For many years the agreement between a company and a search engine was simple: the business provided useful information, Google would index it, and when someone was looking for that information they would see a list of websites. The business then competed for a position close to the top and the reader would click through.
The situation is different now. A person who is looking for an accountant, a payroll system, or a specialist manufacturer can now obtain a fairly complete answer without having to visit a website. Google compiles a summary drawing on a number of sources. The AI assistant investigates the question, compares the different options, and then gives a response together with references to its sources.
It is convenient for the reader, but for the company it poses an unfamiliar question: if people are able to get the answer without going to your website, then what does it mean to be discoverable?
When we published our guide on how to prepare websites for AI search last October, a good deal of it is still relevant, even though we attempted to explain too many things at once and made a simple idea sound more complicated than was necessary. This article is our effort to outline the major change, explain why it is important to ordinary businesses, and indicate where the practical work should start.
What We Said Last Year
The 2025 guide was based on three recommendations: include structured data, especially FAQ markup, so that machines can understand what a page is about; use headings that match the questions that people type; answer each question in between forty and sixty words before going on to explain it; and also included a argument in favour of FAQ pages, a thirty-day plan, and some figures regarding zero-click searches and the results that other businesses had obtained.
The three recommendations still stand; however, the way they are presented has not aged well since it holds that AI search was dealing with a formatting problem, one that could be resolved through the use of markup and structure, and that doing so would reliably result in citations. Moreover, it relies on figures which we had not sourced and which we will return to at the end.
How Does an AI Assistant Decide Which Businesses it Understands and Recommends?
Search has not vanished. People still use Google, go to websites and check information before spending money. The difference now is that more searches include an extra stage between the question and the website, in which a system interprets the question, collects information and then provides the answer on the reader's behalf.
Suppose a business owner is selecting a payroll system for forty employees. In the traditional approach, he or she looks at software companies, review websites and articles that compare the products, and then goes through all of these. Alternatively, an AI assistant could outline the key differences, determine which factors are important at that scale, and provide a few options together with the sources. The owner would then go to one or two websites instead of eight or ten.
The business whose information was used to produce that first response is the one which comes into the reader's awareness before the reader has even arrived. A different business, even if it is just as capable, fails to enter the conversation since its information could not be located, comprehended or regarded as useful enough to include.
It's possible for a website to be well designed, technically sound, and appear prominently in standard search results without having any effect on the answers compiled elsewhere. On the other hand, a smaller business that has real expertise might have some valuable contributions to make, as long as the systems that people use more and more to obtain information are able to locate and interpret that expertise.
The issue isn't anymore about the position of your page in a list; it's about whether your information can help in arriving at an answer and whether anyone has a reason to go back to you.
AEO and GEO, Without the Jargon
Answer Engine Optimisation, often referred to as AEO, involves making content useful and easy for systems that respond directly to questions, such as those which provide summaries or give an answer rather than a list.
Generative Engine Optimisation, often referred to as GEO, involves increasing the chances that generative AI systems will find, use, or cite some content; the terms are still being clarified and overlap to a large extent.
In ordinary language they are solving the same problem: how to make your knowledge visible and useful in the case where a system carries out much of the work of finding and interpreting information on someone else's behalf.
There is no need to give up on search engine optimisation. The basic principles stay the same: accessible pages, a reasonable site structure, descriptive headings, and information that responds to the questions people really ask. Google has stated that there are no extra requirements for its AI features beyond the current guidelines and that no special markup or AI-specific files are necessary. There is no valid reason to regard AEO and GEO as secret fields with their own set of rules. The point at issue is the kind of outcome being aimed for. Traditional SEO investigates whether a page ranks. AEO and GEO likewise look at whether the information is appropriate to include in an answer and whether the source is worth citing.
SEO
Where you appear: a list of links
What the reader does: clicks one
What you need: a page that ranks
AEO
Where you appear: a summary or spoken reply
What the reader does: reads or hears the answer
What you need: content that answers one question clearly
GEO
Where you appear: an AI-written response with sources
What the reader does: reads it, and may follow a reference
What you need: content worth citing, with a credible origin
It cannot be guaranteed that information will be visible or cited. This is because AI systems differ from one another in the way they retrieve information and produce responses. However, the distinction remains a useful way of thinking about the work.
Why an Ordinary Business Should Care
It is not necessary for most businesses to become experts in artificial intelligence; what they need to do is to understand how their customers make decisions and at what points information enters that process.
A company is judged on its ability to meet a specification or to deliver on a large scale; in the case of a professional services firm, this is done because people wish to become familiar with the two methods before hiring an adviser. In both situations the business has information which would assist a potential customer in making a decision. The problem is that a great deal of this information is missing from the website, hidden within general descriptions, or presented in language which conveys very little. A website that talks about its commitment to quality and innovation, without explaining the materials it deals with or the problems it resolves, is not poorly designed; it simply fails to provide anyone who is trying to decide whether the business is relevant to their particular problem with any useful information.
General claims are easy to repeat and do not provide a reason for preferring one company to another. It is specific capability, original evidence, and clearly stated expertise that do. The opportunity is not limited to companies that have large marketing departments; it starts with knowing something worthwhile to explain and then going to the trouble of explaining it properly.
What Determines a Source as Being Worth Citing?
It is tempting to think that AI systems have a preference for short answers and neatly arranged sections. Although a well-structured page helps, it is no different from thousands of similar ones if it merely repeats standard advice without evidence, experience, or personal opinion. What matters is the content itself.
