AI SEO vs Traditional SEO: What Actually Changes in the Work
Two claims dominate this conversation, and both are wrong. The first is that nothing has changed — that AI search is a rebrand, the fundamentals still apply, carry on. The second is that everything has changed — that SEO is dead, a new discipline has replaced it, and you need a separate strategy, a separate budget and a separate agency to survive.
The honest position is narrower and more useful than either. The work is largely the same. The payoff curve has moved sharply. Those two facts together explain almost everything you need to decide, and they point to a handful of genuinely new tasks rather than a wholesale replacement.
This article is the practical comparison: a job-by-job look at what an SEO actually does in a week, and what is different about each job now. It is not a definition piece. If you want to know what generative engine optimisation means, read what is generative engine optimization. If the acronyms are the problem — GEO, AEO, LLMO, AIO — that is untangled in GEO vs AEO vs LLM SEO. This post assumes you already know roughly what the terms mean and want to know what to do differently on Monday.
Start with what Google actually says
Most of the market’s confusion disappears once you read Google’s own guidance on optimising for its generative AI features, published in its Search Central documentation. The key sentence is not hedged:
“In short, yes! The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems.”
On eligibility, Google is equally direct: “To be eligible to be shown in generative AI features on Google Search, a page must be indexed and eligible to be shown in Google Search with a snippet, fulfilling the Search technical requirements.”
And on the special-treatment question that the market keeps trying to sell you:
“Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.”
“You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.”
“There’s no requirement to break your content into tiny pieces for AI to better understand it.”
Take those at face value and one thing follows immediately: the entry ticket to AI search is ordinary technical SEO. Indexable, crawlable, snippet-eligible. If a page fails there, no amount of AI-specific optimisation reaches it. Anyone selling you a GEO retainer before your site is properly indexed has the order wrong.
A caveat worth stating: this is Google’s position on Google’s surfaces. ChatGPT, Perplexity and Claude have their own retrieval stacks and their own crawlers, and Google does not speak for them. But Google is where roughly 92.46% of Singapore searches happen — StatCounter’s July 2026 figure — so its position is the one that governs most of the outcome locally.
So why does it feel like everything changed?
Because the reward for doing the same work well has fallen, and that is a genuinely big deal even when the work is unchanged.
Ahrefs re-ran its click-through study and published the update on 4 February 2026, comparing December 2023 with December 2025 across 300,000 keywords — 150,000 with an AI Overview present, 150,000 informational keywords without one. The position-one click-through rate fell by 58% when an AI Overview was present. Their earlier run of the same study, comparing March 2024 with March 2025, had put the figure at 34.5%. The gap widened.
Read that chart carefully, because the shape matters more than the headline. The losses concentrate at the top. Position one loses 58%; position ten loses 19.4%. AI Overviews are not flattening the results page evenly — they are compressing the premium that used to attach to first place.
Now the counterweight, because a single scary number is how bad strategy gets sold. SparkToro’s analysis of Similarweb clickstream data covering US Google searches from January to April 2026 found 68.01% of searches ended without a click, but also that only 0.34% of searches transitioned into AI Mode. The same dataset produces the alarming figure and the deflating one. Zero-click behaviour is overwhelmingly the old kind — answers in the SERP, refinements, abandonment — not people migrating to a chatbot interface. The study’s authors note it excludes Google’s mobile search app, and cross-study comparisons are not equivalent.
Add one more data point: Alphabet’s Q2 2026 results reported “Google Search and other” revenue of USD 63.3 billion, up 17%. A business collapsing under AI disruption does not post that. What the numbers describe together is a surface that is still enormous, still growing commercially, and paying out less per ranking than it used to.
The job-by-job diff
Here is the comparison that actually helps. Ten things an SEO does, and what is different about each.
