It is entirely possible for a business to rank well on Google and still barely appear when someone asks ChatGPT, Google AI Mode or another AI search tool for recommendations. That does not necessarily mean your SEO is failing, nor does it mean you need to throw out everything you know about search and start again with a completely separate “AI strategy”.
What it usually means is that there is a gap between being able to rank a page for a search query and being easy enough for an AI system to retrieve, understand and confidently include in a broader answer.
That distinction is becoming more important as the way people search changes. Someone searching Google for “math tuition Hougang” is giving the search engine a fairly narrow task. Someone asking ChatGPT, “Which tuition centres in northeast Singapore would you recommend for a Secondary 3 student who is struggling with algebra and prefers smaller classes?” is asking for a much more considered answer. To respond well, the system needs more than a page that happens to contain the right keyword. It may need to understand where the centre is located, which levels it teaches, how its classes are structured, what its teaching approach is, whether there is any evidence that it works well for struggling students, and whether the wider web supports what the centre says about itself.
This is why businesses sometimes look stronger in traditional search than they do in AI search. Their SEO may be good enough to surface individual pages, but the business as a whole is still difficult to understand.
Google’s own guidance now makes this distinction clearer. Its generative search features still rely on the same core search and quality systems, but they can retrieve several related pieces of information and use techniques such as query fan-out to explore different parts of a question before constructing an answer. Google therefore continues to describe good SEO as the foundation of visibility in its AI features rather than something that has suddenly been replaced by AEO or GEO.
The practical implication is fairly simple. If your business appears in Google but not in AI search, the answer is usually not to start stuffing more question-and-answer blocks into your site. It is to make the business easier to find, easier to understand and easier to verify.
First, make sure AI search can actually reach the site
The technical side of this is not especially glamorous, but it is the first thing worth checking because nothing else matters if the important pages cannot be accessed properly.
For ChatGPT Search, OpenAI says that any public website can potentially appear, but recommends allowing OAI-SearchBot so that content can be discovered, surfaced and properly cited. A site may look completely normal to a human visitor and still block automated crawlers through robots.txt, Cloudflare, a firewall, bot protection or another security layer.
Google’s requirements are similarly familiar. A page still needs to be crawlable, indexable and eligible to appear in normal Search before it can appear in Google’s generative AI experiences. There is no separate shortcut around the ordinary technical work, and Google explicitly says that the way it finds and processes pages remains central to how its AI systems access content.
For most businesses, that means checking the same foundations you would check in an SEO audit anyway: whether the important pages are indexed, whether canonical tags point to the right URLs, whether security tools are blocking crawlers, whether internal links make the site’s important pages easy to reach, and whether the content you actually want search systems to understand is present in the page rather than hidden behind an awkward technical implementation.
This is also why I would be cautious about starting an AI visibility project with things like llms.txt, special Markdown versions of every page or elaborate schema experiments. Some of those can be useful as supporting infrastructure, but they should come after the basics. Google now explicitly says that llms.txt does not improve or hurt visibility in Google Search, including its generative AI features, and that there is no requirement to create special AI-specific files in order to appear.
If a site cannot be crawled cleanly, adding more machine-readable files is not going to rescue it.
The bigger problem is often that your website does not explain the business clearly enough
Once the technical side is clean, this is where I would spend most of my time.
Many company websites make perfect sense to the people who built them because they already know what the company does. That does not mean the same thing is obvious to someone arriving for the first time.
A homepage might talk about “transforming possibilities”, “unlocking growth” or “building better outcomes” without ever plainly saying what the business sells. The service pages may be grouped under a heading such as “Solutions”, while the About page focuses on values and company history without establishing who actually does the work or what experience sits behind it.
None of this is unusual, and it is not automatically bad branding. The problem comes when the factual layer of the website is so vague that a search system has to infer almost everything.
If you are a Singapore SEO and content marketing agency, say that clearly somewhere important. If you work specifically with tuition centres, local businesses, SaaS products or publishers, make those relationships easy to find. If the work is delivered by a Singapore-based team rather than being outsourced overseas, that is useful context. If the founder has relevant experience or certifications, those should not be buried in a LinkedIn profile while the website presents itself as a faceless brand.
The same principle applies to the individual services. A company that offers SEO audits, local SEO, Search Console analysis, content strategy and creator marketing should not expect search systems to piece all of that together from a few paragraphs about “integrated digital growth”. The services should be named, described and linked to clearly enough that someone can understand exactly what is being offered without making assumptions.
