SEO is dead again.
It has been dying with remarkable consistency for more than twenty years.
This time the supposed murderer is AI search. Google has AI Overviews and AI Mode. Bing has Copilot and increasingly detailed AI visibility reporting. People are asking conversational questions instead of typing three-word searches. Answers can appear before anybody reaches the traditional blue links.
So naturally, a new industry has appeared to explain that everything you knew about SEO is obsolete and you now require GEO, AEO, LLMO, or whichever collection of letters has the best webinar attached to it.
There is a real change happening.
But the useful response is considerably less dramatic.
Google's advice is almost suspiciously boring
In 2026 Google published dedicated guidance about optimisation for generative AI features. Its central message is not that website owners need a secret new AI strategy.
It is that the fundamentals of SEO still apply.
Google's current documentation says there are no additional technical requirements for appearing as a supporting link in AI Overviews or AI Mode. A page needs to be indexed, eligible to appear in Search with a snippet and built according to the same fundamental SEO practices that already matter.
Google specifically points to crawlability, internal links, page experience, important information being available as text, useful images and video, accurate structured data and up-to-date business information.
It also says you do not need a special AI schema or a new machine-readable AI file to qualify.
The machines have become considerably more sophisticated. They still need to find, understand and trust the page.
That is a useful place to start because it clears away a surprising amount of magic powder.
AI search does not make crawlability less important
There is something wonderfully optimistic about discussing advanced AI optimisation while a website accidentally has noindex turned on.
Before an AI system can use your website as a useful source, the underlying search infrastructure generally needs to discover and understand it.
That means the unglamorous technical foundations still matter: sensible URLs, crawlable navigation, correct status codes, canonical URLs, XML sitemaps, internal links and content that is actually available to crawlers.
Microsoft has been making the same point from the Bing side. Its guidance for AI-powered search continues to emphasise sitemaps, accurate modification dates and IndexNow for notifying participating search engines when content is added, changed or removed.
AI did not repeal HTTP.
Internal architecture matters more when questions become more complex
Traditional SEO often gets described as matching one page to one keyword.
That was always an oversimplification, but AI search makes the limitation particularly obvious.
Google explains that AI Mode and AI Overviews can use a technique it calls query fan-out. A complex question can trigger multiple related searches across subtopics and data sources before the system assembles a response.
That creates an interesting opportunity for well-structured websites.
A business does not need one enormous page desperately trying to mention every possible variation of a subject. It can have a clear service page, relevant project examples, useful documentation, genuine articles and supporting product information, connected with sensible internal links.
Each page has a job. Together they establish context.
This is one reason we have been deliberately building Delicious Web Design as an interconnected site rather than a collection of isolated landing pages. Services connect to projects. Projects demonstrate the services. Software has documentation. Articles can explain the thinking and engineering behind the work.
That architecture is useful to people first. Conveniently, it also gives search systems a much clearer map of what the business actually knows and does.
Original experience has become harder to fake convincingly
AI can generate a perfectly competent article about almost any common subject in seconds.
That changes the value of generic information.
If a thousand websites can publish essentially the same explanation of “five benefits of responsive web design”, producing the thousand-and-first version is not a particularly exciting content strategy.
What is harder to manufacture is experience.
A real migration with real constraints. A plugin you actually built. A problem you actually diagnosed. Screenshots from the work. Decisions you made and why you made them. Measurements, failures, trade-offs and results.
Google's 2026 AI optimisation guidance explicitly highlights the value of non-commodity content. That phrase is useful because it describes exactly what AI abundance has changed.
Information that can be generated anywhere is abundant.
Evidence that you actually did something is not.
Structured data still matters, but it is not an AI cheat code
We like structured data.
We use it on Delicious because it gives machines explicit information about the business and helps remove ambiguity. A business name, type, location, contact information and other defined properties are much easier to interpret when they are represented consistently in machine-readable form as well as visibly on the page.
Google continues to say structured data helps it understand page content and can make pages eligible for supported search features.
What Google does not say is that adding a mysterious AIOptimisedBusiness schema property will summon an LLM carrying a bag of rankings.
In fact, its AI guidance specifically says there is no special schema.org markup required for AI Overviews or AI Mode.
Use structured data because it accurately describes what is on the page. Keep it consistent with visible content. Validate it. Do not treat schema as a place to hide claims you were too embarrassed to put in front of a human.
Clear writing has acquired a second audience
Writing for humans and writing so machines can understand you are increasingly the same job.
Clear headings establish subjects. Descriptive links explain relationships. Direct answers make information easier to extract. Consistent terminology reduces ambiguity. Useful page titles tell everybody what they are about to read.
None of this requires turning prose into robotic SEO copy.
Quite the opposite.
The old habit of awkwardly repeating a target phrase because a plugin's traffic light wanted another occurrence looks even stranger in a world where language models are specifically designed to understand language.
