The Future of SEO and AI: What the Evidence Actually Shows

In February 2024, Gartner predicted that traditional search engine volume would fall 25% by 2026. That single line has anchored more agency decks, conference talks and panicked board meetings than any other claim in our industry. It is now 2026, so the prediction is checkable.

It did not happen. Alphabet’s second-quarter 2026 results put Google Search and other revenue at USD 63.3 billion, up 17% year on year, on consolidated revenue of USD 119.8 billion. Sundar Pichai’s shareholder message describes AI features “driving an incremental increase in Search queries overall”, says AI Mode has “surpassed 1 billion monthly active users” since its global expansion, and states plainly that Google is “now sending billions of clicks to websites every week through AI features in Search”. Search usage hit an all-time high during this year’s World Cup.

So the volume forecast failed. But anyone who has watched their own analytics since 2024 knows something did change. Both things are true at once, and getting the difference right is the whole of planning for the next two years. This guide sets out what the measurable evidence supports, what it does not, and what we would actually put in a Singapore marketing budget on the strength of it.

The forecast that shaped the panic, against the record

It is worth laying the widely repeated claims next to what can actually be checked, because the gap is the reason so many businesses have made bad decisions.

The pattern is consistent: the apocalyptic forecasts have not held, and the specific, measured behavioural changes have. That is not a comforting story or a frightening one. It is a story about where attention moves inside a search result, which is a much more useful thing to plan around than a headline number.

What genuinely changed: ranking and being cited came apart

The most important finding of the last twelve months is not about traffic volume at all. It is that the link between ranking well and being quoted in an AI answer has weakened sharply.

Ahrefs analysed 863,000 keyword SERPs and 4 million AI Overview URLs in March 2026. Of the pages cited in AI Overviews, 38% also rank in the top 10 organic results. The rest split almost evenly: 31.2% come from positions 11 to 100, and 31.0% come from outside the top 100 entirely. In their July 2025 run of the same study, roughly 76% of AI Overview citations came directly from the SERP being shown. Within eight months that halved.

Where AI Overview citations come from, and how fast the sourcing changed between July 2025 and March 2026.

The mechanism behind that shift is documented by Google itself. Its guidance for site owners says AI Overviews and AI Mode “may use a ‘query fan-out’ technique — issuing multiple related searches across subtopics”. Your page is not being matched against the query the user typed. It is being matched against a set of sub-questions the system generated on the user’s behalf, each of which runs its own retrieval. That is why a page sitting at position 40 for the headline term can be cited: it happens to be the best answer to one of the sub-questions.

This has a practical consequence that most rank-tracking dashboards still hide. A page can gain visibility while its tracked position stays flat, and it can lose visibility without dropping a single place. If your reporting still leads with average position, it is measuring a thinner slice of reality than it used to. Our guide to separating vanity metrics from decision metrics covers how to restructure that reporting, and how to show up in AI Overviews goes deeper on the fan-out mechanism itself.

The traffic question, answered honestly

The single most contested question in the industry right now is whether AI-referred traffic is a meaningful business channel yet. The honest answer is that it is small and disproportionately valuable, and that both halves of that sentence are routinely overstated by whoever benefits.

On size: Conductor’s 2026 benchmark, drawn from 13,770 domains and 3.3 billion sessions, put AI referral traffic at an average of 1.08% of sessions, with 87.4% of it coming from ChatGPT and IT and technology sites seeing the highest share at 2.8%. One per cent is not a channel you restructure a business around. It is also not nothing, and it is growing quickly from a base of effectively zero two years ago.

On value, the evidence genuinely conflicts. Semrush reported that AI search visitors convert at roughly 4.4 times the rate of traditional organic visitors — but that figure came from 500-plus high-value digital marketing and SEO topics, a category where AI adoption is unusually high and buying intent is unusually concentrated. Amsive’s statistical analysis of LLM traffic against organic search found no significant difference at all, reporting a p-value of 0.794. Conductor separately found AI-referred sessions lasting 2.3 times longer than search sessions.

What we take from this: do not budget on anyone else’s conversion multiple. Segment your own AI-referred traffic, watch it for a quarter against your real conversion events, and make the call on your own numbers. The GA4 reports worth checking and our conversion tracking guide cover the setup; the trap is that Google’s own AI surfaces are excluded from GA4’s AI Assistants channel and land in Organic Search instead, so “AI traffic” in GA4 means chatbot referrals only.