| The job | What is unchanged | What genuinely shifts |
|---|---|---|
| Keyword research | Demand still comes from a keyword tool. Volume, difficulty and intent still drive prioritisation. | You add a layer of conversational prompts you cannot get volume for. Treat them as a qualitative panel, not a keyword list. |
| Content planning | Topic clusters, one page per intent, internal linking to the money page. | Sub-questions matter more. Google defines query fan-out as “a set of concurrent, related queries generated by the model”, so a page that answers only the headline question is thinner than it looks. |
| Writing | Original, specific, experienced-based content still wins. Generic content still loses. | Self-contained passages beat flowing narrative. A section that needs three earlier sections to make sense is hard to extract and quote. |
| On-page | Titles, headings, internal links, a clear primary topic per page. | Headings framed as real questions, with the answer in the first two sentences below. Cosmetic in 2019, load-bearing now. |
| Technical SEO | Crawlability, indexation, speed, rendering. Unchanged in substance. | Raised in priority, because it is now the literal eligibility gate for AI features as well as blue links. |
| Structured data | Still worth doing for rich results in classic Search. | Not a route to AI citations. Google says so directly, and the independent evidence agrees (see below). |
| Internal linking | Distributes authority, defines site structure. | Largely unchanged. One of the most stable parts of the job. |
| Links and PR | Earned mentions from credible sources still signal authority. | Unlinked brand mentions gain value, because retrieval systems read text about you as well as links to you. |
| Measurement | Sessions, conversions, revenue. Still the point. | Materially harder. Rankings decouple from clicks; some AI referrals arrive with no referrer and land in Direct. |
| Reporting | Business outcomes, not vanity metrics. | Impressions rise while clicks fall, so an unexplained report now looks like failure even when the work is going well. |
Look down the middle column. Most of it says “unchanged”. That is the finding, and it is why the “you need a separate discipline” pitch does not survive contact with the actual task list.
The four things that are genuinely new work
Not marginal adjustments — these are tasks that did not exist in a 2019 SEO plan.
1. Off-site entity presence. Retrieval systems assemble answers from many sources, and the sources they favour skew heavily towards a small set of high-authority domains and community platforms. Ahrefs’ March 2026 analysis of roughly four million AI Overview URLs found 88% of AI Overviews cite three or more sources, and that the top 1% of domains — about twelve sites — take 47% of citations. Being discussed on the platforms that get cited is now part of the job, and it looks more like digital PR than like SEO.
2. Passage-level answerability. Auditing whether each section of a page stands alone as an answer is a new, concrete review task. It is cheap, it is checkable, and almost nobody does it.
3. Citation and mention measurement. Running a fixed panel of buyer questions across assistants on a schedule and recording whether you appear. This is genuinely new infrastructure, and it is noisier than most vendors admit — how to read it without fooling yourself is covered in AI brand visibility monitoring.
4. Crawler policy. Deciding which AI crawlers may access your site is a decision no one had to make before, and the distinction that matters — training crawlers that send nothing back versus retrieval crawlers that drive referrals — is the part most sites get wrong. The detail is in llms.txt and AI crawlers.
The three things people say changed but have not
Schema does not buy citations. This is the most expensive misconception in the category. Ahrefs studied 1,885 pages that added JSON-LD between August 2025 and March 2026, matched against roughly 4,000 control pages, and found no citation lift on AI Overviews, AI Mode or ChatGPT — and a small but statistically significant -4.6% on AI Overviews. Google’s own documentation says structured data is not required. Keep schema for rich results; stop selling it as an AI-visibility tactic.
llms.txt does nothing for AI search. Google states you do not need AI text files, and no major AI provider has committed to reading it in production for search. It has real uses for developer and agent tooling. It is not an AI-search tactic.
“Writing for AI” is not a distinct style. There is no keyword density for language models, no phrasing that flatters a retrieval system. Google explicitly says there is no requirement to break content into tiny pieces. What helps — clear questions, direct answers, specific facts, sources — is what helped human readers already.
What this means for a Singapore business
Three local facts change the emphasis.
Google is the surface that matters. At 92.46% search share in July 2026, AI Overviews and AI Mode reach far more Singaporean buyers than every chatbot combined. If you are prioritising, the Google surfaces come first — which conveniently is also where ordinary SEO already helps you. The detail on that surface sits in AI Overviews in Singapore and Google AI Mode.
Consumer adoption is high; workplace habit is not. Stanford HAI’s 2026 AI Index puts Singapore near the top of global consumer generative-AI adoption at around 61%, against roughly 53% globally. Salesforce research published on 8 July 2026 found Singapore workers among the world’s least AI-sceptical, yet with only about 6% using AI daily at work. If you sell B2C, assistant usage among your buyers is genuinely high. If you sell B2B, the buying committee is probably still on Google.
The local market is small enough that authority compounds. The concentration finding cuts both ways: if a narrow set of sources dominates citations, becoming one of them in a defined Singapore niche is achievable in a way that competing globally is not.
How to reallocate without gambling
A defensible split, given the evidence above:
- Most of the effort stays where it was. Technical health, genuinely useful content, internal linking, earned mentions. This is the entry ticket for both surfaces, and Google’s guidance says so explicitly.
- A modest slice moves to answerability and structure. Rewriting headings as questions, front-loading answers, breaking multi-question pages apart. This is cheap and it improves pages for readers too, so the downside is close to zero.