Write naturally, but be specific.
“We build websites” is understandable.
“We design and develop custom WordPress websites for Perth businesses” tells both a person and a machine considerably more.
Your business information should agree with itself
For local businesses, entity consistency is not glamorous, but it is important.
Your website should clearly identify who you are, what you do, where you operate and how somebody can contact you. Business profiles and relevant directory information should not contradict the website. Structured data should describe the same business the visitor can see.
Google's AI guidance specifically calls out keeping Business Profile information current. Bing's AI guidance similarly emphasises accurate local business information for location-based answers.
This is not about spraying your business name across 300 directories created during the reign of Internet Explorer 6.
It is about reducing ambiguity.
If the web contains multiple trustworthy, consistent signals describing the same real organisation, search systems have a much easier job understanding the entity behind the website.
Images are becoming searchable information
Visual SEO has also become more interesting.
Google has continued expanding multimodal search through tools such as Lens and image-based search, and in September 2026 added Search Console reporting specifically for web multimodal search performance.
That means images are not merely decoration around the “real” searchable content.
Original project photography, useful screenshots, product images and diagrams can all carry information. Descriptive alternative text, sensible filenames, surrounding context and technically accessible image resources help search systems understand what those visuals represent.
For a design business this is particularly relevant. Showing the work has always mattered to humans. Machines are becoming better at understanding it too.
What about llms.txt?
This is a perfect example of why SEO advice needs dates attached to it.
You may have seen recommendations to create an llms.txt file as an essential step for AI visibility.
Google clarified its position in June 2026: llms.txt is not needed for Google Search and does not positively or negatively affect visibility or rankings there. Google also says there is no special AI text file required to appear in its generative search features.
That does not mean the format can never be useful to another system. It means you should not install it because somebody told you Google now requires one.
The distinction between “an interesting emerging convention” and “a ranking requirement” is where a remarkable amount of SEO consulting seems to become creatively flexible.
GEO is useful as a concept, dangerous as a magic trick
Generative Engine Optimisation, or GEO, is not a meaningless idea.
Microsoft itself now uses the term when discussing how web content participates in AI-generated experiences, citations and grounding. Bing Webmaster Tools has also introduced AI performance reporting that can help publishers understand how their pages contribute to AI answers.
So there is clearly a new layer of visibility worth measuring.
Where we become sceptical is when GEO is sold as a completely separate discipline with a bag of proprietary tricks that somehow replaces good SEO.
If the advice is to make your site crawlable, structure information clearly, publish original material, keep facts current, establish who you are, use appropriate structured data and earn genuine references from elsewhere on the web, we are largely describing very good SEO.
The destination has expanded.
The foundations have not disappeared.
Links and mentions still mean something
The web is a network of relationships.
Links remain one of the clearest ways those relationships are expressed. A legitimate organisation linking to a supplier, a publication citing an expert, a project partner naming a developer or a customer crediting the company that built its site all create context beyond what a business says about itself.
AI-generated answers also introduce a more visible version of this idea through citations and supporting links.
That does not mean rushing out to buy 5,000 “AI authority backlinks” from somebody whose email begins with “Dear respected website owner”.
It means genuine mentions still have value because independent sources help establish relationships, reputation and provenance across the web.
What we think will still matter in 2027
Search will keep changing. Interfaces will change. The balance between traditional results, generated answers, multimodal search and AI agents will change.
Some tactics will disappear with them.
But it is difficult to imagine a useful search system deciding that the following things have become undesirable:
Websites that can be crawled. Pages with clear purposes. Accurate information. Original experience. Strong internal relationships. Fast, usable interfaces. Structured information that matches reality. Genuine evidence. Real businesses with consistent identities. Useful references from other sources.
Those are not tricks designed for one algorithm.
They are characteristics of information that is easier to discover, understand and trust.
SEO is becoming less about fooling machines
That may be the biggest change of all.
SEO once had a much larger industry built around finding mechanical weaknesses: repeat this phrase, manufacture these links, create hundreds of thin pages, adjust this density, exploit this feature until the search engine learns to hate it.
Modern search systems are far from perfect, and there will always be people looking for shortcuts.
But AI search pushes the useful part of SEO in a healthier direction.
Make the site technically accessible. Explain things clearly. Connect related information. Demonstrate actual expertise. Keep facts current. Give machines structured clues where standards support them. Build a reputation that exists beyond your own claims.
SEO used to be about convincing a search engine that you knew something. Increasingly, it is about making it exceptionally easy for machines to understand what you genuinely know.
That is considerably harder to fake.
It is also considerably more useful to everybody else.
Sources checked for this article include current Google Search Central guidance for AI features, Google's 2026 Search documentation updates, and current Microsoft Bing Webmaster guidance on AI search, sitemaps, IndexNow and AI Performance reporting. Search and AI features continue to evolve, so technical requirements should always be checked against current platform documentation.