The economics underneath: crawling stopped being a fair trade

There is a structural story beneath the traffic numbers that gets far less airtime, and it will shape the next two years more than any ranking factor.

Cloudflare publishes a crawl-to-refer ratio, which it defines as a measure of “how often a given AI model sends traffic to a site relative to how often it crawls that site”. In the first week of August 2025, measured across its network, Anthropic’s ratio was near 50,000 to 1, OpenAI’s was 887 to 1 and Perplexity’s was 118 to 1. Filtered to news and publications sites, the same three read 2,500 to 1, 152 to 1 and 32.7 to 1. For comparison, a conventional search crawler operates in the single digits.

Crawl-to-refer ratios by platform, all sites and news publishers, from Cloudflare’s network measurements.

The direction of travel is clear in Cloudflare’s 2026 reporting: 52% of crawler requests were for AI training as of June 2026, up from 22% in spring 2025, with mixed-purpose crawlers accounting for more than 36% of activity. More than half of internet traffic is now non-human, and Cloudflare’s own framing is that “for every hour spent online searching for information, only 15 minutes is spent on the open web”.

For a Singapore SME this is not an abstract publisher grievance. It is the reason two decisions now sit on your desk that did not exist in 2023: which AI crawlers you allow, and whether you are comfortable that most of the value you produce is consumed without a visit. The critical detail, which most coverage gets wrong, is that training crawlers and retrieval crawlers are different agents with different consequences. Blocking GPTBot changes what a model learns; blocking OAI-SearchBot removes you from ChatGPT’s search answers entirely. We work through that split in how to rank on ChatGPT.

Five things that are safe to plan on

Strip out the forecasting and a short list of well-evidenced statements remains. These are what we would actually build a two-year plan on.

1. Indexing and snippet eligibility remain the gate. Google’s guidance is unambiguous: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”, and to be eligible “a page must be indexed and eligible to be shown in Google Search with a snippet”. Everything downstream depends on that. If your crawl budget, rendering or internal linking is broken, no AI strategy rescues it — start with technical SEO basics.

2. Off-site brand presence outweighs links for AI visibility. Ahrefs measured 75,000 brands against AI Overview brand mentions using Spearman correlations. Branded web mentions scored 0.664 and branded anchors 0.527, against 0.218 for raw backlink counts and 0.295 for referring domains. Ahrefs are explicit that correlation is not causation, and about 26% of the brands studied had no AI Overview mentions at all. Treat it as a strong hint that being talked about matters more than being linked to, not as a formula.

3. The research query is at risk; the transactional query mostly is not. Whitespark’s study of 540 manual local queries found AI Overviews on 92% of informational searches but only 15% of local-intent ones, where a local pack appeared 93% of the time. The content that fed your funnel is being summarised. The query where someone is ready to call you is largely intact. That is a content problem, not an existential one — see AI local SEO and local SEO in Singapore.

4. Measurement improves, but stays partial. Google announced Search generative AI performance reports for Search Console in June 2026, separating AI Overview and AI Mode impressions from standard organic data for the first time. They launched as a limited rollout showing impressions, pages, countries and devices — no clicks, no click-through rate, no query data. That is progress, and it is still not enough to attribute revenue. Plan for partial visibility as the steady state, not a temporary annoyance.

5. There is no separate technical specification to buy. Google states that “You don’t need to create new machine readable files, AI text files, or markup to appear in these features”. Ahrefs tested this directly on 1,885 pages that added JSON-LD between August 2025 and March 2026 against roughly 4,000 matched controls and found no citation lift, with a small negative effect on AI Overviews. Schema is still worth doing for rich results. It is not an AI visibility lever, whatever the deck says.

What this means for a Singapore marketing budget

Singapore’s position is specific enough that importing an American plan will mislead you. Google holds 92.46% of search referrals here as of July 2026, with Bing at 3.37%. Meanwhile Stanford HAI’s 2026 AI Index places Singapore near the top of global consumer generative-AI adoption at roughly 61%, against about 53% globally. Consumers here use assistants heavily, but the front door to the open web is still overwhelmingly Google. That combination argues for prioritising AI Overviews and AI Mode visibility well ahead of chatbot citation work — the opposite of the emphasis in most imported playbooks.

Market rates for SEO retainers in Singapore run broadly from the high hundreds to several thousand dollars a month depending on scope; we set out how that breaks down in our guide to SEO cost in Singapore. The relevant point for this discussion is that the AI shift should change what sits inside that budget, not add a second invoice beside it. If you want the commercial detail on how we approach this surface, the AI SEO service page sets it out, and the case studies show what the underlying SEO work has produced for other Singapore businesses.

