All posts October 17, 2025

From SEO to GEO - Understand how AI Search works: Be found online in 2025

By Will Price · 8 minute read

The death of the website?

Its not breaking news anymore: 2025 has seen a massive drop-off in website visits with a reported average decrease of 15% and up to 60% in some cases. This is almost entirely attributed to one thing, the emergence of AI Search.

Organisations whose business models are dependent on driving traffic to their websites via search engines are under threat from AI tools like ChatGPT which are designed to fulfil their users needs without them needing to click away from their interface (so called Zero-click search).

Does this spell the death of the website?

No it does not - in the short term there is still plenty of traffic coming from Google and you will also drive visits to your site through campaigns and marketing tactics in other channels.

Its also going to be a while before we delegate agents to execute all of our digital interactions, so we can hook ourselves up to The Matrix, rediscover nature or [insert your favourite theory about the future of humanity here].

Finally, it’s also important to emphasise that giving up on your website is (for the moment) equivalent to throwing in the towel. There may be fewer people visiting your site but the AI bots are out there in force, trying to understand your brand and content to give more meaningful responses to the prompts that drive them. If your content isn’t available to these bots - you definitely will not show up in AI Search.

How does AI Search work?

To improve your chances of getting your brand and content to show up or be cited in AI Search it's crucial to understand the mechanics of how AI Search works.

AI Search tools use four main components:

Large language model (LLM) - An AI system which has been trained on a body of content up to a certain cut off date (for example, ChatGPT-5 cut-off is October 1st 2024). An LLM interprets natural language inputs (prompts) to provide responses based only on the content it has been trained on. Original versions of ChatGPT and other AI tools worked with this component only.
Control layer - This sits in front of the LLM, to interpret the prompt provided and gather any additional context to pass through to the LLM.
Web index retrieval - Makes it possible to search the web for up-to-date content by accessing a continuously updated web index - a snapshot of content, optimised for rapid retrieval. ChatGPT added this capability towards the end of 2024.
Real-time web retrieval - Makes it possible to request content real time from the source (much like a human visiting a website). Not all AI Search tools provide this capability, or only provide it for limited use cases.

When a user enters a prompt, the control layer will first determine intent and establish if up-to-date knowledge is required to supply a response. For example consider the following initial prompts:

“What should I look for when buying new running shoes?” - the control layer will determine that its unlikely that knowledge after the cut-off date is required to give a good quality response and pass the query direct to the LLM to craft a response using it’s ‘baked-in’ knowledge.

“What are the latest innovations in running shoe technology?” - the control layer will detect that the user’s intent is to find up to date information and decide that the web index should be queried. The top results returned will be passed to the LLM as extra context, to formulate a response in natural language.

“What shops have ASICS Metaspeed Sky Tokyo shoes in stock?” - the control layer will detect that real time information is needed (stock) and this may result in a combination of web search plus real-time retrieval being used. Note again, not all AI Search tools provide this capability or will use it for use cases like this.

Important to understand is that often the web index used by AI Search is outside of the AI tool’s own eco-system. For example, at the time of writing, Claude uses Brave Search, Gemini uses Google and ChatGPT uses variously (depending on version, subscription and who you believe) Google, Bing or its own proprietary index.

Also important is that the user’s prompt will be altered when querying the web index for up-to-date information. The control layer will process additional context information if deemed useful, such as the user’s location and conversation history, and execute several different but related searches in parallel (this technique is known as Query fan-out).

How does AI Search understand my brand’s content?

It should now be clearer at what points an AI tool interacts with your brand’s content, and this is key in understanding how to get your content shown or cited.

During training: AI bots will crawl your website and any other sources where your brand is mentioned (for example YouTube, user forums like Reddit and Wikipedia) and provide this information as training data which gets ‘baked-in’ to the tool’s LLM.
When updating web indices: Bots continuously crawl your website and other sources to update their web index which is then used to provide efficient access to up-to-date content. Remember an AI tool may use a web index from another provider, so when using ChatGPT, it might actually be using Google to do the searching.
Direct retrieval from the source: An AI tool may request content from your site real-time and parse and summarise it to assist in forming a response. Note that this is an uncommon form of content retrieval as it is not as efficient as using an index.

If your content doesn’t rank very highly in web indices then it follows that it will not be passed in as context to the LLM producing a response and effectively you do not exist.

With traditional search, the task of improving your ranking (Search Engine Optimisation or SEO) was made easier by:

Simplicity: You can search Google and see exactly what results are shown and in what rank
Google’s dominance: Most organisations could focus purely on ranking on Google and not pay significant attention to other niche search tools
Transparency: Google themselves are reasonably open about how to improve your search ranking, providing tips and tools to do so

For Generative or Answer Engine Optimisation (GEO or AEO) things are more tricky. While Open AI has a large market share with ChatGPT, there are a multitude of other players jostling for dominance like Google (Gemini), Anthropic (Claude) and Perplexity.

Add to this the lack of transparency and constant flux in how underlying web searches are executed and we have a situation where a lot of ‘experts’ in this field are actually operating on little more than guesswork.