This sounds like a small copywriting issue, but it changes how much usable information the site gives both people and machines. The clearer the business is about what it does, who it is for and how the work is delivered, the easier it becomes to associate that business with relevant searches and recommendations.
Ranking for one keyword is not the same as having depth
Another reason a business can look strong in Google but weak in AI search is that its visibility may depend on a surprisingly small number of pages.
A facial spa might rank very well for “extraction facial Singapore”, for example, while the rest of the website consists of little more than a treatment menu, a short About page and a contact form. That can be enough to perform for one valuable commercial term, but it leaves very little material for a search system to work with when someone asks a broader question about blackheads, pore congestion, extraction aftercare, sensitive skin or how often they should have a facial.
The obvious temptation is to interpret this as a reason to publish hundreds of articles in the name of topical authority. I would not.
The better question is whether the business is sharing enough of the knowledge it already has.
A facial spa that performs extractions every day should have more useful insight into pore congestion than a generic lifestyle website. A tuition centre that works with struggling students should be able to explain the mistakes parents often make when choosing a programme, what different class formats are suited to and how they assess whether a student needs help. A SaaS company should know the problems its customers run into before buying the product, during onboarding and after they start using it.
Those are the areas worth developing because they reflect genuine experience. They also create a much richer picture of what the business knows and what kinds of questions it is qualified to answer.
Google’s current guidance for generative search makes a similar point by emphasising what it calls non-commodity content: material that contains genuine experience, a distinctive point of view or information that cannot simply be recreated by summarising what already exists elsewhere. It specifically warns against producing large numbers of pages simply to capture every possible search variation or fan-out query.
The aim is not to cover everything. It is to have something worth retrieving when the subject genuinely overlaps with what the business knows.
Generic content has become much less defensible
This is one of the areas where AI search probably does change the economics of content marketing, even if the basic SEO principles remain the same.
There has always been a lot of interchangeable content online, but generative AI has made it almost effortless to produce even more of it. An article called “10 SEO Tips for Small Businesses” can be generated in seconds, and unless it contains something unusually useful, there is little reason to expect a search engine or AI system to care about your version in particular.
That does not mean informational content is dead. It means there is less value in publishing articles that could just as easily belong to someone else.
A real Search Console analysis of your own site is different. A case study explaining what you changed and what happened afterwards is different. A review of an SEO tool you actually used for 90 days is different. So is an article built around mistakes you repeatedly see when working on real websites, particularly if you can show examples or explain where your opinion differs from the conventional advice.
This is also one of the reasons I would not respond to AI search by making existing content more robotic. There is a growing genre of AEO advice that recommends turning every section into a question, answering it in 40 words, adding an FAQ to every page and breaking longer ideas into tiny “AI-friendly” chunks. Google has now gone out of its way to say that this is unnecessary. Its systems can understand multiple topics, synonyms and nuance within a page, and there is no special need to rewrite content purely for generative AI.
A well-written article can still have clear headings and direct answers without sounding as though it was constructed for a crawler.
In fact, I would argue that the businesses most likely to benefit from AI search over time are the ones that have something specific enough to be worth citing in the first place.
Your website is only one source of information about your business
This becomes particularly important when someone is asking for recommendations rather than simply looking for a piece of information.
Every business describes itself positively. Tuition centres have experienced teachers, spas provide personalised care, agencies are results-driven and restaurants use quality ingredients. Those claims may all be true, but there is very little reason for an AI system to treat them as independent proof simply because they appear on the company’s own website.
The wider web provides additional context. Reviews, press mentions, professional profiles, credible directories, creator coverage, community discussions and references from relevant websites can all help establish that the business exists beyond its own marketing copy.
That should not be interpreted as a reason to manufacture mentions everywhere you can. Google specifically warns that chasing inauthentic mentions is not a useful generative-search tactic, just as artificial link building has never been a particularly durable SEO strategy.
The more useful way to think about it is whether the business has built a credible footprint outside its own domain.
A tuition centre with years of genuine parent reviews, a complete Google Business Profile, teacher profiles, local references and mentions from education publications gives search systems much more to work with than a tuition centre whose entire online existence consists of a polished homepage and a few service pages.