What we would stop doing right now

Four habits have quietly become liabilities.

• Leading a report with average position. It now describes a shrinking share of how you are found. Lead with qualified sessions, assisted conversions and revenue; keep rankings as a diagnostic.
• Buying “GEO” as a separate line. If a proposal cannot name a task inside it that is not already SEO, it is a repackaging exercise. Our breakdown of GEO, AEO and LLM SEO sets out where the terms genuinely differ and where they do not.
• Publishing to a monthly quota. The content most exposed to summarisation is precisely the content produced to fill a slot. Fewer, harder pieces beat more, thinner ones on every surface now.
• Quoting a single vendor index as fact. Most circulating AI-search statistics come from tools vendors measuring their own users. Use the direction, name the source, and never build a forecast on one.

The job is changing more than the discipline

The honest summary is that SEO in 2027 will be recognisably the same discipline with a different centre of gravity. Crawling, indexing, information architecture and content quality all still decide whether you are eligible. What moves is where the leverage sits: away from on-page tuning for a tracked keyword, toward being the source that a synthesis engine has reason to quote, and toward measurement work that can prove any of it happened.

That reallocation is uncomfortable because the new work is harder to invoice against a number. It is much easier to sell a ranking report than a quarter of earned mentions and cleaner attribution. But the businesses we see doing well on AI surfaces in Singapore are not running a separate AI programme. They are running good SEO with more original substance, better off-site presence and honest reporting — which is what the case for SEO rested on before any of this started.

Frequently asked questions

Is SEO dying because of AI?

No, and the specific prediction most often quoted has been falsified. Gartner forecast in February 2024 that traditional search volume would fall 25% by 2026; Alphabet’s Q2 2026 results instead describe record Search usage, AI features increasing total queries, and Search revenue up 17% year on year. What has genuinely changed is click behaviour on individual results and how citations are selected, not the existence of search.

How much traffic actually comes from AI assistants right now?

Conductor’s 2026 benchmark across 13,770 domains and 3.3 billion sessions put AI referral traffic at an average of 1.08% of sessions, with 87.4% of that coming from ChatGPT and IT and technology sites highest at 2.8%. It is small but growing fast. Segment it in your own analytics rather than relying on an industry average.

Do I need to rank in the top 10 to be cited by AI Overviews?

No. Ahrefs found in March 2026 that only 38% of pages cited in AI Overviews rank in the top 10, with 31.2% coming from positions 11 to 100 and 31.0% from outside the top 100. Google’s query fan-out technique matches pages against generated sub-questions rather than the original query, which is why non-ranking pages can be cited.

Should I pay an agency extra for AI SEO or GEO?

Not as a separate retainer on top of SEO. Google’s own documentation states there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode. The work that improves AI visibility is content depth, entity clarity, technical health and earned brand mentions, all of which belong inside a competent SEO engagement.

What should a Singapore business prioritise for AI search?

Google AI Overviews and AI Mode, ahead of chatbot citation work. Google carries 92.46% of Singapore search referrals as of July 2026, so the AI surface most of your buyers meet is still inside Google. Chatbot visibility is worth building, but it should follow rather than lead here.

Does adding schema markup help me get cited by AI?

The available evidence says no. Ahrefs studied 1,885 pages that added JSON-LD between August 2025 and March 2026 against roughly 4,000 matched controls and found no citation lift, with a small negative effect on AI Overviews. Google separately states you do not need new markup or AI text files. Keep schema for rich results, not as an AI visibility tactic.

Where this leaves you

The next two years do not require a new discipline. They require dropping four habits, funding two things more heavily, and refusing to pay twice for one job. Search volume is growing, clicks per result are falling, citations have detached from rankings, and measurement is improving slowly. Plan against those four facts and you will be ahead of most of the market, which is still planning against a forecast that did not come true.

If you want a view on where your own site sits across these surfaces before committing budget, that is the first thing we look at in an AI SEO engagement. Start with the AI SEO guide for Singapore for the full picture, or the tools worth using if you would rather run the diagnostic yourself.

If the question is not what is coming but what to do about it this year, the AI SEO strategy guide turns all of the above into a twelve-month sequence with a budget and a stop condition.

Written by Adrian Tan, Singapore Digital Marketing. Last updated 5 August 2026.

Related guide

Every figure used above, along with the ones we checked and deliberately left out, is collected with its sample size and measurement date in our AI SEO statistics reference.