" at scale using Bing Webmaster Tool data & clustering them into topics, and identifying the trends is a cornerstone of my approach. There are currently certain 'spammy' tactics being discussed / recommended to manipulate LLM responses (e.g. self-serving "listicles" e.g. listing oneself as 1st in a list article on one's own website on the topic of "top 10 online tax return companies") which I will not recommend . This is in my view short term / unsustainable tactics as these articles are "low quality content" which may result in the website reported by competitors & / or being penalised by Google and / or Bing (and as a consequence affect SEO visibility as well).
It may be that for certain prompts negative GEO (Generative Engine Optimisation) techniques are a main technique used to reduce a competitors visibility in LLMs.
Note that in my research I have seen that Google AI Overviews are localised for some prompts e.g. the results for "aluminium window manufacturers" are different in Glasgow, UK to those in Hampshire, UK . This is also the case for Chatgpt even for some prompts without any implicit local intent e.g. “best campervans” - this may be dependent on your (personalisation) settings in the Large Language Model (LLM) / search engine . For "normal" Google ranking tracking one may need to simulate searches from multiple specific locations (e.g. both London, UK & Bristol UK etc) and the same may be true for full LLM visibility monitoring.
Note that randomness in responses is a part of Large Language Models (LLMs), primarily controlled by sampling parameters like temperature and top-p, which allow the models to generate varied and creative, or consistent and predictable outputs . This can provide a challenge for us attempting to monitor visibility for certain prompts.
Key aspects of GEO:
Technical: Chatbots (a.k.a answer engines) like CGPT or Perplexity do not render pages like Googlebot / Bingbot (at time of writing) because it's relatively costly. To be able to turn up in the responses generated from their web crawls / browsing tools, information needs to be available in the initial HTML when the bot visits the site – test this with https://punits.dev/check-server-side-rendered/ .
Content Contextualisation: Ensuring your content is clear, well-structured, and provides sufficient context for AI to understand the topic and generate accurate responses.
Information Synthesis: Focusing on how AI can integrate and synthesize information from your content with other sources to provide comprehensive answers.
User Intent Understanding: Going beyond keyword matching to anticipate and address the deeper intent behind user queries.
Fluency Optimisation: Making sure your website's text flows smoothly, is free from errors, and is easy for both humans and AI to read.
In essence, GEO is about creating content that is:
Comprehensive: Covering all relevant aspects of a topic.
Contextual: Providing clear explanations and background information.
Accurate: Ensuring factual correctness and reliability.
Structured: Using headings, subheadings, and lists to organize information logically.
Accessible: Easy to read and understand for both humans and AI.
Authoritative: Google is likely looking at which other (authoritative) websites link to a website to deem how authoritative a site is.
Why is GEO important?
With the rise of AI-powered search features that provide direct answers and summaries, GEO is becoming increasingly important for online visibility especially due to the prominence of AI overviews on Google for many searches : a late 2024 study by SE Ranking stated that almost 10% keywords (8,718 out of 100K in their sample) have AI Overviews (AIOs). By optimising your content for generative engines, you can:
Increase the chances of your content being used as a source for AI-generated answers.
Enhance your brand's authority and credibility.
Reach a wider audience through AI-powered search experiences.
While traditional SEO remains relevant, GEO represents a significant shift in how we think about content optimisation. It's about creating valuable, informative content that not only ranks well in search results but also serves as a reliable source of information for AI-driven search engines.