“One of the main functions of jargon is to exaggerate expertise.”

Mokokoma Mokhonoana

How can I get my content shown or cited in AI Search?

Much advice around getting your content shown or cited will depend on your particular industry and the AI tool you are targeting. Different AI tools lean more or less heavily on different sources and techniques and I would recommend you do some industry-specific research in this area.

There are, however generic steps you should take to increase the chances of making your content and brand more prominent in AI search.

Monitor AI usage, and understand your audience

If your site has significant traffic, you are almost certainly getting referral traffic from AI Search already and this typically makes up 1-5% of total visits. If you have not done so, make this clear in your web analytics and try to understand what content is being cited where and why. Creating baseline monitoring will ensure you can measure the success of any changes you are making.

Do research on how your target audience use AI Search tools to explore content and brands in your field. Simulate this behaviour and analyse when and from what sources content related to your brand is surfaced. Again, baseline monitoring on different interactions in key AI search tools will help measure the impact of any changes you make.

Some AI Search tools enable you to turn web search capability on and off, and this allows you to compare the impact of ‘web-retrieved’ content in improving the quality of response generated by the tool on top of its ‘baked-in’ knowledge.

Make it easy for AI to access your content

You need to make sure you are not blocking any important bots from crawling and indexing your content. There are a lot of new bots out there, and some Content Delivery Networks (CDNs) and hosting providers block them by default due to the potential load they put on their infrastructure and/or the potential for copyright infringement.

Once you are sure the bots can access your site, the need to be able to make sense of it. Currently many AI crawlers do not execute Javascript, so if content is loaded dynamically it may not be ‘seen’. Web accessibility and SEO guidelines around good content structure remain valid in helping crawlers understand your content better.

Finally, make sure your website performs. Speed of response is a key component to the AI Search tool providing an engaging user experience. If your site is slow to load, AI tools and crawlers can penalise it, particularly in the context of up-to-date knowledge retrieval.

Create great content

AI Search likes authoritative, authentic and up-to-date content which is consistent with your brand. There isn’t really a magic wand for this, but once you have done your best to understood your audience, which AI search tools they use and how their conversations translate into ‘baked-in' content retrieval in LLMs or ‘up-to-date’ web searches you can better align the content you are creating to match the user’s intent.

The best content formats will change over time, but at the moment positive results result from adding supporting quotes, citations, statistics and structured lists.

Don’t limit your content strategy to your own website, you need to get your brand talked about on the channels that AI Search uses to build up a picture of your brand. For example ChatGPT currently leans heavily into Reddit and Gemini makes extensive use of YouTube as supporting content sources.

This will change over time, so again its important to have continuous monitoring to be aware of the intersection between what your audience is asking, and what sources AI Search tools rely most on to provide responses.

Conclusion: Is GEO the same as SEO?

If you have worked in that field you will notice that a lot of the tactics for SEO are still applicable for GEO, so is GEO replacing SEO, do we need to think of them separately or are they actually the same thing?

Traditional search will continue to serve a purpose and will not disappear overnight so SEO is definitely not dead - not least because AI search is, for now largely dependent on querying existing web search indices for up-to-date content and context.

Does this mean that GEO is the same thing as SEO? Right now, I see it as different marketing for the same basic principles. However, the scope of AI Tools such as ChatGPT is widening from purely providing answers to taking actions (facilitating payments for example) and ensuring your brand and content is tightly integrated into this wider eco-system will require many additional considerations than traditional SEO.

Whats undeniable is that AI Search has made the mechanics more complex to understand, and this article should go some way to bring clarity and help you ask critical questions back to experts who claim to have ‘cracked it’.

Regardless of what you call it, if content forms a key part of your business model, and your target audience would expect to interact with your brand through AI search you need to take SEO/GEO seriously and invest in updating your monitoring, tooling, technical readiness and content strategy.

Terminology

Heres a summary of the main AI Search terminology used in this article.

AI Search – Search driven through a conversational interface, and enhanced by artificial intelligence to understand intent, rank results, and provide more relevant or personalised responses.
Zero-click search – Search responses that answer queries directly on the results page, without needing a click-through to the source.
Search Engine Optimisation (SEO) – Optimising content and websites to rank higher in search engine results within tools like Google and Bing.
Answer Engine Optimisation (AEO): A predecessor to GEO - Optimising content to appear in direct answers provided by search engines (for example FAQ content shown directly as a snippet on Google search results).
Generative Engine Optimisation (GEO) – Optimising content to appear more prominently within AI Search responses within tools like ChatGPT, Claude, Perplexity and Gemini.
AI bot / AI crawler – Automated AI system that scans websites to index or analyse content for search, training data or insights.
Content Delivery Network (CDN) - A network of servers that act as local gatekeepers to your website. They improve page load speed by serving a cached version of your content from the closest location and often provide additional security features like blocking bots.
Query fan-out – The process of expanding a single search query into multiple related queries to improve coverage or results.
Large language model (LLM)– AI trained on vast text data to understand and generate human-like language.