- A small, capped slice funds measurement and off-site presence. A monitored prompt panel, plus real participation wherever your buyers actually discuss the category.
- Nothing is spent on schema-for-citations, llms.txt-for-search, or “AI-optimised” rewriting. The evidence does not support any of them.
Market rates for this work look much like SEO rates, because it largely is SEO — the benchmarks in SEO cost in Singapore remain the right reference. Be sceptical of a GEO premium attached to a task list that is 70% ordinary SEO. If you want the budgeting and sequencing version of this decision, that is AI SEO strategy.
The measurement problem you should plan for
One consequence deserves its own note, because it is where good work gets misread. When AI Overviews appear above you, impressions can hold or rise while clicks fall. On the old dashboard that looks like decline. It may be the same visibility paying out less.
Two adjustments help. First, separate the surfaces in reporting so an AI-driven click loss is visible as such rather than blamed on rankings — the mechanics are in tracking AI search traffic. Second, move the headline metric down the funnel. Enquiries and revenue are unaffected by how the click was counted, which is a good reason they should have been the headline all along. The general setup lives in GA4 setup and attribution models.
The honest summary
AI SEO versus traditional SEO is not a choice between two disciplines. It is the same discipline operating against a results page that keeps more of the value than it used to. Google says its generative features run on its core ranking systems and that no special files or markup are needed. The independent evidence agrees on the negative claims — schema does not buy citations, llms.txt does nothing for search — and shows a real, large fall in what a top ranking pays.
So: do the SEO. Do it better than you did, because the margin for a mediocre page is thinner. Add the four genuinely new tasks. Change how you measure, because the old dashboard now misleads. Ignore the rest, and ignore anyone who cannot tell you which of the three columns above their proposal falls into.
If you want that assessment done on your own site — which pages are eligible, which sections fail an answerability check, and where your visibility actually stands — that is what our AI SEO service in Singapore does. The results we report are in our case studies.
Where to go next
- AI SEO in Singapore — the pillar guide to the whole topic
- What is generative engine optimization — the definition and the research behind it
- SEO for AI search — the practical optimisation layer
- AI SEO statistics — every figure in this debate with its sample and date
- Is SEO worth it in Singapore? — the return question, answered with numbers
Frequently asked questions
Is AI SEO replacing traditional SEO?
No. Google’s own guidance states that SEO best practices continue to be relevant because its generative AI features are rooted in its core Search ranking and quality systems, and that a page must be indexed and snippet-eligible to appear in them at all. What has changed is the payoff: an Ahrefs study published in February 2026 found position-one click-through down 58% when an AI Overview is present. Same work, lower yield, plus a small number of genuinely new tasks.
What is the main difference between AI SEO and traditional SEO?
Traditional SEO optimises for a ranked list of links; AI SEO also optimises for being extracted and cited inside a generated answer. In practice the difference shows up in three places: passages must stand alone rather than depend on the rest of the page, off-site mentions matter more because retrieval systems read text about you as well as links to you, and measurement has to account for visibility that produces no click.
Do I need a separate GEO strategy and budget?
Usually not a separate one. Most of the task list overlaps with SEO, so a separate strategy tends to duplicate work and a separate budget tends to fund tactics the evidence does not support. What is worth carving out is a small, capped allocation for citation measurement and off-site presence. Be cautious with any proposal charging a premium for a plan that is mostly ordinary SEO.
Does schema markup help you get cited by AI?
The evidence says no. Google states that structured data is not required for generative AI search and that no special schema.org markup is needed. Ahrefs examined 1,885 pages that added JSON-LD between August 2025 and March 2026 against roughly 4,000 control pages and found no citation lift, with a small but statistically significant 4.6% negative effect on AI Overviews. Schema remains worth implementing for rich results in classic Search.
If AI Overviews reduce clicks, is ranking still worth it?
Yes, for two reasons. Being indexed and snippet-eligible is the stated eligibility requirement for appearing in Google’s generative features, so ranking well is the route into them rather than an alternative to them. And the scale has not collapsed: Alphabet reported Google Search and other revenue of USD 63.3 billion in Q2 2026, up 17%. The realistic planning assumption is lower clicks per ranking, not the end of search traffic.
How is AI SEO different in Singapore?
Google’s share of Singapore search was 92.46% in July 2026, so AI Overviews and AI Mode reach far more local buyers than chatbots do, and the Google surfaces should be prioritised first. Consumer AI adoption is high — Stanford HAI’s 2026 AI Index places Singapore near the top globally at around 61% — but Salesforce research published in July 2026 found only about 6% of Singapore workers using AI daily at work, so B2B buying journeys still run largely through conventional search.