The same applies to an agency. If the founder is visible, the services are clearly documented, case studies contain real numbers, certifications can be verified and the business publishes work based on its own experience, there is simply more evidence available when an AI system is deciding whether it belongs in a recommendation.
Service pages deserve more attention than most businesses give them
A surprising amount of commercial SEO work is still undermined by weak service pages.
Some businesses invest heavily in publishing articles while their main revenue-generating pages remain vague, short or poorly differentiated. A prospective client may land on a service page and still struggle to understand what is included, who the service is for, what the process looks like or whether the company has any particular expertise in that area.
That is already a conversion problem, but it also affects how clearly the business can be understood in search.
If someone asks an AI assistant for a Singapore agency that can analyse Search Console data and identify which existing pages should be refreshed first, a generic “SEO services” page gives the system relatively little to work with. A dedicated Search Console audit page that explains what is reviewed, what problems it is designed to uncover, what the client receives and how the service differs from a generic technical audit is considerably more useful.
This is why I would usually strengthen commercial pages before adding another large batch of informational content. A good service page can rank, convert and provide search systems with a clearer understanding of what the business actually sells.
For local businesses, there is another layer as well. Google explicitly recommends keeping local business information accurate through products such as Google Business Profile, since local information can also surface within its generative search experiences.
AI visibility is not just a blog-content problem. In many cases, the strongest improvements are going to come from making the commercial side of the website clearer.
Make the people and the proof easier to find
Another weakness I see surprisingly often is that businesses have real experience but do a poor job of showing it.
The founder may have worked in the industry for ten years, but the website gives them a two-sentence biography. The team may hold useful certifications, but those credentials only appear on LinkedIn. The company may have produced genuinely strong results for clients, yet the case-study section speaks in vague terms about “improving visibility” rather than showing what changed.
That makes the business harder to evaluate.
It is worth making the people behind the work more visible, particularly for professional services where expertise is part of what the client is buying. Explain who founded the company and why. Show relevant experience. Display certifications where they are meaningful. If the people clients speak to are also the people doing the work, say so. Where you have permission to name clients or show results, use concrete examples rather than general claims.
This is not about adding an enormous E-E-A-T section to every page. It is simply about making the business easier to verify.
Structured data can support that work by clarifying organisations, people, services and relationships between pages, but schema should remain a description of information that already exists rather than a substitute for it. Google also makes this distinction explicit: structured data continues to be useful as part of normal SEO, but there is no special AI schema that gives a site preferential treatment in generative search.
Use the data you already have before chasing new tactics
For an established website, I would still spend a lot of time in Search Console before deciding what needs to change.
That may sound remarkably ordinary for an article about AI search, but Search Console tells you what Google already understands about the site. A service page receiving thousands of impressions but very few clicks may have a positioning or title problem. An article ranking between positions five and twelve for dozens of relevant searches may deserve more attention than a completely new topic. A page appearing for questions it barely answers may be showing you where the existing content is incomplete.
Google now also provides a Generative AI performance report in Search Console for visibility across its generative search features, which makes it possible to examine those signals more directly rather than relying entirely on third-party visibility scores.
This is useful because there is already a growing market of tools that claim to tell businesses how “AI ready” they are. Some can be genuinely helpful as diagnostic tools, but no third-party platform has access to the internal ranking systems used by Google, ChatGPT or other AI search products. A score can point you towards something worth investigating; it should not be mistaken for proof that a particular change will get your business cited.
The same principle applies to traditional SEO audits. Knowing that 87 things could be improved is not the same as knowing which five are worth fixing first.
So why are you still missing from AI search?
If a business is already visible in Google but rarely appears when people ask AI tools for recommendations, I would normally look at the problem in roughly this order.
First, I would make sure the site is technically accessible and that the important pages can actually be crawled and indexed. Then I would look at the homepage, About page and service pages to see whether the business is explained clearly enough for someone who has never heard of it before. I would want to know whether its expertise is visible, whether its services are specific, whether there is evidence behind its claims and whether the wider web provides any independent context about the company.
Only after that would I start thinking about content gaps.
At that stage, the question is not “What else can we publish?” so much as “What does this business genuinely know that is currently missing from the site?” Sometimes that leads to new articles, but it can just as easily lead to stronger service pages, better case studies, a more substantial About page or an update to an older piece of content that is already earning impressions.
That approach is less exciting than promising to “optimise your brand for ChatGPT in 30 days”, but it is also much closer to what the search platforms themselves are telling website owners to do.