Assign a clear author to the information. People have always wanted to know who was responsible for a piece of writing and what kind of experience that person has. Where it is appropriate, name the authors and provide biographies that are meaningful, and make it easy to tell the difference between expert knowledge and mere assertion. Simply stating an author's name does not prove that they are credible, but hiding the author's identity makes it more difficult to judge their credibility.
Publish evidence that you possess. This should include the original measurements, the documented processes, and case studies that have been published with the client's approval. A manufacturer with years of experience is able to explain a technical problem that generic articles only briefly refer to. Since the explanation supports the observation, it has value in itself. A carefully documented example, together with a statement of its limitations, is more informative than a general claim supported by confidence.
You should be ready to form an opinion. A great many websites steer clear of showing a preference out of concern that they might exclude a customer. What then arises is a series of options together with the note that the correct choice will vary from person to person.
The answer should start with a list of options, noting that it will depend on your needs, and this is not the conclusion thereof.
A page that is more helpful will tell you under what circumstances one approach is preferable to another and what you should look at before making your decision. The recommendation has to be based on the evidence, and that evidence must be open to scrutiny.
Make it clear that the conclusions you reach are different from those obtained from other sources. Refer to research, use external statistics, and state when a conclusion is your own interpretation rather than something that is a fact.
That doesn't ensure a citation; in fact, what is produced is more lasting than content designed around a temporary technical trend, since it demonstrates what the business knows and provides others with good grounds for using it.
Which Technical Foundations of Search Still Matter?
Expertise won't be of any use if the systems responsible for locating it cannot get to the pages; technical work must still form the foundation rather than the strategy.
Make sure that the key pages can be crawled, indexed and understood regarding the robots.txt rules, the indexing directives, and the canonical URLs. Since different AI services make use of different crawlers, the appropriate checks will vary according to which services are important to your audience. Blocking one crawler does not mean that your business will be invisible everywhere, and granting access does not ensure that a page will be retrieved.
Structured data allows search engines to understand a business, a product or an article; it must describe what actually appears on the page and is not a way of telling an AI system to recommend you.
The importance of speed, mobile usability and good navigation remains since both people and systems need to use the site. You should find out what your tools are actually able to tell you. The Search Console provides information on Google Search performance; compare the details available for AI features with the current documentation for your property. Although referral data can give some indications it will not pick up every case in which a page had an influence on an answer. The difference lies in measuring what you can observe rather than pretending to measure what you cannot.
What Should You do First?
It isn't a redesign and won't be an article each day; instead, begin with the pages that are most important and check if they address the questions a potential customer is likely to have.
Treat your most important pages as if you were a stranger; choose three or five of them. Can you determine what the business does, who it serves and what sets it apart? Are the claims backed up by evidence, and is the person responsible for them clear?
Gather the questions that customers ask before making a purchase. For example, those regarding compatibility, cost, implementation, risk, and suitability. Such questions are more informative than the keywords you have usually targeted.
Look at the technical aspects, check crawl access, indexing, canonical URLs, and the structured data which tells the truth.
Set a baseline by looking at the existing reports and then testing a range of questions using the AI services that your customers use. Make a note of the questions and the sources referenced, and approach it as an investigation rather than as a rating.
Compose a page's worth of quoted material. Choose a topic on which you are more knowledgeable than anyone else and write out the page so that it would be valuable even if never referenced by any AI system. Outline the problem, present the evidence, and make a recommendation, ensuring that the recommendation is supported by the evidence.
The page has won its place simply because it has helped a customer to make a better decision, no matter how search developments turn out.
Where Euryka Stands
We believe that this kind of discussion is most beneficial when it shifts from focusing on the promise of a new optimisation technique to concentrating on the quality of information that a business provides. A company doesn't need to produce a continuous flow of articles; instead it should work out what it knows, decide which of that knowledge is useful to its customers, and then present it in a form which people can understand, assess and make use of.
The method begins by focusing on current discoveries, identifying what is lacking and highlighting the unique points the business has to make. Next, the technical basics are established, the content that should exist is developed and a practical approach is put in place for monitoring what takes place afterwards. People retain responsibility for the thinking, writing and judgement. While technology assists with research, verification and measurement, it does not take the place of having something to say and the evidence to back it up.
That standard extends to us as well; in our previous guide we had included statistics which were not properly sourced and results that we could not assign to any source. We are retaining the guide, with a note telling readers to consult this page, rather than allowing the original assertions to remain unexamined. It would be of little use to tell businesses to build up their credibility if we are not going to subject our own claims to the same level of scrutiny.
How Does AI Search Change the Role Your Website Plays for Buyers?
It won't be the case that all websites become irrelevant and customers will not cease clicking. What is changing is the role a business plays when it joins the conversation; your information might affect the customer's decision before they even get there, and you may never know to what extent it had influenced them.
That is the reason why the quality, credibility and usefulness of the material you publish are important. A citation included in an answer produced by an AI does have value, but it is not the criterion for success; the true criterion is whether the information helps the appropriate people to understand what you do, to place trust in your expertise and to think of you when they need something that you provide.
If your website already has that feature, then you have a solid base; but if it doesn't, AI search has only made the gap more difficult to ignore.
If you want to find out how your website is appearing in the answers produced by AI and to discover where the most useful improvements are, then send us three pages that are important to your business; that is the point at which we will start.
Sources and Further Reading
Google Search Central, AI features and your website, guidance covering AI Overviews and AI Mode.
Google Search Central, guidance on helpful, reliable, people-first content and structured data.
Google Search Console Help, current documentation on search performance reporting.
OpenAI Help Centre, documentation on ChatGPT search and source citations.
Euryka, How to Make Your Website AI-Ready: The 2025 Guide to AEO and GEO, October 2025.