Good SEO has not suddenly stopped working because AI search exists. What has changed is that the weaknesses of generic, thin or poorly explained websites are becoming harder to hide.
What I would not prioritise
I would be cautious about any AI-search strategy built primarily around publishing hundreds of question pages, rewriting every article into short answer blocks, manufacturing brand mentions or installing increasingly elaborate layers of AI-specific markup.
Google’s current guidance explicitly pushes back against most of those tactics. It says there is no requirement to chunk content for AI, no special structured data needed for generative search, no need to create llms.txt for Google, and no reason to rewrite your content into a special AI-friendly style. It also continues to warn against scaled content created primarily to manipulate search visibility.
That does not mean AEO is meaningless. There is real value in understanding how search behaviour is changing, how conversational queries differ from traditional keywords and how businesses can make their expertise easier to retrieve.
The mistake is turning those observations into a new collection of hacks.
For most businesses, the better work is still more fundamental: explain what you do properly, make the commercial pages strong, publish things based on genuine knowledge, build evidence outside your own website and maintain a technically sound site that search systems can access without difficulty.
Those are not particularly fashionable recommendations, but they are much harder for the next algorithm change to make irrelevant.
Frequently Asked Questions
Why does my business appear in Google but not ChatGPT?
A business can rank in Google without appearing in ChatGPT because the two systems use different retrieval and ranking processes, and the queries people ask can also be very different. A page may be a strong result for a specific keyword while the website still lacks enough context, authority or supporting information to be included in a broader recommendation. OpenAI also recommends allowing OAI-SearchBot so that public website content can be discovered, surfaced and clearly cited in ChatGPT Search.
Can I submit my website to ChatGPT?
There is no equivalent of submitting each URL to ChatGPT in the way you might request indexing through Google Search Console. OpenAI says public websites can appear in ChatGPT Search and recommends making sure OAI-SearchBot is not blocked from accessing the content you want surfaced.
Does ranking first on Google mean I should appear in AI search?
No. Strong Google rankings are useful, particularly because Google’s own generative search features draw on its existing Search index and ranking systems, but a high position for one query does not guarantee that a business will be selected for a different conversational question. AI answers can draw on several sources and several related searches before constructing a response.
Is AEO different from SEO?
AEO is a useful way of thinking about visibility in answer-driven search experiences, but the overlap with SEO is substantial. Google explicitly says that, from its perspective, optimising for generative AI search is still SEO because its AI features rely on the same core Search ranking and quality systems.
Do I need an llms.txt file to appear in AI search?
You do not need one for Google. Google clarified in June 2026 that llms.txt files do not positively or negatively affect visibility or rankings in Google Search, although businesses can still maintain them for other systems that choose to use them.
Does schema improve AI search visibility?
Structured data remains useful for helping search engines understand entities and for eligibility in supported search features, but there is no special schema type required for Google’s generative AI search. Schema should support clear, accurate page content rather than being treated as a shortcut around it.
How long does it take for a business to appear in ChatGPT or AI search?
There is no fixed timeline. Visibility can depend on crawler access, indexing, query relevance, the strength of the website, information available elsewhere on the web and the retrieval systems used by each platform. Any fixed promise that a business will appear in ChatGPT within a particular number of days should therefore be treated with some caution.
How we approach AI search at The Tofu Media Group
We do think AI search deserves attention. People are changing the way they research products, businesses and services, and being easy to discover through Google alone is no longer the full picture.
What we do not think businesses need is another layer of marketing activity simply because the terminology has changed.
When a site already has search visibility but appears to be missing from AI search, we normally start by looking at what the business is already giving search systems to work with. Sometimes the problem is technical, but just as often the company has useful expertise that is poorly explained, service pages that are too generic, evidence that is buried or a substantial content library that nobody has properly reviewed in years.
That is why our approach starts with diagnosis rather than a predetermined list of AEO tasks. We look at what is happening, identify what deserves attention, test what can realistically make a difference and only then decide what is worth scaling.
It is the same thinking behind our Small Steps Growth Framework: Diagnose, Prioritise, Prove, Scale.
If your business is already appearing in Google but you are not sure why it is being overlooked in ChatGPT or other AI search experiences, our free growth review is designed to give you a clearer starting point. We will look at the site, tell you what we would investigate first and highlight the areas we would not spend money on yet.
Small steps before big spend.